Australian Journal Of Physiotherapy

274 Leemans et al: Cost-effectiveness of respiratory physiotherapy in COPD Records identified through Additional records identified through other sources (n = 5) database searching (n = 530) • Medline/PubMed (n = 142) • Web of Science (n = 235) • Wiley Online Library (n = 97) • NHS EED (n = 41) • PEDro (n = 15) Records after duplicates removed (n = 435) Records screened Records excluded (n = 435) (n = 376) Full-text articles assessed Full-text articles excluded for eligibility (n = 59) (n = 48) Studies included in • population (n = 1) qualitative synthesis • intervention (n = 25) (n = 11) • comparison (n = 0) • outcome (n = 3) • study design (n = 19) Studies included in quantitative synthesis (meta-analysis) (n = 0) Figure 1. Flow of studies through the review. programs took place in different healthcare settings: inpatient, to evaluate to what extent these techniques were put into practice. outpatient, community-based/primary care, both inpatient and Looking to ACTs, two studies examined assisted cough and self- outpatient, a combination of outpatient and primary care and home- management/education in ACTs.22,23 A combination of several ACTs based alone (one study each). Second, the duration of the exercise (body positioning, manual percussion, vibration, active cycle of programs varied from short (, 12 weeks, five studies) to a longer breathing techniques) and breathing exercises were reported once. duration (. 12 weeks, four studies) with a minimum of 7 days to a Interestingly, ACTs were only investigated in clinical trials that maximum of 24 months. Third, heterogeneity in training frequency, recruited COPD patients during an acute exacerbation.22,23 intensity and supervision of sessions was observed, ranging from one supervised session per 1 to 3 months for a maintenance program, to Outcomes two to three supervised sessions per week in combination with one The reported outcomes to assess the effect of the interventions to three additional unsupervised sessions at home to enhance behavioural change towards physical activity. Four out of nine trials were general health status (k = 10), disease-specific QOL (seven with exercise programs specified their exercise intensity, which was studies), exercise capacity (two studies), total number of COPD ex- considered high intensity according to the recent guideline on PR.9 In acerbations (one study) and daily physical activity (one study). the study of Boland et al, no duration or training frequency was re- Overall health-related questionnaires were used to derive general ported.20 Fourth, the type of exercise training included either health status expressed as QALYs.28,29 To calculate QALYs, the ma- endurance (one study), endurance and strength training (seven jority of the selected studies used the EQ-5D questionnaire to studies) or was not specified (two studies). The intervention pro- calculate the utilities (seven studies). Three studies derived utility grams also differed in terms of patient education offered, from basic scores for health status by extracting responses from the 36-Item advice by information sheets (one study) to extensive self-manage- Short Form Health Survey (SF-36) and using them to complete a ment programs (nine studies). Regarding the therapeutic methods for six-item health state classification, the SF-6D.21,25,30 The selected increasing physical activity, interventions were activity counselling, outcome measures for disease-specific QOL across studies were theory-based behavioural medicine interventions, training logs, ex- heterogeneous, including St. George Respiratory Questionnaire ercise programs at home, or a combination. However, it was difficult (SGRQ),20,22,31 Chronic Respiratory Questionnaire (CRQ)19,24,32 and the Clinical COPD Questionnaire (CCQ).23 Change in exercise capacity

Table 1 Setting Participants Characteristics of included studies.  Netherlands  COPD based on GOLD guidelines Study  Healthcare and societal  Con: n = 532, age = 68.4 (SD 11.1), Boland et al. 201520 perspective FEV1 = 67.9 (SD 20.5)  Cluster RCT, time horizon 24 mths  2013 Euros  Int: n = 554, age = 68.2 (SD 11.3),  CEA/CUA FEV1 = 67 (SD 20.3) Burge et al (2020)21  Australia  Stable COPD based on FEV1/FVC  Equivalence RCT, 12-mth time  Healthcare and societal ratio of , 70% horizon perspective  Smoking history (current or  CEA/CUA  2017 Australian dollars former) of minimum 10 pack- years. Burns et al. 201624  United Kingdom  Parallel investigator-blind RCT,  Healthcare perspective  Con: n = 77, age = 69 (SD 9), FEV1 =  2012-2013 British pounds sterling 49 (SD 20) 12-mth time horizon  CEA/CUA  United Kingdom  Int: n = 82, age = 71 (SD 9), FEV1 =  Healthcare perspective 53 (SD 18) Cross et al. 201022  2007-2008 British pounds sterling  Equivalence RCT, 6-mth time  Stable COPD based on FEV1 , 70% predicted horizon  CEA/CUA  smoking history of . 20 pack- years   60% attendance rate of standard PR course  Con: n = 75, age = 69.3 (SD 8.9), FEV1 = N/A  Int: n = 73, age = 67.3 (SD 15.1), FEV1 = N/A  COPD based on FEV1 , 80% pre- dicted and FEV1/FVC , 0.7  Acute exacerbation  Con: n = 264, age = 69.6 (SD 9.5), FEV1 = N/A  Int: n = 258, age = 69.1 (SD 9.9), FEV1 = N/A

Control Intervention Outcome measures Research 275 Usual care, based on 2007 national primary care COPD guidelines. Integrated Disease Management  Costs: (IDM)41: Multidisciplinary teams B Intervention Centre-based, outpatient group- consisted of at least three members: B Healthcare resources (eg, based supervised42 PR course with GP, practice nurse and PT with medication prescriptions) components of exercise training and specific certified training in COPD B Travel expenses self-management education. care. Elements of this program were B Productivity loss Dose: 8 weeks, 5 sessions/week with proper diagnosis, optimal medical 2/week supervised sessions/week. adherence, motivational  Effectiveness: SGRQ, total number interviewing, smoking cessation, of COPD exacerbations, CCQ Standard care: encouragement to self-management plans, dietary continue exercises at the conclusion interventions, and guideline-based  Utility: QALY (EQ-5D) of the initial 2-mth PR course and physiotherapeutic reactivations. Follow-up: 0, 6, 9, 12, 18 and 24 attend a local support group for Each team designed their own time- mths. Exacerbations continuously people with lung conditions. contingent individual practice plan over specified time horizon. (no standardisation). Dose: an Airway clearance technique: advice individual patient-specific care plan  Costs: on positioning, cough and sputum negotiated in collaboration with B Healthcare resources (eg, mobilisation in accordance with patient. Intensity and number of IDM staffing) ACBT. This information was elements depended upon health B Personal out-of-pocket reinforced by providing an status and patient’s need. expenses information sheet with the advice. Home-based PR course42: first week a PT home visit to establish exercise  Effectiveness: D6MWD (m) goals and supervise first exercise  Utility: QALY (SF-6D) session followed by seven once-  Comorbidities weekly structured telephone calls Follow-up: 0, 8 weeks, 12 mths from a PT, using structured modules  Costs: and motivational interviewing. Dose: 8 weeks, 5/week with 1/week B Healthcare resources (eg, follow-up session by call. staffing) Low-intensity maintenance course: 2 hr maintenance session at 3, 6 and 9 B Social services mths after completion of standard PR B Out-of-pocket expenses course, comprising 1 hr of education B Productivity loss and 1 hr of structured exercise in  Effectiveness: CRQ addition to standard care. Patients  Utility: QALY (EQ-5D-3L) received an individually tailored Follow-up: 0, 3, 6, 9 and 12 mths exercise prescription, to be undertaken at home, which was  Costs: reviewed at each session and B Healthcare resources (eg, modified as appropriate. staffing) Airway clearance technique: MCP B Personal social services during thoracic expansion exercises and vibration during expiration.  Effectiveness: SGRQ: symptoms, Treatment interspersed with periods activity, impact and total score of relaxed abdominal breathing and FET in accordance with ACBT.  Utility: QALY (EQ-5D) Following MCP, the PT provided the Follow-up: 0, 6 weeks and 6 mths patient with advice on positioning, with ACBT. This information was reinforced by providing an information sheet with the advice.

Table 1 (Continued) Setting Participants Study  Ireland  COPD based on GOLD guidelines. Gillespie et al. 201319  Healthcare perspective  Con: n = 172, age = 68.4 (SD 10.3),  Cluster RCT, 5.5-mth  2009 Euros time horizon FEV1 = 59.7 (SD 13.8)  CEA/CUA  Canada  Int: n = 178, age = 68.8 (SD 10.2),  Societal perspective Goldstein et al. 199732  1989 Canadian dollars FEV1 = 57.6 (SD 14.3)  RCT, 6-mth time horizon  CEA  Netherlands  Severe stable COPD defined by  Healthcare and societal FEV1 , 40% pred.43 Goossens et al. 201323  RCT, time horizon 3 mths perspective  Con: n = 44, age = 65 (SD 8), FEV1 =  CEA/CUA  2009 Euros 34.6 (SD 11.8) Griffiths et al. 200130  United Kingdom  Int: n = 45, age = 66 (SD 7), FEV1 =  RCT, time horizon 12 mths  Healthcare perspective 34.8 (SD 14.5)  CUA  2001 British pounds sterling  COPD based on GOLD guidelines  Acute exacerbation  Smoking history  10 pack-years  Con: n = 69, age = 67.80 (SD 11.30), FEV1 = N/A  Int: n = 70, age = 68.31 (SD 10.34), FEV1 = N/A  COPD with FEV1 , 60% pred.44  Con: n = 101, age = 68.3 (SD 8.1), FEV1 = 39.4 (SD 16.4)  Int: n = 99, age = 68.2 (SD 8.2), FEV1 = 39.7 (SD 16.2)

Control Intervention Outcome measures 276 Leemans et al: Cost-effectiveness of respiratory physiotherapy in COPD Standard care in Irish general practices Structured education PR course: 2-hr  Costs: group-based session each week for 8 B Healthcare resources (eg, GP Usual care: GP and/or respiratory weeks delivered jointly by practice visit) specialist with no special attention to nurse and PT. The PT focused on B Medication rehabilitation. delivering the exercise component. B Travel expenses Afterwards, participants were B Time input Early assisted discharge with home followed up formally via telephone care: all patients received usual care call at 4 weeks after completion of  Effectiveness: CRQ between days 1 and 3. The PT the course and via a 1-hr group  Utility: QALY (EQ-5D-3L) instructed the patient in breathing, session at 10 weeks. Follow-up: 0 and 5.5 mths coughing techniques and Supervised PR: 2 mths inpatient reactivation. The PT instructed the followed by 4 mths combination of  Costs: patient to follow the written outpatient and home exercises. The B Healthcare resources (eg, med- instructions at home. The previously inpatient activities included ical care, home-care services) described treatment was continued supervised exercise training, patient B Transportation costs at home and supervised by nurses.34 education and psychosocial support. Duration: 7 days During the next 4 mth, daily home  Effectiveness: CRQ Usual care: outpatient or primary exercise program with 3 to 4 home- Follow-up: 0 and 6 mths care follow-up. care visits and outpatient appointments for 4 to 5 supervised  Costs: exercise training by PT. B Healthcare resources (eg, GP Usual inpatient hospital treatment: visit) all patients received usual care B Non-healthcare costs (eg, un- between days 1 and 7. The PT paid domestic help) instructed the patient in breathing, B Productivity loss coughing techniques and reactivation.  Effectiveness: CCQ Duration: 7 days.  Utility: QALY (EQ-5D-5L) Follow-up: 0, 7 days and 3 mths Outpatient PR program: patients received rehabilitation on three half  Costs: days per week for 6 weeks. Each B Healthcare resources (eg, PR session lasted for approximately 2 service) hrs and included educational B Transport activities, exercise periods and psychosocial sessions. Individual  Utility: QALY (SF-6D) goal setting, dietary intervention, PT Follow-up: 0, 6 weeks and 12 mths and occupational therapy were also included. After the 6-week program, patients were invited to join a patient-run group meeting.

Table 1 (Continued) Setting Participants Study  Netherlands  COPD GOLD II-III  Societal and third-party payer’s  Impaired exercise capacity , 70% Hoogendoorn et al 201031  RCT, time horizon 24 mths perspective predicted  CEA/CUA  2007 Euros  Con: n = 88, age = 67 (SD 9), FEV1 = 58 (SD 17)  Int: n = 87, age = 66 (SD 9), FEV1 = 60 (SD 15) Waterhouse et al 201025  United Kingdom  COPD based on GOLD guidelines  RCT, 2x2 factorial design, time  Healthcare perspective  Con: n = 129, age = 69.1 (SD 7.5),  2010 British pounds sterling horizon 18 mths FEV1 = 48.3 (SD 19.3)  CUA  Int: n = 111, age = 68.7 (SD 8.3), FEV1 = 45.1 (SD 16.3) Zwerink et al 201626  Netherlands  COPD based on GOLD guidelines  RCT, time horizon 24 mths  Healthcare perspective   three exacerbations or one  CEA/CUA  2009 Euros hospitalisation respiratory prob- lems in 2 yrs preceding study entry  Con: n = 76, age = 64.1 (SD 7.7), FEV1 = 50.5 (SD 17.0)  Int: n = 77, age = 63.1 (SD 8.1), FEV1 = 49.6 (SD 14.2) ACBT = active cycle breathing techniques, CCQ = Clinical COPD Questionnaire, CEA = cost-effectiveness analysis, Con = control expiration technique, FEV1 = forced expiratory volume in 1 second (% predicted), GP = general practitioner, IDM = integrated d not available in published manuscript, PR = pulmonary rehabilitation, PT = physiotherapist, QALY = quality-adjusted life year

Control Intervention Outcome measures Usual care: pharmacotherapy Interdisciplinary community-based  Costs: Research 277 according to accepted guidelines, a management course45: B Healthcare resources (eg, GP short smoking cessation advice 4-mth, supervised, intensive visit) session from the respiratory intervention: 2x/week individual B Non-healthcare costs (eg, un- physician and short nutritional exercise training sessions by PT. paid domestic help) advice. Patients were instructed and B Travel expenses motivated to perform exercises at B Productivity loss PR in hospital setting: home and to walk/cycle 2x/day. Acute: exercises training in Smoking cessation counselling,  Effectiveness: SGRQ, total number combination with education leading education, self-management skills of COPD exacerbations to a total of 2 hrs/session, two times/ provided by respiratory nurses. week, for 6 weeks. Patients were 20-mth active maintenance: patients  Utility: QALY (EQ-5D) encouraged to exercise between visited the PT 1x/mth to monitor Follow-up: 0, 4, 12 and 24 mths. formal classes. They kept an exercise exercise capacity and adherence to Exacerbations continuously during diary at home between sessions. the training and to provide 24 mths Long-term: telephone follow-up encouragement to continue the versus ‘standard follow-up’. An exercise training at home.  Costs: exercise booklet, individualised for PR in community setting: B Healthcare resources (eg, GP the level of exercise the participant Acute: exercises training in service) had achieved during the sessions, combination with education leading B Non-healthcare costs (eg, home was provided and the research to a total of 2 hrs/session, two times/ help) participants were encouraged to week, for 6 weeks. Patients were keep up with the booklet exercises. encouraged to exercise between  Utility: QALY (SF-6D) Intervention calls gave formal classes. They kept an exercise Follow-up: 0, 6 weeks, 6, 12 and 18 encouragement to exercise, not diary at home between sessions. mths general healthcare advice. Long-term: telephone follow-up Self-management program: first versus ‘standard follow-up’. An  Costs: mth, patients attended four 2-hr exercise booklet, individualised for B Healthcare resources (eg, self-management sessions. the level of exercise the participant medication prescriptions) had achieved during the sessions, was provided and the research  Effectiveness: ISWT, daily step participants were encouraged to count keep up with the booklet exercises. Intervention calls were orientated to  Utility: QALY (EQ-5D) giving encouragement to exercise,  Follow-up: 0, 7, 12, 18 and 24 mths not general healthcare advice. Community-based exercise program within a self-management program: first mth, patients attended four 2-hr self-management sessions. Afterwards followed by an exercise program by PT of private practices, consisting of a 6-mth ‘compulsory’ period (3 sessions/week) and subsequently a 5-mth ‘optional’ period (2 sessions/week). Since COPE-active was intended to change behaviour with regard to exercise, one session/week in both periods consisted of unsupervised home- based exercise training. l group, CRQ = Chronic Respiratory Questionnaire, CUA = cost-utility analysis, EQ-5D = EuroQol five dimensional, FET = forced disease management, Int = Interventional group, ISWT = incremental shuttle walk test, MCP = manual chest percussion, N/A = rs, RCT = randomised controlled trial, SGRQ = St. George Respiratory Questionnaire, 6MWD = 6-minute walk distance.

278 Leemans et al: Cost-effectiveness of respiratory physiotherapy in COPD Table 2 Critical appraisal of the included clinical trials, assessed using the JBI Critical Appraisal Checklist for Economic Evaluation.16 Study Q1 Q2 Q3 Q4 Q5 Q6 Q7 Q8 Q9 Q10 Q11 Boland et al 201520 111111 1 11 1 1 1 1 Burge et al 202021 111111 1 11 1 1 0 - Burns et al 201624 111 - 11 1 11 1 1 1 - Cross et al 201022 111 - 11 1 1 - 1 1 1 1 Gillespie et al 201319 1 1 1 1 1 1 N/A 1 1 1 1 1 - Goldstein et al 199732 111111 1 1 - 1 1 Goossens et al 201323 111111 1 11 Griffiths et al 200130 111111 0 11 Hoogendoorn et al 201031 1 1 1 1 1 1 1 11 Waterhouse et al 201025 1 1 1 1 1 1 0 1 1 Zwerink et al 201626 111111 1 11 1 = yes, - = inadequate methodology, 0 = insufficient information, N/A = not applicable. Q1 = Is there a well-defined question?, Q2 = Is there comprehensive description of alternatives?, Q3 = Are all important and relevant costs and outcomes for each alternative identified?, Q4 = Has clinical effectiveness been established?, Q5 = Are costs and outcomes measured accurately?, Q6 = Are costs and outcomes valued credibly?, Q7 = Are costs and outcomes adjusted for differential timing?, Q8 = Is there an incremental analysis of costs and consequences?, Q9 = Were sensitivity analyses conducted to investigate uncertainty in estimates of cost or consequences?, Q10 = Do study results include all issues of concern to users?, Q11 = Are the results generalisable to the setting of interest in the review? was assessed by two different field walking tests, with the walking Cost-effective analyses distance covered as the primary outcome. Burge et al21 used the distance (6-minute walk distance) of the self-paced 6-minute walk All 45 ICERs were extracted across the three different perspec- test as a measure of functional exercise capacity, while Zwerink et al26 tives: 25 ICERs from a healthcare perspective, 17 ICERs from a societal used the externally paced incremental shuttle walk test to report the perspective and three ICERs from a third-party payer perspective. All maximal exercise capacity. The total number of exacerbations was ICERs were converted to 2019 Euros (Tables 4 to 6). Based on the calculated as the sum of both moderate and severe exacerbations. extracted data, the intervention was interpreted to be cost-effective Similar definitions were used in both clinical trials.20,31 Mean daily in four studies,21,23,30,32 while in two it was not.25,33 The cost- step count was assessed with a pedometer to unravel any clinical effectiveness was uncertain or dependent on the willingness to pay effect on daily physical activity.26 in five studies.19,22,24,26,31 Effect of interventions Healthcare perspective More than half of all the reported ICERs for this perspective (15 of Five studies reported statistically significant changes regarding disease-specific QOL19,23,31,32 and health status30 (see Tables 4 to 6). 25, 60%) were located in the eastern quadrants of the cost- Four of these studies compared a structured multicomponent inter- effectiveness plane, suggesting that the intervention was more vention with exercise training versus usual care.19,30–32 While Gil- effective than the control group. Twenty percent of the ICERs were lespie et al (n = 350) reported only a statistically significant, positive dominant (south-east quadrant) indicating that the intervention was change in the mean total CRQ, both Hoogendoorn et al31 (n = 175) and more effective and less expensive than the comparison. For example, Goldstein et al32 (n = 89) showed statistically and clinically relevant Griffiths et al (n = 200) showed that their 6-week outpatient PR changes in disease-specific QOL. Griffiths et al (n = 200) demonstrated program was more effective (DQALY, 10.03, 95% CI 0.002 to 0.058) a positive, albeit small, significant change in QALYs.30 Goossens et al and potentially cost saving (D cost, –£152, 95% CI –880 to 577) (n = 139) compared the usual inpatient hospital care, including res- compared with standard care.30 In contrast with these unambiguous piratory physiotherapy, versus an early assisted discharge program at findings, Burns et al24 (n = 148) concluded that while the 12-month home. A statistically significant difference was observed between the low-intensity maintenance program was less expensive (D cost, groups in the probability of having a clinically relevant improvement £–204.04, 95% CI –1,522.18 to 1,114.10), it yielded incremental QALY in the CCQ score between days 3 and 7 (51.3% in the usual hospital gains (10.015, 95% CI –0.050 to 0.079) but had losses in disease- care group versus 31.7% in the early discharge group).23 specific QOL (DCRQ, –0.007, 95% CI –0.461 to 0.447) (south-west quadrant). Also, only one ICER in Cross et al’s study22 (n = 522) on Costs manual percussion dominated the control group for disease-specific QOL (DSGRQ, symptoms score, –0.09). Overall, percussion was asso- Costs reported in the studies could be divided into three sub- ciated with lower health service costs by savings associated with groups: intervention, direct medical (healthcare, out-of-pocket) and fewer hospital admissions among participants assigned to receive indirect costs (productivity loss). All studies reported the intervention percussion. However, interpretation of this saving should be exam- and direct medical cost. Interestingly, only four studies investigated ined in the light of the location of the other ICERs, which demon- the effect of their intervention on productivity loss.20,23,24,31 Only in strated clinical changes in favour of the control group (south-west two out of 11 studies, the mean costs were significantly higher than quadrant).22 The majority of the ICERs in the eastern quadrant were the alternative interventions (p , 0.05).19,20 Gillespie et al assessed located in the north-east quadrant (40%), indicating that the inter- the cost-effectiveness of a structured education PR program relative vention was more effective but also more costly than the alternatives to usual practice in primary care from a healthcare perspective. The for positive changes in daily physical activity,26 exercise capacity,26 intervention was estimated to result in a statistically significant in- disease-specific QOL19,23 and health status (utilities).19,23,26 In crease in mean cost per patient of V944 (95% CI 489 to 1,400) in total contrast to the other studies, the integrated care program by Boland healthcare costs.19 Boland et al compared a disease management et al (n = 1,086) was not cost-effective in primary care since it was program to usual care. The adjusted mean total costs were signifi- dominated by the alternative (usual care) for all clinical outcomes and cantly higher in the disease management group than in the usual care related costs (north-west quadrant).20 group by V584 (95% CI 86 to 1,046) from a healthcare perspective and V645 (95% CI 28 to 190) from a societal perspective.20 Societal perspective Twelve of 17 ICERs (71%) were located in the eastern quadrants of the cost-effectiveness plane, suggesting that the intervention was more effective than the control group. Twenty-four percent of the

Research 279 Table 3 quadrant, meaning that the interventions were dominated by Use of respiratory physiotherapy interventions per syllabus items described in the usual care.20,31 European Respiratory Society core curriculum in respiratory physiotherapy. Third-party payer’s perspective Core curriculum in respiratory physiotherapy for adult Studies Only Hoogendoorn et al reported ICERs from a third-party payer’s patients perspective. Two of three ICERs for this perspective were located in Module Syllabus item  Cross et al 201022 the north-east quadrant, and slightly lower than the societal 2 Airway clearance Body positioning  Cross et al 201022 perspective: V7,774.82 per additional patient with a relevant Manual clapping  Cross et al 201022 improvement in SGRQ total score or V27,769.26 per QALY. Never- techniques Vibration  Cross et al 201022 theless, a higher mean number of exacerbations (0.84, 95% CI –0.07 to ACBT  Cross et al 201022 1.78) led to an ICER in the north-west quadrant.31 3 Respiratory muscle Assisted cough  Goossens et al 201323 training, breathing  Cross et al 201022 Discussion strategies and Self-management/  Goossens et al 201323 techniques education ACT  Goossens et al 201323 Of the 11 included studies, four studies were interpreted to be for lung expansion Breathing exercises cost-effective,21,23,30,32 while two were not.25,33 In five studies, the  Boland et al 201520 cost-effectiveness was uncertain or dependent on the willingness to 4 Exercise training and Exercise training  Burge et al 202021 pay.19,22,24,26,31 For example, Zwerink et al (n = 153) reported that physical activity  Burns et al 201624 their community-based exercise program could not be considered  Gillespie et al 201319 cost-effective based on the primary outcome. Nonetheless, the Physical activity  Goldstein et al 199732 ICERs for the secondary outcomes on physical activity and QALY  Griffiths et al 200130 were generally considered to be acceptable.26 Since respiratory  Hoogendoorn et al physiotherapy interventions are aimed at improving health-related QOL,35 this specific treatment goal was investigated in more detail. 201031 Regardless of the economic perspectives, this systematic review  Waterhouse et al showed that 67% of all QOL-related ICERs were situated in the north-east or south-east quadrants of the cost-effectiveness plane. 201025 Moreover, 71% of all the QALYs were also located in these quadrants.  Zwerink et al 201626 When examining potential new treatments and ever-growing  Boland et al 201520 medical costs for an efficient healthcare system, cost-effectiveness  Burge et al 202021 thresholds are often used. While the advantages and disadvan-  Burns et al 201624 tages of cost-effectiveness thresholds as a formal measure are  Goldstein et al 199732 known, many countries are still using cost-effectiveness thresholds  Goossens et al 201323 within the World Health Organization guidelines of one-to-three  Griffiths et al 200130 times the gross domestic product per capita.36 Cameron et al  Hoogendoorn et al confirmed this approach in a recent systematic review and identi- fied an average willingness to pay per QALY of US$77,509.37 201031 Comparing the ICERs located in the north-east and south-east  Waterhouse et al quadrants with the specified cost-effectiveness threshold of US$77,509 per QALY, six studies21,23,24,26,30,31 could be seen as cost- 201025 effective in terms of cost per QALY when offering respiratory  Zwerink et al 201626 physiotherapy interventions focusing on exercise training in com- bination with enhancing physical activity levels. ACBT = active cycle breathing techniques, ACT = airway clearance technique. Although most ICERs indicated that the interventions were cost- ICERs were dominant (south-east quadrant) indicating that the effective and since statistical and clinical significance was hard to intervention was more effective and less expensive than the com- demonstrate, even with the applied sensitivity analysis, these ICERs parison. The hospital care in the Goossens et al (n = 139) study should be interpreted with caution for several reasons. First, the included respiratory physiotherapy for all participants with breathing complex and heterogeneous nature of COPD as a disease could limit and coughing instructions and reactivation for a treatment period of 7 the generalisability of the reported results. Patients with very severe days. A standardised written instruction was developed, ensuring COPD (FEV1 , 30% predicted) were not enrolled in the selected identical instruction in the participating hospitals.23,34 This hospital trials. Although obtaining a homogeneous sample was prioritised in care was less expensive (D mean cost V–880, 95% CI –580 to 2,268) this study design, it does raise concerns about the generalisability than the early assisted-discharge group after 3 months of follow-up towards those very severe patients seen in clinical practice. and yielded incremental gains in disease-specific QOL (DCCQ Furthermore, Cross et al22 reported that the use of manual chest –0.041, 95% CI –0.41 to 0.48) and health status (DQALY 0.005, 95% CI therapy for airway clearance did not appear to affect QOL, while it –0.021 to 0.0095).23 Although at 3 months all ICERs were located in was cost-effective. Since much uncertainty was associated with the the south-east quadrant, the ICERs after the 7-day treatment period latter, it would be difficult to justify providing ACT on the basis of were located in the north-east quadrant. The cost in the hospital care cost-effectiveness alone. Nevertheless, ACTs seem to be safe and group was now slightly higher than the early assisted-discharge generate small benefits on short-term reductions in the need for group, leading to an incremental cost difference of V65 (95% CI increased ventilatory assistance, duration of ventilatory assistance –152 to 25) in combination with a positive clinical effect on disease- and hospital length of stay. Therefore, selecting better-defined specific QOL (DCCQ –0.290, 95% CI –0.03 to 0.61).23 The same ICERs subgroups in the future (ie, patients prone to the retention of se- were observed for the 24-month INTERdisciplinary COMmunity- cretions as a treatable trait)38 could potentially demonstrate both based COPD management program (INTERCOM) in the study by the cost-effectiveness and therapeutic value of ACT.39,40 Second, Hoogendoorn et al,31 leading to a total of 47% ICERs located in the from all 11 economic evaluations included in the present review, north-east quadrant. More specifically, the INTERCOM group had 30% only five studies reported statistically significant positive changes in (95% CI 3 to 56) more patients with a clinically relevant improvement disease-specific QOL19,23,31 and health status.30 The majority of these in disease-specific QOL (SGRQ) and 0.08 (95% CI –0.01 to 0.18) more studies compared a structured multicomponent intervention with QALYs per patient. Mean total 2-year costs were higher for INTERCOM than for usual care, which resulted in an ICER of V9,960.48 per additional patient with a relevant improvement in SGRQ or V35,577 per QALY.31 All ICERs related to the cost per exacerbation avoided within a time horizon of 24 months were located in the south-west

Table 4 Healthcare perspective: cost and effects (mean) for each group, mean (95% CI) difference between groups, incremental cost- Study Outcome, time horizon Costs Intervention minus Intervention Control Daily physical activity Zwerink et al 2016 Cost per additional patient with 500 V6,949 V6,511 V438 ( steps/day, 24 mths Exercise capacity Cost per D6MWD, 8 weeks $15,447 $19,944 $–4,497 (212,2 Cost per additional patient with a V6,949 V6,511 V438 ( Burge et al 2020 clinically relevant improvement in ISWT, Zwerink et al 2016 24 mths Quality of life Cost per point additional improvement V1,463 V1,219 V244 (–315 Goossens et al 2013d in mean CCQ score, 7 days Cost per additional patient with V1,463 V1,219 V244 (–315 Goossens et al 2013d improved CCQ score, 7 days Cost per point additional improvement V4,297 V4,129 V168 (–1,25 Goossens et al 2013d in mean CCQ score, 3 mths Cost per additional patient with V4,297 V4,129 V168 (–1,25 Goossens et al 2013d improved CCQ score, 3 mths Cost per CRQ total score, 22 weeks V2,357 V1,505 V944 (489 Gillespie et al 2013 Cost per SQRQ symptoms score, 6 mths £5,870.31 £6,281.10 £–410.79 Cross et al 2010 Cost per SQRQ activity score, 6 mths £5,870.31 £6,281.10 £–410.79 Cross et al 2010 Cost per SQRQ impacts score, 6 mths £5,870.31 £6,281.10 £–410.79 Cross et al 2010 Cost per SQRQ total score, 6 mths £5,870.31 £6,281.10 £–410.79 Cross et al 2010 Cost per DCRQ, 12 mths Burns et al 2016 Cost per additional patient with clinically £3,726 £3,122 £–204.04 (–1,522 Boland et al 2015 relevant improvement in CCQ, 24 mths V5,119 V4,535 V584 (86 t Cost per additional patient with a Boland et al 2015 clinically relevant improvement in SGRQ, V5,119 V4,535 V584 (86 t 24 mths Utility Cost per QALY gained, 22 weeks V2,357 V1,505 V944 (489 Cost per QALY, 8 weeks $15,447 $19,944 $–4,497 (212,2 Gillespie et al 2013 Cost savings per incremental QALY lost, 3 V4,297 V4,129 Burge et al 2020 mths V168 (–1,25 Goossens et al 2013d Cost per QALY, 6 mths £5,870.31 £6,281.10 Cost per QALY, 12 mths £4,002 £2,686 £–410.79 Cross et al 2010 Cost per QALY, 24 mths V5,119 V4,535 £–204.04 (–1,522 Burns et al 2016 Cost per additional QALY, 24 mths V6,949 V6,511 Boland et al 2015 Cost per QALY, 18 mths V584 (86 t Zwerink et al 2016 Cost per QALY, 12 mths £3,643.74 £4,511.21 V438 ( Waterhouse et al 2010 £1,674 £1,826 Griffiths et al 2001 £–867 (–2,366 £–152 (–88 Exacerbations Boland et al 2015 Cost per exacerbation avoided, 24 mths V5,119 V4,535 V584 (86 t aAdjusted with missing data imputed, bEstimation of incremental effectiveness at follow-up was undertaken using GEE regre cConversion of original ICERs to 2019 Euros by CCEMG-EPPI-Centre Cost Converter with IMF as source dataset for PPP value a usual care involved physiotherapy interventions. CCQ = Clinical COPD Questionnaire, CE = cost effectiveness, CRQ = Chronic Respiratory Questionnaire, ICERs = incremental co years. SGRQ = St. George Respiratory Questionnaire, S/E = south-east, S/W = south-west, 6MWD = 6-minute walk distance. * Significant, p , 0.05.

-effectiveness ratios and cost-effectiveness plane per outcome. 280 Leemans et al: Cost-effectiveness of respiratory physiotherapy in COPD Effect Cost-effectiveness s control (95% CI) Intervention Control Intervention minus control (95% CI) ICERs (2019 Euros)b CE plane (N/A) 0.09 –0.19 0.28 (–0.01 to 0.59) V1,783.16 N/E 250 to 3,257) 24 11 14 (211 to 39) Dominant S/E (N/A) –0.28 –0.35 0.07 (–0.18 to 0.33) V7,133.77 N/E 5 to –168) –0.303 –0.013 –0.290 (–0.03 to 0.61) V959.99 N/E 5 to –168) 0.513 0.327 0.1941 (–0.3625 to –0.5)* V1,433.14 N/E 53 to 922) 0.024 0.065 –0.041 (–0.41 to 0.48) V4,672.24 N/E 53 to 922) 0.399 0.358 0.0417 (–0.2194 to 0.1527) V4,560.51 N/E to 1,400)* 20.82 19.10 1.11 (0.35 to 1.87)*,b V917.52 N/E 9 (N/A) –11.28 –11.19 –0.09 (N/A) Dominant S/E 9 (N/A) –5.50 –6.00 0.50 (N/A) V1,154.04 S/W 9 (N/A) –6.06 0.91 (N/A) S/W 9 (N/A) –5.15 –6.10 0.89 (N/A) V636.56 S/W 2.18 to 1,114.10)a –5.22 –0.819 V654.95 S/W to 1,046)* –0.537 –0.007 (–0.461 to 0.447)b V38,0 06.90 N/W 0.12 –0.02 (–0.06 to 0.02) Dominated to 1,046)* 0.11 N/W 0.27 –0.01 (–0.07 to 0.04) Dominated 0.26 to 1,400)* 0.337 0.305 0.002 (–0.006 to 0.011)b V509,492.18 N/E 250 to 3,257) 0.645 0.621 0.025 (20.038 to 0.086) Dominant S/E 53 to 922) 0.175 0.170 0.005 (–0.021 to 0.0095) N/E V35,470.49 9 (N/A) 0.018 0.020 –0.002 (N/A) S/W 2.18 to 1,114.10)a 0.581 0.555 0.015 (–0.050 to 0.079)b V334,658.35 S/E to 1,046)* –0.04 (20.07 to 20.01)* Dominant N/W 1.40 1.44 N/E (N/A) 1.53 1.49 0.04 (–0.10 to 0.18) Dominated S/W 6.11 to 631.17) 1.51 1.54 –0.03 (–0.13 to 0.07) V12,484.39 S/E 80 to 577) 0.381 0.351 0.03 (0.002 to 0.058)* V39,425.41 Dominant to 1,046)* 0.78 0.65 –0.14 (–0.30 to 0.06) Dominated N/W ession models, assuming a Gaussian variance function and controlling for treatment arm, baseline EQ5D score and clustering, and selected country Belgium, dControl group and intervention group are reversed compared with original publication since ost-effectiveness ratios, ISWT = incremental shuttle walk test, N/E, north-east, N/W, north-west, QALY = quality-adjusted life

Table 5 Societal perspective: cost and effects (mean) for each group, mean (95% CI) difference between groups, incremental cost-eff Study Outcome, time horizon Costs Intervention m Intervention Control Quality of life Cost per point additional improvement in mean V1,463 V1,398 V65 Goossens et al 2013b CCQ score, 7 days V1,463 V1,398 V65 Goossens et al 2013b Cost per additional patient with improved CCQ V5,395 V6,304 V–880 Goossens et al 2013b score, 7 days V5,395 V6,304 V–880 Goossens et al 2013b Cost per point additional improvement in mean V5,750 V5,105 V645 Boland et al 2015 CCQ score, 3 mths V5,750 V5,105 V645 Boland et al 2015 Cost per additional patient with improved CCQ V13,565 V10,814 V2,751 Hoogendoorn et al 2010 score, 3 mths V12,311 Goldstein et al 1997 Cost per additional patient with clinically V12,311 V654 $1 Goldstein et al 1997 relevant improvement in CCQ, 24 mths V12,311 V654 $1 Goldstein et al 1997 Cost per additional patient with a clinically V12,311 V654 $1 Goldstein et al 1997 relevant improvement in SGRQ, 24 mths V654 $1 Cost per additional patient with a clinically relevant improvement in SGRQ, 24 mths Cost per unit change in CRQ domain dyspnoea, 6 mths Cost per unit change in CRQ domain fatigue, 6 mths Cost per unit change in CRQ domain emotional function, 6 mths Cost per unit change in CRQ domain mastery, 6 mths Utility Goossens et al 2013b Cost savings per incremental QALY lost, 3 mths V5,395 V6,304 V–880 Burge et al 2020 Cost per QALY, 12 mths N/A N/A $ –4,316 ( Hoogendoorn et al 2010 Cost per QALY, 24 mths Boland et al 2015 Cost per QALY, 24 mths V13,565 V10,814 V2,751 V5,750 V5,105 V645 Exacerbations Boland et al 2015 Cost per exacerbation avoided, 24 mths V5,750 V5,105 V645 Hoogendoorn et al 2010 Cost per exacerbation avoided, 24 mths V13,565 V10,814 V2,751 aConversion of original ICERs to 2019 Euros by CCEMG-EPPI-Centre Cost Converter with IMF as source dataset for PPP value a usual care involved physiotherapy interventions. CCQ = Clinical COPD Questionnaire, CRQ = Chronic Respiratory Questionnaire, CE = cost effectiveness, ICER = incremental co Questionnaire, S/E = south-east, S/W = south-west. * Significant, p , 0.05.

fectiveness ratios and cost-effectiveness plane per outcome. Effect Cost-effectiveness minus control (95% CI) Intervention Control Intervention minus control (95% CI) ICER (2019 Euros)a CE plane 5 (–152 to 25) –0.303 –0.013 –0.290 (–0.03 to 0.61) V224 N/E 5 (–152 to 25) 0.513 0.327 0.1941 (–0.3625 to –0.5)* V335 N/E 0.024 0.065 Dominant S/E (–580 to 2,268) 0.399 0.358 –0.041 (–0.41 to 0.48) Dominant S/E (–580 to 2,268) 0.11 0.0417 (–0.2194 to 0.1527) Dominated N/W 5 (28 to 190)* 0.26 0.12 Dominated N/W 5 (28 to 190)* 0.13 0.27 –0.02 (–0.06 to 0.02) V9,078 N/E 1 (–631 to 6,372) – –0.17 –0.01 (–0.07 to 0.04) V19,011 N/E 11,597 (N/A) – V35,142 N/E 11,597 (N/A) – – 0.3 (0.03 to 0.56)* V26,357 N/E 11,597 (N/A) – – 0.61 (N/A)* V16,567 N/E 11,597 (N/A) – 0.33 (N/A) – 0.44 (N/A)* 0.70 (N/A)* (–580 to 2,268) 0.175 0.170 0.005 (–0.021 to 0.0095) Dominant S/E Research 281 (216,328 to 6,780) 0.645 0.621 0.025 (20.038 to 0.086) Dominant S/E 1 (–631 to 6,372) N/E 5 (28 to 190)* 1.62 1.54 0.08 (–0.01 to 0.18) V32,425 N/W 1.40 1.44 –0.04 (20.07 to 20.01)* Dominated 5 (28 to 190)* 0.78 0.65 –0.14 (–0.30 to 0.06) Dominated N/W 1 (–631 to 6,372) 3.02 2.18 0.84 (–0.07 to 1.78) Dominated N/W and selected country Belgium, bControl group and intervention group are reversed compared with original publication since ost-effectiveness ratios, N/E, north-east, N/W, north-west, QALY = quality-adjusted life years, SGRQ = St. George Respiratory

282 Leemans et al: Cost-effectiveness of respiratory physiotherapy in COPDCost-effectivenessIntervention Control Intervention minus control (95% CI) Intervention Control Intervention minus control (95% CI) ICER (2019 Euros)a CE planeV7,774.82 N/E V27,769.26 N/E N/W exercise training and physical activity versus usual care.19,30–32 In Dominated general, the reported content, timing and organisational aspects Table 6 between all evaluated exercise interventions also differed greatly, 0.3 (0.03 to 0.56)* 0.08 (–0.01 to 0.18) 0.84 (–0.07 to 1.78) Third-party payer’s perspective: cost and effects (mean) for each group, mean (95% CI) difference between groups, incremental cost-effectiveness ratios and cost-effectiveness plane per outcome. which limited comparison and thus drawing conclusions. Therefore, no particular intervention type was found to be more effective thanEffect aConversion of original ICERs to 2019 Euros by CCEMG-EPPI-Centre Cost Converter with IMF as source dataset for PPP value and selected country Belgium. another intervention. Third, the value for QALYs produced as a result CE = cost-effectiveness, ICER = incremental cost-effectiveness ratios, N/E, north-east, N/W, north-west. QALY = quality-adjusted life years, SGRQ = St. George Respiratory Questionnaire. of the described interventions appeared small in absolute terms. –0.17 1.62 1.54 3.02 2.18 This could be explained by the relative insensitivity to change in * Significant, p , 0.05. the generic QOL instruments to detect a clinically meaningful 0.13 improvement in COPD health status compared with larger changes seen in multi-dimensional disease-specific QOL outcome mea-Costs V2,147 (–1,091 to 5,649) V2,147 (–1,091 to 5,649) V2,147 (–1,091 to 5,649) sures.19,30 Fourth, cost-effectiveness ratios were influenced by the chosen time horizon: of the six studies that reported a cost- V12,145 V9,998 V12,145 V9,998 V12,145 V9,998 effectiveness threshold lower than US$77,509 per QALY, only one had a time horizon of less than 1 year.23 Although the chosen timeOutcome, time horizonHoogendoorn et al 2010 Cost per additional patient with a clinically relevant Hoogendoorn et al 2010 Cost per QALY, 24 mths Hoogendoorn et al 2010 Cost per exacerbation avoided, 24 mths horizons of the other five studies were between 1 and 2 years, the improvement in SGRQ, 24 mths cost-effectiveness ratios might still be less favourable in the longerStudyQuality of life Utility Exacerbations term. Finally, a comparison of the various estimates of cost- effectiveness remains complicated because economic measures depend on other factors as well, like local regulations and national health policies. Consequently, generalisation of the findings to other countries may be limited. It is believed that this is the first systematic review to focus on economic evaluations of respiratory physiotherapy interventions in COPD. The methodology was enhanced by using both PRISMA guidelines and the five-step approach for systematic reviews on economic evaluations11,12 to ensure the comprehensiveness of the systematic literature search. Although this extensive approach resulted in 11 trials of high methodological quality, with the ma- jority focusing on exercise training and physical activity, this small number of studies may have limited the significance of this review. For some interventions such as ACT or respiratory muscle training, no conclusions could be made. Nevertheless, these results may offer a solid basis for future reviews and meta-analyses. Another aspect was the heterogeneous nature of the multicomponent in- terventions identified in the economic evaluations, which limited comparison and thus drawing conclusions. As expected, it was inappropriate to carry out a meta-analysis. In order to shed light on this heterogeneity, the recent ERS harmonised respiratory physio- therapy curriculum was used as a framework for structuring the different components of all interventions.14 This enabled the au- thors to compare the different studies, resulting in a detailed descriptive analysis of health economic outcomes on current practices across different areas of respiratory physiotherapy in COPD. Regarding the external validity of the described costs, to provide a best possible comparison of studies, all costs were re- ported in 2019 Euros. This systematic review highlights the diffi- culties in designing and undertaking cost-effectiveness studies and raises the question of whether it is possible to determine the economic measures of individual respiratory physiotherapy in- terventions versus the total package in multi-modal interventions. For future research, this study supports the recent recommenda- tions by Rodrigues et al40 that well-designed respiratory physio- therapy intervention studies are still needed with larger samples based on size estimation (eg, very severe COPD with multi- morbidity), optimal and well-described protocols, improved blinding strategies, responsive outcome measures to demonstrate effectiveness and long-term data collection (eg, costs) across the whole healthcare spectrum. In conclusion, this sample of 11 studies shows that respiratory physiotherapy interventions focusing on exercise training in combi- nation with enhancing physical activity levels have the potential to be cost-effective in terms of costs per unit QOL gained and QALYs. Despite these results, there is still uncertainty on various estimates of cost-effectiveness, due to differences in content and intensity of the type of interventions, outcome measures and comparators. In the future, more studies investigating cost-effectiveness of respiratory physiotherapy interventions in terms of costs per QALY gained from a societal perspective over a long-term horizon are needed to reach definite conclusions.

Research 283 What was already known on this topic: Chronic obstructive 18. The Campbell and Cochrane Economics Methods, Evidence for Policy and Practice pulmonary disease is one of the most prevalent chronic respira- Information and Coordinating Centre. CCEMG – EPPI-Centre Cost Converter. http:// tory diseases worldwide and associated with a significant social eppi.ioe.ac.uk/costconversion/default.aspx. Accessed 20 April, 2020. and economic burden. What this study adds: Respiratory physiotherapy in- 19. Gillespie P, O’Shea E, Casey D, Murphy C, Devane D, Cooney A, et al. The cost- terventions focusing on exercise training in combination with effectiveness of a structured education pulmonary rehabilitation programme for enhancing physical activity levels are likely to be cost-effective chronic obstructive pulmonary disease in primary care: the PRINCE cluster ran- in terms of costs per unit gain in health-related quality of life domized trial. BMJ Open. 2013;3:e003479. and quality-adjusted life years. 20. Boland MRS, Kruis AL, Tsiachristas A, Assendelft WJJ, Gussekloo J, Blom CMG, et al. Footnotes: a Mendeley Desktop, version 1.19.5, Mendeley Ltd, Cost-effectiveness of integrated COPD care: The RECODE cluster randomized trial. London, UK. b EPPI-Centre Cost Converter, version 1.6, London, UK. BMJ Open. 2015;5. eAddenda: Appendix 1 can be found online at https://doi.org/10. 21. Burge AT, Holland AE, McDonald CF, Abramson MJ, Hill CJ, Lee AL, et al. Home- 1016/j.jphys.2021.08.018 based pulmonary rehabilitation for COPD using minimal resources: an economic analysis. Respirology. 2020;25:183–190. Ethics approval: Not applicable. Competing interests: Nil. 22. Cross J, Elender F, Barton G, Clark A, Shepstone L, Blyth A, et al. A randomised Source(s) of support: Not applicable. controlled equivalence trial to determine the effectiveness and cost-utility of Acknowledgements: The authors would like to acknowledge Kris manual chest physiotherapy techniques in the management of exacerbations of Ides for helping with the initial literature search. chronic obstructive pulmonary disease (MATREX). Health Technol Assess. Correspondence: Glenn Leemans, Faculty of Medicine and 2010;14:1–147. iii-iv. Health Sciences, University of Antwerp, Belgium e-mail: [email protected] 23. Goossens LMA, Utens CMA, Smeenk FWJM, van Schayck OCP, van Vliet M, van Litsenburg W, et al. Cost-effectiveness of early assisted discharge for COPD exac- References erbations in the Netherlands. Value Heal. 2013;16:517–528. 1. Global Initiative for Chronic Obstructive Lung Disease. Global Strategy for the 24. Burns DK, Wilson ECF, Browne P, Olive S, Clark A, Galey P, et al. The cost effec- Diagnosis. Management and Prevention of COPD; 2021. tiveness of maintenance schedules following pulmonary rehabilitation in patients with chronic obstructive pulmonary disease: an economic evaluation alongside a 2. Agusti A, Bel E, Thomas M, Vogelmeier C, Brusselle G, Holgate S, et al. Treatable randomised controlled trial. Appl Health Econ Health Policy. 2016;14:105–115. traits: toward precision medicine of chronic airway diseases. Eur Respir J. 2016;47:410–419. 25. Waterhouse JC, Walters SJ, Oluboyede Y, Lawson RA. A randomized 2 3 2 trial of community versus hospital pulmonary rehabilitation for chronic obstructive pul- 3. Soriano JB, Abajobir AA, Abate KH, Abera SF, Agrawal A, Ahmed MB, et al. Global, monary disease followed by telephone or conventional follow-up. Health Technol regional, and national deaths, prevalence, disability-adjusted life years, and years Assess (Rockv). 2010;14:1–164. lived with disability for chronic obstructive pulmonary disease and asthma, 1990–2015: a systematic analysis for the Global Burden of Disease Study 2015. 26. Zwerink M, Effing T, Kerstjens HAM, van der Valk P, Brusse-Keizer M, Zielhuis G, Lancet Respir Med. 2017;5:691–706. et al. Cost-effectiveness of a community-based exercise programme in COPD self- management. COPD J Chronic Obstr Pulm Dis. 2016;13:214–223. 4. European Respiratory Society. The economic burden of lung disease. In: The Eu- ropean Lung White Book; 2013. https://www.erswhitebook.org/about/ 27. Troosters T, Langer D, Burtin C, Chatwin M, Clini EM, Emtner M, et al. A guide for respiratory physiotherapy postgraduate education: presentation of the harmonized 5. Guarascio AJ, Ray SM, Finch CK, Self TH. The clinical and economic burden of curriculum. Eur Respir J. 2019;53. https://doi.org/10.1183/13993003.00320-2019 chronic obstructive pulmonary disease in the USA. Clin Outcomes Res. 2013;5:235– 245. 28. NICE. Glossary; 2020. https://www.nice.org.uk/glossary?letter=q. Accessed 3 June, 2020. 6. ur Rehman A, Ahmad Hassali MA, Muhammad SA, Shah S, Abbas S, Bin Hyder Ali IA, et al. The economic burden of chronic obstructive pulmonary disease (COPD) 29. Whitehead SJ, Ali S. Health outcomes in economic evaluation: The QALY and in the USA, Europe, and Asia: results from a systematic review of the literature. utilities. Br Med Bull. 2010;96:5–21. Expert Rev Pharmacoecon Outcomes Res. 2020;20:661–672. 30. Griffiths TL, Phillips CJ, Davies S, Burr ML, Campbell IA. Cost effectiveness of an 7. Faulkner E, Annemans L, Garrison L, Helfand M, Holtorf A-P, Hornberger J, et al. outpatient multidisciplinary pulmonary rehabilitation programme. Thorax. Challenges in the development and reimbursement of personalized medicine- 2001;56:779–784. payer and manufacturer perspectives and implications for health economics and outcomes research: a report of the ISPOR personalized medicine special interest 31. Hoogendoorn M, van Wetering CR, Schols AM, Rutten-van Molken MP. Is INTER- group. Value Heal. 2012;15:1162–1171. disciplinary COMmunity-based COPD management (INTERCOM) cost-effective? Eur Respir J. 2010;35:79–87. 8. Hoogendoorn M, Corro Ramos I, Baldwin M, Gonzalez-Rojas Guix N, Rutten-van Mölken MPMH. Broadening the perspective of cost-effectiveness modeling in 32. Goldstein RS, Gort EH, Guyatt GH, Feeny D. Economic analysis of respiratory chronic obstructive pulmonary disease: a new patient-level simulation model rehabilitation. Chest. 1997;112:370–379. suitable to evaluate stratified medicine. Value Heal. 2019;22:313–321. 33. Boland MRS, Kruis AL, Huygens SA, Tsiachristas A, Assendelft WJJ, Gussekloo J, et al. 9. Spruit MA, Singh SJ, Garvey C, ZuWallack R, Nici L, Rochester C, et al. An official Exploring the variation in implementation of a COPD disease management pro- American Thoracic Society/European Respiratory Society statement: key concepts gramme and its impact on health outcomes: a post hoc analysis of the RECODE and advances in pulmonary rehabilitation. Am J Respir Crit Care Med. cluster randomized trial. NPJ Prim Care Respir Med. 2015;25. https://doi.org/10. 2013;188:e13–64. 1038/npjpcrm.2015.71 10. Zoumot Z, Jordan S, Hopkinson NS. Emphysema: time to say farewell to therapeutic 34. Utens CMA, Goossens LMA, Smeenk FWJM, van Schayck OCP, van Litsenburg W, nihilism. Thorax. 2014;69:973–975. Janssen A, et al. Effectiveness and cost-effectiveness of early assisted discharge for chronic obstructive pulmonary disease exacerbations: the design of a randomized 11. Liberati A, Altman DG, Tetzlaff J, Mulrow C, Gotzsche P, Ioannides JPA, et al. The controlled trial. BMC Public Health. 2010;10:1–12. PRISMA statement for reporting systematic reviews and meta-analyses of studies that evaluate health care interventions: explanation and elaboration. PLoS Med. 35. McCarthy B, Casey D, Devane D, Murphy K, Murphy E, Lacasse Y. Pulmonary 2009;6:e1000100. rehabilitation for chronic obstructive pulmonary disease. Cochrane Database Syst Rev. 2015;2:2. 12. van Mastrigt GAPG, Hiligsmann M, Arts JJC, Broos PH, Kleijnen J, Evers SMAA, et al. How to prepare a systematic review of economic evaluations for informing 36. Marseille E, Larson B, Kazi DS, Kahn JG, Rosen S. Thresholds for the cost– evidence-based healthcare decisions: a five-step approach (part 1/3). Expert Rev effectiveness of interventions: alternative approaches. Bull World Health Organ. Pharmacoecon Outcomes Res. 2016;16:689–704. 2015;93:118–124. 13. Drummond MF, Sculpher MJ, Claxton K, Stoddart GL, Torrance GW. Methods for the 37. Cameron D, Ubels J, Norström F. On what basis are medical cost-effectiveness Economic Evaluation of Health Care Programmes. Oxford: Oxford University Press; thresholds set? Clashing opinions and an absence of data: a systematic review. 2015. Glob Health Action. 2018;11. https://doi.org/10.1080/16549716.2018.1447828 14. Troosters T, Tabin N, Langer D, Burtin C, Chatwin M, Clini EM, et al. Introduction of the 38. McDonald VM, Osadnik CR, Gibson PG. Treatable traits in acute exacerbations of harmonised respiratory physiotherapy curriculum. Breathe. 2019;15:110–115. chronic airway diseases. Chron Respir Dis. 2019;16. https://doi.org/10.1177/ 1479973119867954 15. Wijnen B, Van Mastrigt G, Redekop W, Majoie H, De Kinderen R, Evers SMAA. How to prepare a systematic review of economic evaluations for informing evidence- 39. Osadnik CR, McDonald CF, Jones AP, Holland AE. Airway clearance techniques for based healthcare decisions: data extraction, risk of bias, and transferability (part chronic obstructive pulmonary disease. Cochrane Database Syst Rev. 3/3). Expert Rev Pharmacoecon Outcomes Res. 2016;16:723–732. 2012;3:CD008328. 16. Gomersall JS, Jadotte YT, Xue Y, Lockwood S, Riddle D, Preda A. Conducting systematic 40. Rodrigues A, Muñoz Castro G, Jácome C, Langer D, Parry SM, Burtin C. Current reviews of economic evaluations. Int J Evid Based Healthc. 2015;13:170–178. developments and future directions in respiratory physiotherapy. Eur Respir Rev. 2020;29:200264. 17. Spruit MA, Pitta F, Garvey C, ZuWallack RL, Roberts CM, Collins EG, et al. Differ- ences in content and organisational aspects of pulmonary rehabilitation pro- 41. Kruis AL, Boland MRS, Schoonvelde CH, Assendelft WJJ, Rurren-van Molken MPMH, grammes. Eur Respir J. 2014;43:1326–1337. Gussekloo J, et al. RECODE: Design and baseline results of a cluster randomized trial on cost-effectiveness of integrated COPD management in primary care. BMC Pulm Med. 2013;13:1–15. 42. Holland AE, Mahal A, Hill CJ, Lee AL, Burge AT, Moore R, et al. Benefits and costs of home-based pulmonary rehabilitation in chronic obstructive pulmonary disease - a multi-centre randomized controlled equivalence trial. BMC Pulm Med. 2013;13:1. 43. Goldstein RS, Gort EH, Avendano MA, Stubbing D, Guyatt GH. Randomised controlled trial of respiratory rehabilitation. Lancet. 1994;344:1394–1397. 44. Griffiths TL, Burr ML, Campbell A, Lewis-Jenkins V, Mullins J, Shiels K, et al. Results at 1 year of outpatient multidisciplinary pulmonary rehabilitation: a randomized controlled trial. Lancet. 2000;355:362–368. 45. Van Wetering CR, Hoogendoorn M, Mol SJM, Rutten-Van Molken MPMH, Schols AM. Short- and long-term efficacy of a community-based COPD manage- ment programme in less advanced COPD: A randomized controlled trial. Thorax. 2010;65:7–13.

Journal of Physiotherapy 67 (2021) 238–239 j o u r n a l h o m e p a g e : w w w. e l s ev i e r. c o m / l o c a t e / j p hy s Editorial Sex and gender in physiotherapy research K Jane Chalmers a, Mark R Elkins b,c a IIMPACT in Health, Allied Health and Human Performance, University of South Australia, Adelaide, Australia; b Editor, Journal of Physiotherapy; c Faculty of Medicine and Health, University of Sydney, Sydney, Australia Evidence from preclinical and clinical research demonstrates that gender only (n = 1), even though the condition being investigated sex and gender can impact disease presentation, prognosis and affects both females and males. Of these 39 trials, only 17 studies response to treatment.1 Sex refers to the biological attributes that provided a rationale as to why they were investigating a single sex or define humans as male, female, or intersex, including chromosomes, gender; the remaining 22 studies failed to provide a rationale. gene expression, hormone levels and reproductive anatomy.2 Gender Interestingly, the high proportion of female-only trials is in contrast refers to the socially constructed norms, behaviours and roles of girls, to what has been seen in other health disciplines, where trials are women, boys, men and gender-diverse people.2 Gender is a contin- more likely to be male-only9 or have a higher proportion of male uum and not a static concept; gender can change over time. The participant inclusion.11,12 mechanisms by which sex and gender impact health are complex and not yet fully understood, but include an interaction between epige- Sex and gender diversity in physiotherapy research netic, genetic, environmental, social and behavioural factors, among others.3 The 191 trials with sex/gender diversity among participants included data on 19,390 participants. Nineteen trials, involving 1,049 Disaggregation of data by sex and/or gender helps to unpack the participants, failed to specify the male/female split and so could not complex biological and psychosocial interactions that might influence be included in the analysis of sex or gender ratio. For the remaining disease presentation and progression. In a clinical setting, unpacking trials, the percentage of female participants was calculated. Among this interaction helps us to screen for disease more effectively, and trials reporting the sex of participants, the mean percentage of female choose the best evidence-based treatment course for patients. Dif- participants was 50% (SD 9), indicating that these trials typically had ferences between females/women and males/men have already been roughly equal participation of males and females. Among trials identified across several disciplines of physiotherapy, including reporting the gender of participants, the mean percentage of female neurology,4 musculoskeletal5 and cardiopulmonary.6 participants was 55% (SD 22). While this result again indicates roughly equal participation of males and females overall, the larger In clinical research, recruiting sex-diverse and gender-diverse SD indicates that more trials would have had substantial imbalance. populations is essential to enable disaggregation of data. Historical- Also, no trials reported including any gender-diverse participants ly, health data have been collected on males and generalised to fe- beyond binary males or females. Population-based studies estimate males.7 Such patterns are still observed in certain health fields such the prevalence of gender nonconformity to occur in up to 4.6% of the as pain research,8 exercise science9 and pharmaceutical trials.10 To general population.13 That no trials included gender-nonconforming remedy this, the governmental funding bodies of many countries participants highlights a gap in current physiotherapy research. Pre- have mandated equality in the collection, analysis and reporting of vious research has identified that LGBTIQ1 individuals often have sex-specific and gender-specific health. It is still debateable whether uncomfortable experiences with physiotherapists and physiothera- or not these policies translate into research, so that the complex pists lack knowledge specific to the needs of these individuals.14 biological and psychosocial interactions of sex and gender with These findings may reflect: poor physiotherapist training on gender health can be unpacked.9 However, the first step in this process for diversity and inclusion; poor participation or recruitment of gender- clinical research must be the recruitment of sex-diverse and gender- diverse participants; and/or poor data collection and reporting diverse populations. regarding gender in research. It also indicates some deficiency in the evidence that physiotherapists have with which to tailor their clinical It is currently unknown whether physiotherapy research trials decision making to individual patients. recruit sex-diverse and gender-diverse populations, or whether data are reported as sex and/or gender disaggregated. To gauge the current When describing demographic data, sex was used as a descriptor but pre-pandemic status of sex and gender reporting in physio- in 132 trials and gender as a descriptor in 59 trials. Eleven trials therapy trials, the reports of 250 randomised trials published in 2019 conflated sex and gender, using the terms interchangeably were randomly sampled from among those on the Physiotherapy throughout the manuscript. Twenty-seven trials did not use either Evidence Database (PEDro). Among these, four trial reports provided term, but instead provided a number or percentage of males and/or no information or conflicting information on sex and gender; these females (n/% of males/females n = 14, n/% of males n = 7, n/% of fe- were not analysed further. Fifty-five trials studied females only or males n = 6). No trial included a definition of either sex or gender, or males only, and the remaining 191 trials included some degree of sex stated why one variable was chosen over the other to be collected or gender diversity among the participants. from participants. In the 55 trials that studied females only or males only, the mi- nority were related to conditions that are female-specific (breast cancer n = 6, menopause n = 5, pregnancy n = 2, puberty n = 1) or male-specific (prostatitis n = 1). The other 39 trials involved partici- pants with female sex only (n = 27), male sex only (n = 11) or female https://doi.org/10.1016/j.jphys.2021.08.015 1836-9553/© 2021 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/).

Editorial 239 Sex and gender disaggregation in physiotherapy research researchers to run analyses to determine possible interaction effects due to sex and/or gender, preserving statistical power, and present None of the 191 trials disaggregated their data by sex or gender. disaggregated data where such an interaction effect exists. Some trials used statistical methods to account for potential sex or gender differences. Twenty trials used statistical adjustments for sex If a sex or gender effect is identified, it is important to determine differences (18% of trials reporting the sex of participants) and five whether it is a consistent finding. This could be assessed in future trials used statistical adjustments for gender differences (9% of trials replications of the study. It could also be assessed easily in existing reporting the gender of participants). While statistical adjustment for similar trials if they publish the individual participant data with sex/ demographic variables is helpful, adjusting for sex and gender is gender data. See, for example, Table 4 in the eAddenda of the trial by insufficient in health research. Statistical adjustments do not show Scheer et al in this issue.17 where specific differences between sex and gender exist, and un- derstanding these differences is vital for clinical decision making. The Journal of Physiotherapy encourages researchers to consider whether sex and/or gender may moderate the effects of the inter- Recommendations for future research vention. Where it is plausible that sex and/or gender may be treat- ment effect moderators, researchers should plan, a priori, to This analysis of physiotherapy trials has highlighted strengths and investigate this using appropriate analyses (eg, test for an interaction weaknesses in that research. The equal recruitment of males and effect based on sex and/or gender). If a significant interaction is females by sex and gender is encouraging and is something our found, disaggregated data should be reported. We also encourage profession should continue to strive to do in research. However, there authors to adopt this approach when submitting research to other was a lack of inclusion of gender non-conforming participants. Future journals. physiotherapy research should look to recruit even or representative numbers of women and men (by gender), and also include gender- Competing interests: Nil. nonconforming participants. Source(s) of support: Nil. Acknowledgements: Nil. Sex and gender were often used interchangeably, and no studies Provenance: Invited. Not peer reviewed. provided definitions of sex or gender, or a description of why either Correspondence: Mark R Elkins, Centre for Education & Workforce sex or gender was chosen over the other as a demographic charac- Development, Sydney Local Health District, Sydney, Australia. Email: teristic. Researchers are encouraged to consider sex and gender var- [email protected] iables in the planning stage of trials, and determine which characteristic may be more important to the primary research aim. References Recruitment of participants by either sex or gender, or even both, should relate to any hypotheses around whether these characteristics 1. McGregor AJ, et al. Biol Sex Differ. 2016;7:61–72. may influence the outcomes of the trial. Gendered Innovations in 2. Canadian Institutes of Health Research. What is gender? What is sex? https://cihr- Science, Health & Medicine, Engineering, and Environment15 provides comprehensive resources for health researchers to plan for analysing irsc.gc.ca/e/48642.html. Accessed 28 July, 2021. sex and gender differences, and the Canadian Institutes of Health 3. Wainer Z, et al. Med J Aust. 2020;212:57–62. Research offer free online training courses to assist researchers in 4. Haast RA, et al. J Cereb Blood Flow Metab. 2012;32:2100–2107. appreciating the differences between and across sexes and genders.16 5. George SZ, et al. J Orthop Sports Phys Ther. 2006;36:354–363. 6. Krishnan JA, et al. Arch Intern Med. 2001;161:1660–1668. While some of the included trials statistically adjusted for sex or 7. Perez CC. Invisible women: Exposing data bias in a world designed for men. Random gender variables, no trials disaggregated their data according to sex or gender. Disaggregating data in this way ensures that differences in House; 2019. outcomes can be observed between and across sexes and genders, 8. Mogil JS. Nat Rev Neurosci. 2020;21:353–365. which is vital information for clinicians to more effectively screen for 9. Costello JT, et al. Eur J Sport Sci. 2014;14:847–851. disease and choose the best evidence-based treatment options for 10. Richardson SS, et al. Proc Natl Acad Sci. 2015;112:13419–13420. presenting patients. However, we recognise that analysis of sub- 11. Prakash VS, et al. J Womens Health. 2018;27:1342–1348. groups by sex or gender results in a loss of statistical power, which is 12. Feldman S, et al. JAMA Netw Open. 2019;2:e196700. an important consideration in clinical trials. Therefore, we encourage 13. Kuyper L, et al. Arch Sex Behav. 2014;43:377–385. 14. Ross MH, et al. J Physiother. 2019;65:99–105. 15. Gendered Innovations. http://genderedinnovations.stanford.edu/index.html. Accessed 9 August, 2021. 16. Canadian Institutes of Health. Sex and Gender Training Modules. https://www. cihr-irsc-igh-isfh.ca/. Accessed 9 August, 2021. 17. Scheer A, et al. J Physiother. 2021;67:284–290.

Journal of Physiotherapy 67 (2021) 252–262 j o u r n a l h o m e p a g e : w w w. e l s ev i e r. c o m / l o c a t e / j p hy s Research Some types of exercise are more effective than others in people with chronic low back pain: a network meta-analysis Jill A Hayden a, Jenna Ellis a, Rachel Ogilvie a, Samuel A Stewart a, Matthew K Bagg b,c, Sanja Stanojevic a, Tiê P Yamato d,e, Bruno T Saragiotto d,e a Department of Community Health & Epidemiology, Dalhousie University, Halifax, Canada; b Centre for Pain IMPACT, Neuroscience Research Australia, Sydney, Australia; c WA SportsMed Physiotherapy, Perth, Australia; d Musculoskeletal Health Sydney, School of Public Health, University of Sydney, Sydney, Australia; e Masters and Doctoral Programs in Physical Therapy, Universidade Cidade de São Paulo, São Paulo, Brazil KEY WORDS ABSTRACT Chronic low back pain Question: What are the effects of specific types of exercise treatments on pain intensity and functional Functional limitation limitation outcomes for adults with chronic low back pain? Design: Systematic review with network meta- Exercise analysis of randomised controlled trials. Participants: Adults with non-specific low back pain for  12 weeks. Network meta-analysis Intervention: Exercise treatments prescribed or planned by a health professional that involved conducting Physical therapy specific activities, postures and/or movements with a goal to improve low back pain outcomes. Outcome measures: Pain intensity (eg, visual analogue scale or numerical rating scale) and back-related functional limitations (eg, Roland Morris Disability Questionnaire or Oswestry Disability Index), each standardised to range from 0 to 100. Results: This review included 217 randomised controlled trials with 20,969 participants and 507 treatment groups. Most exercise types were more effective than minimal treatment for pain and functional limitation outcomes. Network meta-analysis results were compatible with moderate to clinically important treatment effects for Pilates, McKenzie therapy, and functional restoration (pain only) and flexi- bility exercises (function only) compared with minimal treatment, other effective treatments and other exercise types. The estimated mean differences for these exercise types compared with minimal treatment ranged from 215 to 219 for pain and from 210 to 212 for functional limitation. Conclusion: This review found evidence that Pilates, McKenzie therapy and functional restoration were more effective than other types of exercise treatment for reducing pain intensity and functional limitations. Nevertheless, people with chronic low back pain should be encouraged to perform the exercise that they enjoy to promote adherence. Registration: DOI:10.1002/14651858.CD009790. [Hayden JA, Ellis J, Ogilvie R, Stewart SA, Bagg MK, Stanojevic S, Yamato TP, Saragiotto BT (2021) Some types of exercise are more effective than others in people with chronic low back pain: a network meta-analysis. Journal of Physiotherapy 67:252–262] © 2021 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 but generally do not include all study information and often ignore or are unable to account for important treatment heterogeneity in Exercise is a common treatment approach for chronic low back design and delivery characteristics.7 Network meta-analysis methods, pain recommended by clinical practice guidelines as a first line of comparing multiple treatments simultaneously and considering other care.1,2 However, there is limited evidence to support the use of one potential sources of heterogeneity, have the potential to better type of exercise or program characteristic over another. Our recent identify the best approach for low back pain management. Cochrane review including 249 randomised clinical trials found consistent, moderately strong evidence that exercise treatment was Therefore, the specific research question for this systematic re- more effective than no treatment or usual care for the management of view was: chronic low pain.3 Exercise treatments investigated in the included trials were heterogeneous and varied in specific exercise types, pro- What are the effects of specific types of exercise treatments on gram design, dose, delivery format and whether they were combined pain intensity and functional limitation outcomes for adults with with other conservative treatments.3 chronic low back pain? Traditional methods for meta-analysis cannot answer important Methods questions about which treatment works best,4–6 and as they compare only two treatments at one time, do not allow full analysis of trials This study is reported according to the Network Meta-Analysis investigating multiple treatment groups within studies. Traditional extension of the Preferred Reporting Items for Systematic Reviews meta-analysis methods synthesise important overarching questions, https://doi.org/10.1016/j.jphys.2021.09.004 1836-9553/© 2021 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 253 and Meta-Analyses (PRISMA-NMA) reporting guidelines.8 The Box 1. Systematic review selection criteria, and additional PRISMA-NMA checklist is presented in Appendix 1 on the eAddenda. criteriaa for inclusion and exclusion in this network meta- analysis. Identification and selection of studies Study design This study builds upon a recent Cochrane review that examined  Randomised controlled trials exercise therapy for chronic low back pain.3 The literature search,  Completed and published which is described fully in the published Cochrane review,3 included Population a comprehensive electronic search of nine databases (Appendix 2 on  Adults the eAddenda). Selection criteria and operationalised definitions for  Chronic non-specific low back pain ( 12 weeks) population, intervention, comparison, outcome characteristics and Interventions study design are described (Box 1). We followed the standard pro-  Exercise treatments prescribed or planned by a health tocol for study selection and data extraction as recommended by Cochrane Back and Neck Group Methods Guidelines.9 Further professional description of the study selection and data extraction processes can  Involved specific activities, postures and/or movements with a be found in the published Cochrane review.3 goal to improve low back pain outcomes Assessment of characteristics of studies  Categorised as: Participants  Core strengthening/motor control Data were extracted on participant setting, age, sex, duration of  Mixed exercise types  General strengthening low back pain episode, presence of radicular symptoms, and pain and  Aerobic exercises functional limitations at baseline.  Pilates  Stretching Exercise treatments  Yoga Exercise treatments were described by the type of exercise, program  Functional restoration  McKenzie therapy design, delivery approach, dose (intensity and duration) and inclusion of  Flexibility additional interventions (Appendix 3 on the eAddenda). We included  Other specific exercises trials that allocated participants to any of 11 categories of exercise Comparators treatments listed in Box 1. In cases where the intervention involved  Placebo, no treatment or usual care several types of exercise, it was judged whether: a single type domi-  Education nated, in which case that exercise type was assigned as the main  Manual therapy category for the primary analyses; or no type of exercise clearly domi-  Back school nated, in which case this intervention was assigned to the mixed cate-  Electrotherapy gory. These 11 categories were used as nodes in the primary analyses.  Mixed physiotherapy (not involving exercise)  Psychological therapy Comparison interventions  Anti-inflammatory/analgesics We also included trials that allocated participants to any of the  Relaxation Outcomes exercise treatments and to at least one of nine comparison in-  Pain intensity (eg, visual analogue scale, numerical rating terventions shown in Box 1. In the related Cochrane review, we further scale) categorised these comparison interventions into three groups: no  Back-related functional limitations (eg, Roland Morris Disability treatment/usual care (including placebo and education), ineffective Questionnaire, Oswestry Disability Index) interventions (electrotherapy) and other conservative treatment Integrity (psychological therapy, anti-inflammatory/analgesics, relaxation,  Trials were excluded from the review if they were judged to be manual therapy, physiotherapy, back school), according to guideline either plagiarised or published in a presumed predatory journal recommendations.10,11 For this network meta-analysis, we combined in addition to at least one other research integrity concern (eg, no treatment/usual care and ineffective interventions into one group high risk of bias, inadequate reporting) referred to as ‘minimal treatment’ node, which was used with the Data availabilitya ‘other conservative treatment’ node in the primary analyses.  Trials were excluded from the analysis if they did not have any outcome follow-up of  4 weeks Outcomes  Trials were excluded from the analysis if they did not have data The primary analyses in this study assessed the effect of treat- for analysis in either pain or functional limitations (allowing for standard deviation imputation and carry-through of sample size ments on pain intensity (eg, measured with a visual analogue scale or from baseline) numeric rating scale) and back-related functional limitations (eg, measured with the Roland Morris Disability Questionnaire12 or median was used to estimate the mean. If sample size information Oswestry Disability Index13). Pain and functional limitation outcomes was unavailable for follow-up, the sample size was carried forward were self-reported as continuous measures and analysed on the from baseline. continuous scale. To facilitate synthesis across trials and interpret- ability of the results, each trial’s pain and functional limitation Risk of bias outcome data were converted to standardised 0 to 100 (maximum) The risk of bias assessment for trials was conducted using the scales. Existing outcome data for all available follow-up time points were collected. The primary meta-analysis and meta-regression an- criteria in the Cochrane Collaboration Risk of Bias Tool (version 1).14,15 alyses used the most complete data available for each outcome We assessed potential bias related to: randomisation; treatment separately. Outcomes at the available follow-up period closest to allocation concealment; blinding of participants, care providers and short-term (post-treatment time-period  4 weeks, closest to 3 outcome assessors; drop-out rate; intention to treat; seletive months) were used for these analyses. outcome reporting; similarity at baseline; avoidance of co- interventions; compliance; and similar timing of outcome assess- When trial authors did not respond to a request for missing data, ment. Individual criteria were scored as ‘high risk’, ‘low risk’ or missing variance scores were imputed using the mean variance from ‘unclear risk’. A trial with a low risk of bias was defined as fulfilling six trials with similar populations of people with low back pain. Where or more of the 12 criteria items, and with no other fatal flaws. For data were reported as a median and interquartile range (IQR), the additional detail regarding risk of bias assessment see the published Cochrane review.3

254 Hayden et al: Exercise in chronic low back pain Data analysis heterogeneity and incoherence in the primary analyses. First, exploratory network meta-analyses were conducted with the nine The primary analyses included descriptive analyses, pairwise specific comparison treatments uncategorised. Next, included trials meta-analyses and network meta-analyses for the 11 specified exer- were restricted to more homogeneous measures of pain intensity cise types, and two comparison categories: minimal treatment and (including trials using the visual analogue scale or numerical rating other conservative treatment. We used two software packagesa,b for scale only), and functional limitations (including trials using the data preparation and analyses. Statistical code is provided in Roland Morris Disability Questionnaire12 or Oswestry Disability In- Appendix 4 on the eAddenda. dex13 only). Third, we excluded trials judged as having high risk of bias. Fourth, we excluded trials judged to have an improbable or Descriptive analyses outlying mean outcome if the absolute difference between any ex- The individual and summary characteristics of all included trials ercise group and any comparison group over all available follow-ups was greater than a predetermined threshold of 30/100 for pain and were described using the appropriate descriptive statistics, and rea- 20/100 for functional limitations, selected based on clinical sons for exclusion of trials that were included in the related Cochrane judgement. review were documented. For all sensitivity and exploratory analyses, we assessed changes Effectiveness of exercise types in the interpretation of pairwise treatment effects for the analysis Meta-analyses were conducted for each pairwise comparison of compared to the primary network meta-analysis and changes in overall model heterogeneity and incoherence observed. treatments for which there were data available. DerSimonian and Laird random-effects models were used to estimate the pooled effects Certainty of the evidence of intervention (MD and 95% CI) and measures of heterogeneity (s2, I2) for each comparison. The sample size of the repeated group The CINeMA web application (which adapts GRADE domains to was evenly split across instances where a comparison group network meta-analysis) was used to evaluate confidence in findings contributed to multiple observations within the same meta-analysis from the primary network meta-analyses due to: risk of bias within (ie, in comparisons involving trials with more than two groups). comparisons, publication bias, indirectness, imprecision, heteroge- Egger’s test was used for asymmetry when  10 intervention groups neity and incoherence.20 A detailed description of the assessment contributed to a comparison. process is provided in Appendix 7 on the eAddenda. Network meta-analysis A clinically important difference in outcome between low back We conducted network meta-analyses to estimate the effects pain treatments was interpreted as a difference in pain of 15 points out of 100, and difference in function of 10 points. These were of the interventions on pain and functional limitation outcomes calculated as the smallest worthwhile effects based on a 30% reduc- separately. A frequentist inconsistency model16 was fitted using tion in outcome,21 from the average baseline pain (50.9, 95% CI 49.1 to contrast-based linear mixed-effects modelling. The models included: 52.8), and average baseline functional limitations (38.9, 95% CI 35.8 to fixed-effect parameters for the effects of intervention; baseline 42.0) for included trials. Differences were considered statistically outcome value and their interaction; and random-effects terms to significant at the 5% level. We defined five categories of results account for correlation between observed effects in trials with more (categories 1 to 4 favouring one treatment): 1. Clinically important than two groups. We specified random-effects terms using a difference; 2. Moderate difference compatible with a clinically compound symmetric covariance structure with rho = 0.5.17 important difference; 3. Moderate difference; 4. Small difference; and 5. No difference. We assessed changes in interpretation of results for Assessment of network transitivity sensitivity analyses with an algorithm to identify changes in the We assumed that all participants in the included trials were interpretation of the effect direction, size and compatibility with a clinically important difference (Appendix 8 on the eAddenda). We equally likely to be randomised to any of the interventions in the defined important changes in interpretation of the results for sensi- observed trials (ie, we assumed that the transitivity assumption was tivity analyses from primary analysis results based on the number of plausible). Nonetheless, we considered participant setting, duration changes in the results category: no or one change = no concerns; two of the low back pain episode, radicular symptoms and baseline or three changes = some concerns; and four or more changes = major outcome values as modifiers of treatment effects. We also considered concerns. exercise treatment intervention dose/intensity, delivery format and presence of additional interventions to be effect modifiers. Accord- Results ingly, the distribution of the population variables across network comparisons was assessed (Appendix 5 on the eAddenda) and each This review provides an up-to-date assessment of the effective- treatment effect modifier covariate was modelled (Appendix 6 on the ness of exercise treatment of chronic non-specific low back pain. It eAddenda). This assessment of network transitivity and considering included 249 randomised trials (24,486 participants), with 217 of theoretical mechanisms of modification led us to adjust for baseline these trials (87%) providing sufficient data for meta-analysis (20,969 outcome values in the primary analyses, and explore the impact of participants at baseline) (Figure 1). Citations for the included studies exercise dose and additional co-interventions in sensitivity analyses. are presented in Appendix 9 on the eAddenda. The reasons that trials were excluded from this study are presented in Appendix 10 on the Assessment of incoherence eAddenda. There was a total of 507 treatment groups in the included We evaluated incoherence (ie, agreement between direct and in- trials. In total, 126 trials compared exercise to non-exercise compar- isons, and 91 compared only groups receiving different types of ex- direct evidence) of the pain and function outcome networks globally ercise interventions (Figure 2). and evaluated local incoherence for each treatment comparison using side-splitting18 and by evaluating statistical incoherence of the Characteristics of included trials network separately in every closed loop.19 In the loop-specific approach, loops formed only by multi-arm trials were excluded and Table 1 describes summary characteristics of included trial pop- correlation induced by multi-arm trials was mitigated by dropping ulations and Appendix 11 (on the eAddenda) provides detailed in- the direct comparison with the largest number of trials from multi- formation about each trial included in this analysis. The included arm trials when it appeared in a particular loop. Local incoherence trials were mostly conducted in healthcare settings (56%, 122 trials); was considered to be statistically significant if loop-specific 95% 65 trials (30%) were from general populations or mixed settings and confidence intervals did not include zero. Sensitivity analyses Four types of sensitivity analyses were conducted to explore the impact of methodological decisions and to further explore residual

Research 255 Records identified through database searches (n = 21,714) • Medline (n = 4,072) • CINAHL (n = 4,969) • Embase (n = 4,165) • PsycINFO (n = 339) • CENTRAL (n = 5,186) • SportDiscus (n = 746) • PEDro (n = 467) • ICTRP (n = 421) • ClinicalTrials.gov (n = 404) Records after duplicates removed (n = 12,833) Additional records identified Unique records screened by title (n = 13,087) Excluded (n = 4,969) from other sources (n = 204) Records screened by title and abstract (n = 8,118) Excluded (n = 7,015) Records awaiting further Records screened by full text (n = 1,103) assessment (n = 172) Excluded (n = 568) • not full text (n = 97) • not randomised trial (n = 129) • not chronic non-specific pain (n = 165) • not exercise versus comparison (n = 79) • no eligible outcome measures (n = 3) • duplicate record (n = 95) Linked publications (n = 40) Excluded (n = 74) • not chronic non-specific pain (n = 29) • not exercise versus comparison (n = 14) • no eligible outcome measures (n = 3) • multiple research integrity concerns (n = 30) Studies included (n = 249) Trials with data available for meta-analysis (n = 217) Figure 1. Flow of trials through the review. 21 trials (10%) were from occupational settings. The number of par- groups. There were between five and 110 exercise groups available for ticipants in included trials ranged from 10 to 722, with a median each of the predefined 11 types of exercise categories. The most sample size of 70 participants (IQR 43 to 120). The trial participants common type of exercise investigated was core strengthening (30%), reported a mean pain severity at baseline of 44/100 (95% CI 43 to 45), followed by treatment groups comprising mixed types of exercises with 67 included trials (31%) describing participants with chronic low (ie, three or more types) (26%) and general strengthening exercises back pain of moderate symptom duration (mean duration 12 weeks (12%). We classified 45% of exercise treatments as ‘back specific’ and to 3 years) and the same number including participants with longer 29% as ‘whole body’ exercises. The exercise program design was duration of chronic low back pain of . 3 years (83 trials did not classified as individualised for 19%, partially individualised for 36% specify the duration of chronic low back pain). The median time point and standardised for 45%. The exercise programs that were investi- of the short-term outcome follow-up reported in this project was 12 gated were mostly delivered in a supervised group setting (40%) or weeks (IQR 7 to 12). with individual healthcare provider supervision (39%). The programs had a median treatment time of 12 hours (IQR 8 to 20) delivered over Of the 217 included trials, 108 (50%) were judged to be at risk of a median period of 8 weeks (IQR 6 to 12). Most of the exercise selection bias, 172 (79%) were judged to be at risk of performance treatment groups included other additional interventions (57%), the bias, 158 (73%) were judged to be at risk of detection bias, 38 (18%) most common being advice/education (31%). were judged to be at risk of attrition bias, 12 (6%) were judged to be at risk of reporting bias and 150 (69%) were assessed as susceptible to The 138 comparison groups included in the trials were categorised other potential sources of bias. The full risk of bias assessments for as minimal treatment (62%) and effective treatment/unclear effec- each included trial are provided in Appendix 12 on the eAddenda. tiveness (38%). The minimal treatment category included comparison groups that provided placebo, no treatment or usual care (62%), ed- Characteristics of treatments ucation only (33%) or electrotherapy (6%). The effective or unclear effectiveness category of comparisons included treatments such as Included trials had between one and four exercise treatment manual therapy (27%), mixed non-exercise physical therapy (48%) groups, totalling 369 exercise groups across the 217 trials. Table 2 and back school (12%). A detailed description of comparison groups is describes the summary characteristics of the exercise treatment provided in Appendix 11.

256 Hayden et al: Exercise in chronic low back pain Number of comparison Number of exercise Number of Table 1 groups in study groups in study studies Characteristics of included trial populations. 4321 1234 Characteristic All trials (n = 217) 79 Population source, n (%) 122 (56) healthcare 21 (10) 77 occupational 65 (30) general or mixed 1 (, 1) 33 other not specified 8 (4) 9 Total participants at baseline (n) 20,969 8 pooled 70 (43 to 120; 10 to 722) per study, median (IQR; range) 4 44 (43 to 45) Age of participants (y), mean (95% CI) 44 (41 to 47) 3 Male participants (%), mean (95% CI) Category of low back pain, n (%) 189 (87) 3 27 (12) chronic 1 (, 1) 1 mixed chronic not specified 17 (8) Figure 2. Breakdown of exercise treatment and comparison treatment group numbers Recurrent pain, n (%) 200 (92) for trials available for the meta-analyses presented in this study (217 included trials). yes no 67 (31) Comparative effectiveness of exercise treatment types Pain duration, n (%) 67 (31) moderate 83 (38) The overall certainty of the evidence for pain intensity outcomes long 51 (49 to 53) was judged to be low for 64% of pairwise comparisons (49 of 77) and not specified moderate for 36% of comparisons (28 of 77). For functional limitation Baseline pain severity (0 to 100), mean (95% CI) 0 (0) outcomes, the overall certainty of the evidence was low for 16% of Leg pain or neurological symptoms, n (%) 90 (42) pairwise comparisons (12 of 77), moderate for 82% of comparisons all 73 (34) (63 of 77) and high for 3% of comparisons (2 of 77). some 54 (25) none Pain intensity outcomes not specified 507 In total, 198 trials measured pain intensity outcomes (including Treatment groups (n) 2 (2 to 3; 2 to 5) pooled 466 treatment groups, 17,534 participants), with 166 providing data per study, median (IQR; range)a 369 (73) for meta-analyses (399 groups, 15,553 participants). This meant that exercise groups, n (%)a 138 (27) direct evidence was available for 52 of the possible 77 treatment comparison groups, n (%)a comparisons in the network. The number of trials available for each Outcomes assessed, n (%)b 202 (93) comparison ranged from one (24 participants) to 18 (1,739 partici- pain intensity 199 (92) pants). Two exercise types – core strengthening and mixed exercise functional limitations 30 (14) type – and the minimal treatment comparison type had direct com- work status 67 (31) parison pairings available with all other treatment types. Egger’s test health-related quality of life 66 (30) suggested possible publication bias in 29% of meta-analyses with adverse outcomes 40 (18)  10 trials available (two of seven meta-analyses). global perceived recovery Follow-up periods, n (%) 34 (16) Direct pairwise meta-analyses: Pairwise meta-analyses for all immediate (, 6 weeks) 168 (77) comparisons are provided with detailed individual trial-level infor- short term (6 to 12 weeks) 116 (54) mation, summarised in Appendix 5 and presented in forest plots in moderate (13 to 47 weeks) 63 (29) Appendix 13 on the eAddenda. The data available for pain outcomes long term ( 48 weeks) directly comparing each of the 11 exercise types with minimal treatment ranged from one trial (110 participants) for flexibility ex- a For primary network meta-analyses, exercise groups of the same exercise type and ercises to 17 trials (1,614 participants) for mixed exercises. The mean difference in pain intensity at short-term follow-up from direct comparison groups of the same type were combined, giving 422 treatment groups (head-to-head) evidence favoured each exercise type compared with minimal treatment for all types other than flexibility exercises (MD (290 exercise groups, 132 comparison groups). 5.0, 95% CI 28.9 to 18.9; one trial favouring minimal treatment). The b Trial reported outcome measurement. largest mean differences in pain intensity from direct evidence (for comparisons with more than one trial providing data) were observed Direct pairwise evidence was available for nine of the 11 exercise for Pilates (MD 221.8, 95% CI 229.6 to 214.1, 11 trials, 800 partici- types compared with other effective treatments, with evidence pants, I2 = 91.1), McKenzie therapy (MD 214.1, 95% CI 227.7 to 20.4, ranging from one trial (75 participants) for functional restoration to two trials, 170 participants, I2 = 71.2), stretching (MD 214.0, 95% 18 trials (1,429 participants) for core strengthening exercises. Three or CI 221.1 to 26.8, six trials, 354 participants, I2 = 54.6), general fewer trials provided direct evidence for all except core strength- strengthening (MD 213.4, 95% CI 220.6 to 26.2, nine trials, 433 ening, mixed exercise type, general strengthening and aerobic exer- participants, I2 = 79.8) and core strengthening exercises (MD 212.8, cises compared with other effective treatments. The mean difference 95% CI 217.8 to 27.9, 17 trials, 1,545 participants, I2 = 82.3). in pain intensity at short-term follow-up compared with other effective treatments from direct evidence favoured Pilates (MD 218.3, 95% CI 223.4 to 213.1, two trials, 161 participants), McKenzie therapy (MD 217.6, 95% CI 233.6 to 21.5, two trials, 170 participants, I2 = 86.6) and functional restoration (MD 229.0, 95% CI 241.0 to 217.0, one trial, 75 participants). Moderate to no differences in treatment effect were observed for other exercise types (Appendix 5 and Appendix 13). Network meta-analysis (primary analyses): The network was well connected, with 66% of comparisons having direct evidence (Figure 3). The network meta-analysis results comparing exercise types and minimal treatment are presented in Figure 4. The mean difference in pain intensity at short-term follow-up favoured each exercise type compared with minimal treatment, with mean treat- ment effects ranging from 26.8 to 218.7 (decreased pain intensity with exercise treatment). Outcomes for the following exercise types were compatible with a clinically important treatment effect

Research 257 Table 2 exercises (MD 26.5, 95% CI 210.6 to 22.5) and flexibility exercises Description of exercise group characteristics in included trials (217 trials, 368 exercise (MD 25.9, 95% CI 216.2 to 4.3), with low certainty evidence overall. groups). McKenzie therapy was moderately more effective than stretching (MD 28.0, 95% CI 215.8 to 20.2), aerobic exercises (MD 27.9, 95% Characteristic Exercise groups (n = 368)a CI 215.0 to 20.9) and flexibility exercises (MD 27.4, 95% CI 219.1 to 4.4), with moderate certainty evidence overall. Functional restoration Types of exercise, n (%) 110 (30) exercises were moderately more effective than stretching (MD 27.8, core strengthening 96 (26) 95% CI 215.5 to 20.2), aerobic exercises (MD 27.8, 95% CI 214.2 mixed exercise types 44 (12) to 21.3) and flexibility exercises (MD 27.4, 95% CI 219.1 to 4.4), with general strengthening 25 (7) low to moderate certainty evidence. There was a small to moderate aerobic 24 (7) difference in pain outcomes comparing Pilates to McKenzie therapy, Pilates 17 (5) core strengthening exercises and functional restoration, with Pilates stretching 15 (4) appearing more effective (Figure 5). other specific exercises 13 (4) yoga 10 (3) Network meta-analysis (exploring all comparisons): The network functional restoration meta-analysis comparing the effectiveness of all 11 exercise types and McKenzie therapy 9 (3) all nine specific comparison treatment types included 162 pairwise flexibility 5 (1) comparisons with direct and indirect data from 166 trials (399 groups, 15,553 participants) for pain intensity outcomes (Appendix 14 Specificity of exercise, n (%) 108 (29) on the eAddenda). There was direct evidence for 81 of the possible whole body 167 (45) treatment comparisons (50%). Eleven of 41 direct comparisons with back specific 64 (17) two or more trials (27%) had I2 values . 75%. Sparseness of the both network led to wide confidence intervals for some treatment com- not specified 29 (8) parisons; however, we observed moderate differences compatible with a clinically important difference in pain outcomes for Pilates Exercise program design, n (%) 69 (19) compared with all other conservative treatments (most likely MD individualised 131 (36) range 29.3 to 217.5). partially individualised 166 (45) standardised 2 (, 1) Functional limitation outcomes not specified 19 (5) The network for this study included data about functional limi- Primary delivery format, n (%) 25 (7) tation outcomes from 187 trials (433 treatment groups, 16,926 par- independent exercise 147 (40) ticipants). Direct and indirect pairwise data from 149 trials (355 independent exercise with follow-up 143 (39) treatment groups, 14,220 participants) about functional limitation group supervision 34 (9) outcomes were available. This represented direct evidence for 49 of individual supervision 12.0 (8.0 to 20.0; 0.3 to 156.0) the possible 77 treatment comparisons. The number of trials available not specified for pairwise comparisons ranged from one (24 participants) to 16 8 (6 to 12; 1 to 52) (1,749 participants). Two exercise types – core strengthening and Duration of intervention (hr), median mixed exercise type – and minimal treatment comparison type had (IQR; range) (284 of 368) 120 (33) direct comparison pairings available with all other treatment types. Duration of intervention (wk), median 246 (67) Direct meta-analyses that included two or more trials (57%, 28 of 49 (IQR; range) (361 of 368) 2 (, 1) direct meta-analyses) had statistical heterogeneity ranging from 0 to Dose of intervention, n (%) 93.1% measured with the I2 statistic. Eleven pairwise meta-analyses 208 (57) had I2 values , 50% and nine had values . 75%. Egger’s test sug- high dose ( 20 hours) 136 (37) gested possible publication bias in 13% of meta-analyses with  10 low dose (, 20 hours) trials available (one of eight meta-analyses). not specified 24 (7) Other additional interventions, n (%) Direct pairwise meta-analyses: Pairwise meta-analyses for each yes exercise type compared with minimal treatment are presented in no Figure 4. For functional limitation outcomes, data available for each of not specified the 11 exercise types compared with minimal treatment ranged from one trial (90 participants) for functional restoration exercises to 16 a For the primary network meta-analyses, exercise groups of the same exercise were trials (1,749 participants) for mixed exercise type. The mean differ- combined giving 290 exercise groups. ence in functional limitations at short-term follow-up from direct evidence favoured each exercise type compared with minimal compared with minimal treatment (ie, 95% CI suggesting . 15-point treatment. The largest reduction in functional limitations from direct improvement in pain outcomes): Pilates, McKenzie therapy, func- evidence was observed for Pilates (MD 213.1, 95% CI 218.6 to 27.7, 10 tional restoration, core strengthening and other specific exercise trials, 780 participants, I2 = 88.3). types. The network meta-analysis mean difference estimate for Pilates (MD 218.7, 95% CI 224.4 to 213.1) was most likely to be Direct pairwise evidence was available for nine of the 11 exercise compatible with clinically important improvement compared with types compared with other effective treatments, with evidence minimal treatment. The certainty of this evidence was judged to be ranging from one trial (28 participants) for ‘other’ specific exercises to moderate overall. 14 trials (1,716 participants) for mixed exercise type. Four or fewer trials provided direct evidence for all except core strengthening and The mean difference in pain intensity at short-term follow-up mixed exercise type compared with other effective treatments. The compared with other effective treatments from network meta- mean difference in functional limitations at short-term follow-up analysis was compatible with a clinically important difference for compared with other effective treatments from direct evidence fav- Pilates (MD 211.2, 95% CI 217.2 to 25.3) and moderate treatment oured functional restoration (MD 225.3, 95% CI 238.1 to 212.5, one effect for McKenzie therapy (MD 27.3, 95% CI 214.1 to 20.5) and trial, 75 participants), McKenzie therapy (MD 216.1, 95% CI 219.5 functional restoration (MD 27.2, 95% CI 213.9 to 20.4). Small to no to 212.8, one trial, 45 participants) and Pilates (MD 28.3, 95% treatment effects were observed for other exercise types compared CI 220.9 to 4.4, two trials, 161 participants, I2 = 72.9). Small to no with other effective treatments. The certainty of this evidence was differences in treatment effect were observed for other exercise types. judged to be low to moderate overall. Pairwise meta-analyses for all comparisons are summarised in Comparing the various exercise types with each other, Pilates, core strengthening, McKenzie therapy and functional restoration exercises had larger improvement in pain intensity compared with several other exercise types. Pilates was more effective than all other exercise types, compatible with a clinically important difference in pain in- tensity, compared with stretching exercises (MD 211.9, 95% CI 218.2 to 25.6), aerobic exercises (MD 211.8, 95% CI 217.6 to 26.0), flexi- bility exercises (MD 211.3, 95% CI 222.5 to 0.0), yoga (MD 210.7, 95% CI 217.4 to 23.9) and mixed exercises (MD 29.9, 95% CI 215.2 to 24.7), with low to moderate certainty evidence overall. Core strengthening exercises were moderately more effective than stretching exercises (MD 26.6, 95% CI 211.8 to 21.3), aerobic

258 Hayden et al: Exercise in chronic low back pain Figure 3. Network plot presenting the trial data contributing evidence comparing exercise treatment types for short-term outcomes: A. pain intensity: 166 trials, 359 groups, 15,553 participants. B. functional limitations: 149 trials, 322 groups, 14,220 participants. The size of the nodes represents how many times the exercise appears in any comparison about that treatment and the width of the edges represents the total sample size in the comparisons it connects. Appendix 5 and provided with detailed individual trial-level infor- exercise types and minimal treatment found mean difference in mation in Appendix 13. functional limitations at short-term follow-up favoured each exercise type compared with minimal treatment, with mean treatment effects Network meta-analysis (primary analysis): The network plot for ranging from 23.4 to 211.7 (decreased functional limitations with functional limitation outcomes primary analysis was well connected, exercise treatment), with treatment effects for four of 11 exercise with 62% of comparisons having direct evidence and no disconnected types most compatible with a clinically important effect (ie, 95% CI nodes (Figure 3). The network meta-analysis results comparing

Research 259 Exercise type Pain intensity (0 to 100) Functional limitations (0 to 100) Study data available Network meta-analysis estimate Study data available Network meta-analysis estimate (studies; groups; MD (95% CI) (studies; groups; MD (95% CI) participants) participants) Core strengthening 61; 69; 2,476 –13.4 (–17.2 to –9.6) 56; 63; 2,320 –6.6 (–9.0 to –4.3) Mixed types 48; 59; 2,478 –8.8 (–12.5 to –5.1) 46; 56; 2,473 –4.4 (–6.8 to –2.0) General strengthening 30; 36; 1,102 –10.9 (–14.7 to –7.0) 29; 35; 1,065 –5.0 (–7.5 to –2.4) Aerobic 18; 25; 845 –6.9 (–10.9 to –2.9) 16; 22; 626 –4.3 (–7.0 to –1.6) Pilates 17; 20; 719 –18.7 (–24.4 to –13.1) 15; 17; 667 –10.2 (–13.8 to –6.6) Stretching 13; 15; 401 –6.8 (–11.9 to –1.8) 11; 13; 353 –3.4 (–6.4 to –0.3) Other specific exercises 14; 18; 820 –10.7 (–15.8 to –5.5) 11; 14; 558 –3.5 (–6.7 to –0.4) Yoga 10; 15; 542 –8.1 (–14.1 to –2.0) 11; 16; 698 –5.3 (–9.0 to –1.6) Functional restoration 9; 11; 459 –14.7 (–21.3 to –8.1) 7; 9; 304 –7.4 (–11.9 to –2.9) McKenzie 8; 9; 428 –14.8 (–21.4 to –8.2) 7; 7; 419 –11.7 (–16.7 to –6.7) Flexibility 4; 4; 163 –7.5 (–17.5 to 2.5) 4; 4; 151 –11.0 (–17.2 to –4.8) –30 –20 –10 0 10 –30 –20 –10 0 10 Favours exercise Favours control Favours exercise Favours control Figure 4. Summary network meta-analysis results for each exercise type compared with minimal treatment comparisons for short-term outcomes: pain intensity, functional limitations. The hashed line indicates clinically important difference. suggesting . 10-point improvement in functional limitation out- observed in functional limitation outcomes for Pilates, McKenzie comes): McKenzie therapy (MD 211.7, 95% CI 216.7 to 26.7), flexi- therapy and flexibility exercises compared with most other conser- bility exercises (MD 211.0, 95% CI 217.2 to 24.8), Pilates (MD 210.2, vative treatments (most likely mean difference range 25.3 to 212.4). 95% CI 213.8 to 26.6) and functional restoration exercise (MD 27.4, 95% CI 211.9 to 22.9) (Figure 4). The certainty of this evidence was Sensitivity analyses judged to be moderate to high overall. We summarised sensitivity analysis results for exercise types The mean difference in functional limitations at short-term compared with minimal treatment (Appendix 15 on the eAddenda; follow-up compared with other effective treatments from network full output for sensitivity analyses is available on request). There were meta-analysis was compatible with a clinically important difference minimal changes in the interpretation of pain and functional limita- for McKenzie therapy (MD 27.1, 95% CI 212.1 to 22.1) and functional tion outcomes for exercise types compared with minimal treatment restoration (MD 26.4, 95% CI 212.7 to 20.1), and moderate treatment for each of the sensitivity analyses (all 66 sensitivity analysis results effect for Pilates (MD 25.5, 95% CI 29.4 to 21.7). The certainty of this judged to have no concerns). Overall, for network meta-analyses for evidence was judged to be moderate. Small to no treatment effects pairwise sets of two treatments there were small changes to the size were observed for other exercise types compared with other effective of the network meta-analytic effects (higher or lower); however, treatments. interpretation of results was unchanged for 98% of pain sensitivity analyses (226 of 231 pairwise comparison network meta-analysis Comparing the various exercise types with each other, McKenzie results) and 99% of functional limitation sensitivity analyses (228 of therapy, Pilates and flexibility exercises had larger improvement in 231 pairwise comparison network meta-analysis results). functional limitations, compatible with a clinically important differ- ence, compared with other exercise types investigated. There were no Exploring potential effect modifying characteristics (Appendix 6), observed differences in functional limitation outcomes between there were no clear differences in the interactions of symptom these three exercise types at short-term follow-up (Figure 5). duration and inclusion of participants with leg pain or neurological symptoms (covariates) with intervention effects across comparisons. Network meta-analysis (exploring all comparisons): Network meta- A high dose of most exercise treatments appeared to reduce pain and analysis comparing the effectiveness of the 11 exercise types and nine functional limitation outcomes more than low dose; more so for specific comparison treatment types included 162 pairwise compar- Pilates compared with minimal treatment than for other types, isons of direct and indirect data (Appendix 14). Functional limitation although the observed confidence intervals overlapped. Co- outcome data from 149 trials (355 treatment groups, 14,220 partici- interventions appeared to improve effectiveness of most exercise pants) provided evidence for 74 of the possible treatment compari- types for pain and function but not for stretching exercises. Network sons (46%). Ten of 37 direct comparisons with two or more trials meta-analysis models including adjustment for dose and additional (27%) had I2 values . 75%. Sparseness of the network led to imprecise co-interventions as treatment effect modifying covariates were con- estimates for some treatment comparisons; however, moderate dif- ferences compatible with clinically important differences were

260 Hayden et al: Exercise in chronic low back pain A Core strengthening –4.6 Mixed (> 2 (–7.7 to –1.5) types) –2.5 2.1 General (–6.2 to 1.1) (–1.7 to 5.8) strengthening –6.5 –1.9 –4.0 Aerobic (–8.3 to 0.4) (–10.6 to –2.5) (–6.0 to 2.3) 5.3 9.9 7.9 11.8 Pilates (0.4 to 10.2) (4.7 to 15.2) (2.2 to 13.6) (6.0 to 17.6) –6.6 –2.0 –4.0 –0.1 –11.9 Stretching (–9.7 to 1.7) (–5.9 to 5.7) (–18.2 to –5.6) (–11.8 to –1.3) (–7.4 to 3.5) –2.8 1.9 –0.2 3.8 –8.1 3.8 Other specific (–7.7 to 2.2) (–3.2 to 6.9) (–5.4 to 5.0) (–1.6 to 9.1) (–14.5 to –1.6) (–2.7 to 10.4) exercises –5.4 –0.7 –2.8 1.2 –10.7 1.2 –2.6 Yoga (–10.9 to 0.2) (–6.4 to 4.9) (–8.9 to 3.3) (–4.9 to 7.2) (–17.4 to –3.9) (–5.5 to 8.0) (–9.0 to 3.8) 1.3 5.9 3.8 7.8 –4.1 7.8 4.0 6.6 Functional (–4.8 to 7.3) (–0.1 to 11.9) (–2.1 to 9.8) (1.3 to 14.2) (–11.4 to 3.3) (0.2 to 15.5) (–3.2 to 11.2) (–1.0 to 14.3) restoration 1.4 6.0 4.0 7.9 –3.9 8.0 4.2 6.8 0.1 McKenzie (–5.0 to 7.9) (–0.6 to 12.7) (–3.0 to 10.9) (0.9 to 15.0) (–11.2 to 3.3) (0.2 to 15.8) (–3.5 to 11.8) (–1.3 to 14.9) (–8.3 to 8.6) –5.9 –1.3 –3.4 0.6 –11.3 0.6 –3.2 –0.6 –7.2 –7.4 Flexibility/ mobilising (–16.2 to 4.3) (–11.6 to 8.9) (–13.8 to 7.0) (–9.7 to 10.8) (–22.5 to 0.0) (–10.4 to 11.7) (–14.2 to 7.8) (–12.0 to 10.9) (–18.9 to 4.5) (–19.1 to 4.4) –13.4 –8.8 –10.9 –6.9 –18.7 –6.8 –10.7 –8.1 –14.7 –14.8 –7.5 Minimal treatment (–17.2 to –9.6) (–12.5 to –5.1) (–14.7 to –7.0) (–10.9 to –2.9) (–24.4 to –13.1) (–11.9 to –1.8) (–15.8 to –5.5) (–14.1 to –2.0) (–21.3 to –8.1) (–21.4 to –8.2) (–17.5 to 2.5) Effective –5.9 –1.3 –3.3 0.6 –11.2 0.7 –3.1 –0.5 –7.2 –7.3 0.1 treatment (–9.8 to –2.0) (–5.2 to 2.6) (–7.4 to 0.7) (–3.5 to 4.8) (–17.2 to –5.3) (–4.7 to 6.1) (–8.6 to 2.3) (–6.9 to 5.9) (–13.9 to –0.4) (–14.1 to –0.5) (–10.1 to 10.2) B Core Mixed (> 2 General Aerobic strengthening types) strengthening 5.9 Pilates –2.2 0.5 –0.7 (2.0 to 9.7) (–4.1 to –0.3) (–1.9 to 2.9) (–3.6 to 2.3) –0.9 –6.8 Stretching –1.7 –0.1 5.2 (–4.6 to 2.8) (–10.6 to –3.0) (–4.0 to 0.6) (–2.9 to 2.6) (1.5 to 8.9) –0.8 –6.6 0.2 Other specific –2.3 5.7 –1.6 exercises (–5.0 to 0.3) (2.3 to 9.1) (–5.1 to 1.9) (–4.5 to 3.0) (–10.9 to –2.4) (–3.9 to 4.2) 1.7 3.5 –1.0 –1.4 1.0 –4.9 1.9 (–2.3 to 5.7) Yoga (0.3 to 6.7) (–4.3 to 2.2) (–4.8 to 2.0) (–2.8 to 4.8) (–9.1 to –0.6) (–2.1 to 5.9) 3.9 –3.2 –0.9 0.3 3.1 –2.8 4.0 (–1.1 to 8.9) 2.1 Functional (–6.3 to –0.2) (–4.2 to 2.4) (–3.5 to 4.1) (–1.3 to 7.6) (–7.8 to 2.3) (–1.0 to 9.0) (–3.0 to 7.2) restoration 8.2 –3.1 0.9 2.4 7.4 1.5 8.3 (2.6 to 13.7) 6.4 4.3 McKenzie (–6.3 to 0.2) (–2.6 to 4.3) (–1.6 to 6.5) (2.2 to 12.6) (–4.1 to 7.2) (2.8 to 13.8) (0.7 to 12.2) (–1.9 to 10.5) 7.5 –1.3 3.0 6.7 6.7 0.8 7.6 (0.6 to 14.3) 5.7 3.6 –0.7 Flexibility/ (–4.8 to 2.1) (–1.3 to 7.2) (1.6 to 11.9) (0.1 to 13.3) (–6.2 to 7.9) (0.9 to 14.4) (–1.3 to 12.7) (–3.9 to 11.1) (–8.5 to 7.1) mobilising –3.5 0.8 7.3 6.0 –4.3 –10.2 –3.4 (–6.7 to –0.4) –5.3 –7.4 –11.7 –11.0 Minimal (–3.4 to 5.0) (2.4 to 12.2) (–0.5 to 12.5) (–7.0 to –1.6) (–13.8 to –6.6) (–6.4 to –0.3) treatment 1.1 (–9.0 to –1.6) (–11.9 to –2.9) (–16.7 to –6.7) (–17.2 to –4.8) 5.1 6.6 –5.0 (–2.4 to 4.6) Effective (0.3 to 9.9) (0.3 to 12.8) (–7.5 to –2.4) 0.3 –5.5 1.2 –0.7 –2.8 –7.1 –6.4 treatment (–2.5 to 3.1) (–9.4 to –1.7) (–2.1 to 4.6) (–4.6 to 3.3) (–7.4 to 1.8) (–12.1 to –2.1) (–12.7 to –0.1) 4.4 –4.4 –0.3 (–1.9 to 10.6) (–6.8 to –2.0) (–3.1 to 2.4) –6.6 0.2 (–9.0 to –4.3) (–2.3 to 2.7) –2.0 (–4.5 to 0.5) Clinically important difference No Clinically important difference favouring column treatment difference favouring row treatment Figure 5. League table of network meta-analysis results for all comparisons between exercise and non-exercise interventions. Effects are expressed as the mean difference (95% CI) between interventions on short-term outcomes A. Pain intensity, 0 to 100 (198 trials, 466 groups, 17,534 participants). B. Functional limitations, 0 to 100 (187 trials, 433 groups, 16,926 participants). Blue shading indicates that the intervention listed in the column is more effective than that in the row, whereas orange shading indicates the intervention listed in the row is more effective. The depth of shading indicates the size of the treatment effect and likely compatibility with a clinically important effect (Appendix 7 describes the shading algorithm). ducted. Adjusting for these exercise characteristic covariates did not CINeMA framework, is reported in Appendix 17 on the eAddenda. modify the conclusions about either Pilates or stretching exercises Result characteristics that lowered our certainty in the evidence (Appendix 16 on the eAddenda). In these pain and function models, included major concerns about within-study bias in 31% of compar- Pilates remained consistently better than other exercise types; core isons, and some concerns in the remaining 69% of comparisons (53 of strengthening exercise types were consistently better than other 77). Heterogeneity was common, with some concerns or major con- exercise types, McKenzie therapy was not consistently better, and cerns identified in all pairwise comparisons (40 of 77 had some stretching exercises were consistently worse than other types and concerns, 37 had major) for pain outcomes and 91% of functional non-exercise comparisons. limitation outcome comparisons (62 of 77 had some concerns, eight had major). Examining the coherence of the network, direct and in- Certainty of the network meta-analysis evidence direct estimates of treatment effect were found to agree with respect to direction and size; interpretation of direct and indirect meta- The overall certainty of the evidence available for each compari- analysis results differed for three of 52 comparisons (pain) (one son in the primary network meta-analyses, assessed using the major concern, two some concerns) and eight of 49 comparisons

Research 261 (functional limitations) (all some concerns) with direct data available. characteristics; this decreased confidence in the available evidence. Loop-specific coherence was observed for 92 of 101 loops with direct Incomplete reporting of trial and population characteristics, differing data available for pain intensity outcomes, and 79 of 90 loops with opinions about treatment type classifications, and potential misclas- direct data available for functional limitation outcomes (Appendix 18 sification of exercise types and population characteristics are addi- on the eAddenda). The global test of incoherence did not show sig- tional limitations. It should also be noted that the interventions that nificant incoherence for pain outcomes (c2 = 49.8, 74 df, p = 0.99) or appeared to be the most effective were also interventions that are for functional limitation outcomes (c2 = 80.2, 68 df, p = 0.15). There costly to deliver and to ‘purchase’ for patients. It is possible that our was minimal observed change in heterogeneity or incoherence for results were conflated with other factors related to higher socioeco- any sensitivity analyses (Appendix 19 on the eAddenda). nomic status in these patient groups (eg, physical labour, other healthcare access and health status). Discussion Our study also had several strengths. It used a comprehensive It is believed that this study is the largest network meta-analysis search, robust selection criteria and detailed data extraction pro- in the field of low back pain. The primary network meta-analysis cesses. The large number of included trials increased generalisability showed pain and functional limitation outcomes compatible with and provided a dense, well-connected network for analyses. It clinically important effects for Pilates, McKenzie therapy and func- considered quality/integrity issues when selecting trials, used best tional restoration exercise types. methods for analyses, and comprehensively explored methodological issues and potential effect modifiers. The results are reported in a All exercise types, other than stretching exercises when adjusted complete and transparent way. for dose and additional co-interventions, were consistently more effective than minimal care and most other comparison treatments Incoherence, often a concern in network meta-analysis because it for reducing pain intensity and improving functional limitations in relies on the assumption of transitivity, did not seem to be a problem people with chronic low back pain. Pilates exercises were found to be with the evidence available, as evidenced by global and local tests of more effective for pain outcomes than other comparison treatments incoherence. While there was some evidence of exercise character- and other types of exercise treatment. In the primary models, istic effect modifying covariate differences (dose and additional co- McKenzie therapy and Pilates were found to more effective for interventions) that may indicate lack of transitivity in the network, functional limitation outcomes than other comparisons and other adjusting for these covariates did not modify our overall conclusions. types of exercise treatment. McKenzie therapy was no longer better than other comparisons when adjusted for dose and additional co- Implications for clinical practice interventions. High dose of most exercise treatments appeared to reduce pain and functional limitation outcomes more than low dose, This network meta-analysis provides further support for exercise and the addition of co-interventions appeared to improve effective- treatment being effective and guideline-recommended care. It found ness of most exercise types for pain and functional limitation that most exercise types are more effective than minimal treatment; outcomes. to increase dose, patients should be encouraged to perform the ex- ercise that they enjoy and will take part in consistently. We found Comparison with other studies evidence that Pilates, McKenzie therapy and functional restoration were more effective than other types of exercise treatment for There are many systematic reviews of specific exercise treatment reducing pain intensity and functional limitations; the observed ef- types for persistent low back pain, including published Cochrane fect of McKenzie therapy may be related to higher dose and/or co- reviews,22–25 upcoming Cochrane protocols26–28 and many recent interventions in included trials. If the observed pain and function systematic reviews that have been published outside of the Cochrane outcomes align with the patient’s goals, it may be appropriate to Library.29–36 These reviews have included five to 29 trials on treat- recommend these types of exercise programs if they are available and ments for chronic low back pain, a small proportion of the trials financially feasible for the patient. included here, with more focused review questions and selection criteria, and some differences in methods. Similar to the current Implications for research findings, the Cochrane reviews have reported low to moderate quality evidence that the specific exercise type investigated produced small One considerable advantage of network meta-analysis and meta- to moderate improvements in outcomes compared with minimal regression over traditional aggregate meta-analysis is the ability to intervention. However, these Cochrane reviews reported limited ev- consider all evidence from trials. This includes direct and indirect idence of important differences compared with other types of exer- evidence, as well as other population and delivery characteristics. The cise and other conservative treatments. strengths of network meta-analysis were used to include a large number of trials and borrow strength from indirect evidence. How- Owen and colleagues (2019) conducted a similar overarching ever, potential challenges are similar to traditional meta-analyses network meta-analysis of exercise treatment for chronic low back where the evidence is limited by the reporting and quality of the pain to compare effectiveness of different exercise types.37 That re- trials included. There is a continued goal to improve the planning, view included 89 trials comparing exercise and non-exercise treat- conduct and reporting of the trials. ments; they used stricter inclusion criteria and a less comprehensive search than our review. The considerable difference in number of Future trials on exercise treatment should evaluate other relevant included trials makes it challenging to compare findings. However, patient outcomes aligned with the proposed mechanisms of exercise Owen and colleagues similarly reported low-quality evidence for treatment and cost-effectiveness, as this will guide individuals and their network meta-analysis due to risk of bias and heterogeneity, clinicians in their choice for the best treatment. and reported Pilates to be effective for improving pain. This review explored several sources of heterogeneity, including Strengths and limitations population and treatment characteristics, outcomes and methodo- logical characteristics. Statistical heterogeneity was moderate to Our results were limited by the quality of the included evidence. substantial in most analyses, not reduced by sensitivity, subgroup The evidence available for most treatment comparisons was judged analyses or considering suspected treatment effect modifiers. Exam- to be of moderate certainty, due to within-study risk of bias and ining individual-level patient characteristics and their relationship heterogeneity. Unexplained heterogeneity remained, despite explo- with effectiveness of exercise with individual participant data was ration of multiple population, exercise and methodological beyond the scope of this review; however, this would be important to consider in future research.

262 Hayden et al: Exercise in chronic low back pain In conclusion, our findings were compatible with moderate to 9. Furlan AD, Malmivaara A, Chou R, Maher CG, Deyo RA, Schoene M, et al. 2015 clinically important treatment effect for Pilates and McKenzie therapy Updated method guideline for systematic reviews in the Cochrane Back and Neck compared with minimal treatment, other effective treatments and Group. Spine. 2015;40:1660–1673. other exercise types. This analysis should help guide primary care clinicians in their patient management and referral practices. 10. de Campos TF. Low back pain and sciatica in over 16s: assessment and manage- ment NICE Guideline [NG59]. J Physiother. 2017;63:120. What was already known on this topic: Clinical practice guidelines recommend exercise as first-line management for 11. Hauk L. Low Back Pain: American College of Physicians Practice Guideline on chronic low back pain. However, there is limited evidence to Noninvasive Treatments. Am Fam Physician. 2017;96:407–408. support the use of one type of exercise or program characteristic over another. 12. Roland M, Fairbank J. The Roland-Morris Disability Questionnaire and the What this study adds: Most exercise types were more Oswestry Disability Questionnaire. Spine. 2000;25:3115–3124. effective than minimal treatment for pain and functional limita- tion outcomes. Pilates, McKenzie therapy and functional resto- 13. Fairbank JC, Couper J, Davies JB, O’Brien JP. The Oswestry low back pain disability ration were more effective than other types of exercise questionnaire. Physiotherapy. 1980;66:271–273. treatment for reducing pain intensity and functional limitations. 14. Higgins J. Cochrane handbook for systematic reviews of interventions. Version 5.1. Footnotes: a Stata SE v14.2, Stata Corp, College Station, USA. 0 [updated March 2011]. 2021. The Cochrane Collaboration. www cochrane- b R software V4.0.3, R Core Team, Vienna, Austria. handbook org eAddenda: Appendices 1 to 19 can be found online at https://doi. 15. Higgins JP, Altman DG, Gøtzsche PC, Jüni P, Moher D, Oxman AD, et al. The org/10.1016/j.jphys.2021.09.004. Cochrane Collaboration’s tool for assessing risk of bias in randomised trials. BMJ. 2011;343:d5928. Ethics approval: Not applicable. Competing interests: Nil. 16. Law M, Jackson D, Turner R, Rhodes K, Viechtbauer W. Two new methods to fit Source(s) of support: Funding was received from Research Nova models for network meta-analysis with random inconsistency effects. BMC Med Scotia to carry out analyses related to the Cochrane review. Res Methodol. 2016;16:1–14. Acknowledgements: Dr. Yamato is supported by the São Paulo Research Foundation, FAPESP (2019/10330-4), São Paulo, Brazil. Dr. 17. Higgins JP, Whitehead A. Borrowing strength from external trials in a meta-anal- Saragiotto is supported by the São Paulo Research Foundation, FAPESP ysis. Stat Med. 1996;15:2733–2749. (2020/14231-8), São Paulo, Brazil. Provenance: Not invited. Peer reviewed. 18. Dias S, Welton NJ, Caldwell DM, Ades AE. Checking consistency in mixed treatment Correspondence: Jill A Hayden, Department of Community Health comparison meta-analysis. Stat Med. 2010;29:932–944. & Epidemiology, Dalhousie University, Halifax, Canada. Email: [email protected] 19. Chaimani A, Salanti G. Visualizing assumptions and results in network meta- analysis: the network graphs package. Stata J. 2015;15:905–950. References 20. Nikolakopoulou A, Higgins JPT, Papakonstantinou T, Chaimani A, Del Giovane C, 1. Chou R, Qaseem A, Snow V, Casey D, Cross JT, Shekelle P, et al. Diagnosis and Egger M, et al. CINeMA: An approach for assessing confidence in the results of a treatment of low back pain: a joint clinical practice guideline from the American network meta-analysis. PLoS Med. 2020;17:e1003082. College of Physicians and the American Pain Society. Ann Intern Med. 2007;147:478–491. 21. Ferreira ML, Herbert RD, Ferreira PH, Latimer J, Ostelo RW, Grotle M, et al. The smallest worthwhile effect of nonsteroidal anti-inflammatory drugs and physio- 2. Toward Optimized Practice (TOP) Low Back Pain Working Group. Evidence- therapy for chronic low back pain: a benefit-harm trade-off study. J Clin Epidemiol. informed primary care management of low back pain: Clinical practice guideline. 2013;66:1397–1404. 3rd edition; 2017. Edmonton, AB: Toward Optimized Practice. Available from: https://actt.albertadoctors.org/CPGs/Lists/CPGDocumentList/LBP-guideline.pdf. 22. Saragiotto BT, Maher CG, Yamato TP, Costa LO, Menezes Costa LC, Ostelo R, et al. Accessed 1 September, 2021. Motor control exercise for chronic non-specific low-back pain. Cochrane Database Syst Rev. 2016. 3. Hayden J, Ellis J, Ogilvie R, Malmivaara A, van Tulder M. Exercise therapy for chronic low back pain. Cochrane Database Syst Rev. 2021;CD009790:In 23. Schaafsma FG, Whelan K, van der Beek AJ, van der Es-Lambeek LC, Ojajärvi A, Press. Verbeek JH. Physical conditioning as part of a return to work strategy to reduce sickness absence for workers with back pain. Cochrane Database Syst Rev. 2013. 4. Ferreira ML, Smeets RJ, Kamper SJ, Ferreira PH, Machado LA. Can we explain het- erogeneity among randomized clinical trials of exercise for chronic back pain? A 24. Wieland LS, Skoetz N, Pilkington K, Vempati R, D’Adamo CR, Berman BM. Yoga meta-regression analysis of randomized controlled trials. Phys Ther. 2010;90:1383– treatment for chronic non-specific low back pain. Cochrane Database Syst Rev. 2017. 1403. 25. Yamato TP, Maher CG, Saragiotto BT, Hancock MJ, Ostelo R, Cabral CMN, et al. 5. Crequit P, Trinquart L, Yavchitz A, Ravaud P. Wasted research when systematic Pilates for low back pain. Cochrane Database Syst Rev. 2015. reviews fail to provide a complete and up-to-date evidence synthesis: the example of lung cancer. BMC Med. 2016;14:8. 26. Liu H, Yao K, Zhang J, Li L, Wu T, Brox JI, et al. Sling exercise therapy for chronic low- back pain. Cochrane Database Syst Rev. 2013. 6. Elliott JH, Turner T, Clavisi O, Thomas J, Higgins JP, Mavergames C, et al. Living systematic reviews: an emerging opportunity to narrow the evidence-practice gap. 27. Machado L, Lin CWC, Clare H, van Tulder MW. The McKenzie method for (sub)acute PLoS Med. 2014;11:e1001603. non-specific low-back pain (protocol). Cochrane Database Syst Rev. 2012. 7. Hayden JA, van Tulder MW, Tomlinson G. Systematic review: strategies for using 28. Machado L, van Tulder MW, Lin CWC, Clare H, Hayden JA. The McKenzie method exercise therapy to improve outcomes in chronic low back pain. Ann Intern Med. for chronic non-specific low-back pain (protocol). Cochrane Database Syst Rev. 2005;142:776–785. 2012. 8. Hutton B, Salanti G, Caldwell DM, Chaimani A, Schmid CH, Cameron C, et al. The 29. Alhakami AM, Davis S, Qasheesh M, Shaphe A, Chahal A. Effects of McKenzie and PRISMA extension statement for reporting of systematic reviews incorporating stabilization exercises in reducing pain intensity and functional disability in in- network meta-analyses of health care interventions: checklist and explanations. dividuals with nonspecific chronic low back pain: a systematic review. J Phys Ther Ann Intern Med. 2015;162:777–784. Sci. 2019;31:590–597. 30. Namnaqani FI, Mashabi AS, Yaseen KM, Alshehri MA. The effectiveness of McKenzie method compared to manual therapy for treating chronic low back pain: a systematic review. J Musculoskelet Neuronal Interact. 2019;19:492–499. 31. Niederer D, Mueller J. Sustainability effects of motor control stabilisation exercises on pain and function in chronic nonspecific low back pain patients: a systematic review with meta-analysis and meta-regression. PLoS One. 2020;15:e0227423. 32. Shi Z, Zhou H, Lu L, Pan B, Wei Z, Yao X, et al. Aquatic exercises in the treatment of low back pain: a systematic review of the literature and meta-analysis of eight studies. Am J Phys Med Rehabil. 2018;97:116–122. 33. Vanti C, Andreatta S, Borghi S, Guccione AA, Pillastrini P, Bertozzi L. The effec- tiveness of walking versus exercise on pain and function in chronic low back pain: a systematic review and meta-analysis of randomized trials. Disabil Rehabil. 2019;41:622–632. 34. Wewege MA, Booth J, Parmenter BJ. Aerobic vs. resistance exercise for chronic non- specific low back pain: a systematic review and meta-analysis. J Back Musculoskelet Rehabil. 2018;31:889–899. 35. Zhang Y, Loprinzi PD, Yang L, Liu J, Liu S, Zou L. The beneficial effects of traditional Chinese exercises for adults with low back pain: a meta-analysis of randomized controlled trials. Medicina. 2019;55:118–141. 36. Zou L, Zhang Y, Yang L, Loprinzi PD, Yeung AS, Kong J, et al. Are mindful exercises safe and beneficial for treating chronic lower back pain? a systematic review and meta-analysis of randomized controlled trials. J Clin Med. 2019;8:628–654. 37. Owen PJ, Miller CT, Mundell NL, Verswijveren SJ, Tagliaferri SD, Brisby H, et al. Which specific modes of exercise training are most effective for treating low back pain? Network meta-analysis. Br J Sports Med. 2020;54:1–12.

Journal of Physiotherapy 67 (2021) 298–307 j o u r n a l h o m e p a g e : w w w. e l s ev i e r. c o m / l o c a t e / j p hy s Research The Transmural Trauma Care Model can be implemented well but some barriers and facilitators should be considered during implementation: a mixed methods study Suzanne H Wiertsema a, Marianne H Donker b, Johanna M van Dongen b, Edwin Geleijn a, Frank W Bloemers c, Raymond WJG Ostelo b,d, Vincent de Groot a a Amsterdam UMC, Vrije Universiteit Amsterdam, Department of Rehabilitation Medicine, Amsterdam Movement Sciences, Amsterdam, Netherlands; b Vrije Universiteit Amsterdam, Department of Health Sciences, Faculty of Science, Amsterdam Movement Sciences, Amsterdam, Netherlands; c Amsterdam UMC, Vrije Universiteit Amsterdam, Department of Trauma Surgery, Amsterdam Movement Sciences, Amsterdam, Netherlands; d Amsterdam UMC, Vrije Universiteit Amsterdam, Department of Epidemiology and Biostatistics, Amsterdam Movement Sciences, Amsterdam, Netherlands KEY WORDS ABSTRACT Trauma Questions: What is the reach, dose delivered, dose received and fidelity of the Transmural Trauma Care Fractures Model (TTCM)? What are the barriers and facilitators associated with the implementation of the TTCM? Rehabilitation Design: Mixed-methods process evaluation with quantitative evaluation of the extent to which the TTCM Process evaluation was implemented as intended and qualitative evaluation of barriers and facilitators to its implementation. Framework method Participants: Focus group participants included trauma patients, trauma surgeons, hospital-based physio- Implementation therapists and primary care network physiotherapists. Outcome measures: Implementation was assessed with reach, dose delivered, dose received and fidelity. Data analysis: A framework method was used to analyse the focus groups and the ‘constellation approach’ was used to categorise barriers and facilitators into three categories: structure, culture and practice. Results: The TTCM’s reach was 81%, its dose delivered was 99% and 100%, and its dose received was 95% and 96% for the multidisciplinary TTCM consultation hours at the outpatient clinic for trauma patients and the primary care network physiotherapists, respectively. Various fidelity scores ranged from 66 to 93%. Numerous barriers and facilitators associated with the implementation of the TTCM were identified and categorised. Conclusion: This process evaluation showed that the TTCM was largely implemented as intended. Furthermore, various facilitators and barriers were identified that need to be considered when implementing the TTCM more widely. Differences were found among stakeholders but they were generally of the opinion that if the barriers were overcome, the quality of care and patient satisfaction were likely to improve significantly after implementing the TTCM. Registration: NTR5474. [Wiertsema SH, Donker MH, van Dongen JM, Geleijn E, Bloemers FW, Ostelo RWJG, de Groot V (2021) The Transmural Trauma Care Model can be implemented well but some barriers and facilitators should be considered during implementation: a mixed methods study. Journal of Physiotherapy 67:298–307] © 2021 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 To improve the rehabilitation process of trauma patients, the Transmural Trauma Care Model (TTCM) was developed and imple- Traumatic injury accounts for 9.6% of global mortality.1–3 Major mented at a Level 1 trauma centre in the Netherlands. The TTCM is an trauma, in particular, is the most important cause of long-term advanced rehabilitation model consisting of a continuous feedback functional limitations in adults younger than 45 years.4 Traumatic loop, in which a multidisciplinary hospital-based team supervises a injury mainly affects younger individuals and, as a consequence, ac- network of primary care physiotherapists during the rehabilitation counts for the highest number of lost productive years of life process of trauma patients.10 Evidence on the effectiveness and cost- compared with other conditions.5 During the last two decades, effectiveness of the TTCM compared with regular care has been mortality due to traumatic injury has decreased considerably by be- published elsewhere.11,12 The results showed that the TTCM was tween 15 and 25%.6–8 Consequently, the focus of trauma care has associated with better patient outcomes and that it may be consid- moved from reducing mortality to improving quality of life and ered cost-effective compared with regular care, depending on the outcome, which in turn has resulted in a growing interest in decision-makers’ willingness to pay and the probability of cost- improving the quality of trauma rehabilitation.9 effectiveness that they perceive as acceptable. https://doi.org/10.1016/j.jphys.2021.08.017 1836-9553/© 2021 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 299 It is recommended that process evaluations be conducted along- extent to which trauma patients actively engaged in the intervention side clinical trials, as process evaluations can provide important in- (ie, the proportion of included TTCM participants who eventually formation for interpreting their results.13–16 Moreover, process visited their scheduled multidisciplinary TTCM appointment at the evaluation results can be used to further improve the intervention outpatient clinic and the proportion of included TTCM participants and facilitate the transition of research evidence into clinical prac- who eventually visited the PCNP they were referred to). Fidelity was tice.17,18 In the field of trauma treatment and trauma rehabilitation, defined as the extent to which the intervention was delivered as process evaluations are rarely performed. One mixed-method study planned (ie, the extent to which the intervention protocol was fol- assessed the relationship between participant-related factors and lowed by the various care providers). Various fidelity scores were adherence to osteoporosis medication, vitamin D supplementation, assessed (eg, whether a secure email was sent from the hospital- and participation in physical activity in older patients with fragility based physiotherapist (HBP) to the PCNP after each multidisci- fractures.19 Moreover, a recent focus group study among trauma pa- plinary TTCM visit and vice versa). A complete overview of all the tients suggested that inadequate aftercare negatively influenced criteria used to assess fidelity is presented in Table 2. To explore trauma patients’ perceived quality of life at least 1 year after the four process evaluation components, data were collected from the trauma.20 It is noteworthy that the majority of patients participating patients’ electronic patient records (eg, the number of secure emails, in this study were aged . 65 years and it is believed that process the use of standardised referral forms, the setting of individual evaluations in younger patients with traumatic injury are lacking. functional goals) and from the care providers’ schedules. Even though results suggest that the TTCM could improve patient Qualitative assessment outcomes and healthcare efficiency,11,12 it is less clear how to implement this model in practice. It is unknown how the TTCM could Barriers and facilitators are defined as ‘factors that hampered or be implemented in other trauma regions with their own structures, enhanced the implementation of an intervention’, respectively.24 To cultural norms/values and practical routines.21 These considerations explore the barriers and facilitators associated with the imple- led to the current study and process evaluation being performed. mentation of the TTCM, homogeneous focus groups were conducted among trauma patients, trauma surgeons, HBPs and PCNPs. Partici- Therefore, the research questions for this mixed-methods process pants were selected purposively. This sampling method allows re- evaluation were: searchers to use their own judgement to select individuals who are able to provide in-depth information pertaining to the research 1) What is the reach, dose delivered, dose received and fidelity of the questions. This study used focus groups instead of in-depth in- TTCM? terviews because more in-depth information can be obtained from a group context, in which members influence each other (‘the whole is 2) What are the barriers and facilitators associated with the imple- greater than the sum of its parts’).25,26 Another strength of focus mentation of the TTCM? groups is that they provide access to shared social meaning and norms, and how these are enacted.27 This study used homogeneous Method focus groups to avoid existing professional and/or personal hierarchy structures (eg, between surgeons and physiotherapists and patients) Design to influence the results. Homogeneous focus groups create a safe environment in which participants are more likely to speak freely and This process evaluation was conducted alongside a clinical trial openly.28 evaluating the effectiveness and cost-effectiveness of the TTCM compared with regular practice.22 The clinical trial was conducted at Focus groups were conducted at a time and location convenient to the outpatient clinic for trauma patients of a Level 1 trauma centre the participants. Prior to the focus groups, participants were assured (Amsterdam UMC, location VUmc, Amsterdam, the Netherlands). In of confidentiality and were asked to provide informed consent. Focus this study, operatively and non-operatively treated trauma patients groups were guided by two experienced qualitative researchers who with at least one fracture and aged  18 years were included. This were familiar with the TTCM, but were not involved in the TTCM as a process evaluation used a mixed-methods design. Quantitative pro- care provider. During each focus group, three round table discussions cess evaluation data were collected from the intervention group were held: the first aimed to identify possible facilitators, the second participants’ electronic patient records to assess the TTCM’s reach, aimed to identify possible barriers and the third aimed to comple- dose delivered, dose received and fidelity. Additionally, qualitative ment and validate the barriers and facilitators identified in the first data were collected by conducting focus groups among various two rounds of discussion. During all round table discussions, a topic stakeholders to explore the barriers and facilitators related to the list was used as a guide. Every round started by asking participants to implementation of the TTCM. independently write down facilitators and barriers on adhesive notes to frame the personal perspective of the participants and avoid The Transmural Trauma Care Model groupthink. Subsequently, participants were free to discuss all topics they considered important. All focus groups were audiotaped and The TTCM consists of four linked components shown in Table 1. A transcribed verbatim. more detailed description of the TTCM is provided elsewhere.10 Data preparation and analysis Quantitative assessment Quantitative analysis The extent to which the TTCM was implemented as intended was To assess the reach, dose delivered, dose received and fidelity of explored by assessing four process evaluation components of Linnan and Steckler, including its reach, dose delivered, dose received and fi- the TTCM, summary statistics were prepared in commercial delity.23 Reach was defined as the proportion of the intended target softwarea. audience that participated in the intervention (ie, the proportion of potentially eligible trauma patients who eventually participated in Qualitative analysis the TTCM during the clinical trial period). Dose delivered was defined Focus group data were analysed using the framework method. as the number of units of the intervention delivered (ie, the propor- tion of intended multidisciplinary TTCM consultation hours that This is a hierarchical, matrix-based method for ordering and syn- eventually took place at the outpatient clinic and the proportion of thesising qualitative data.29,30 The framework method enables sys- included TTCM participants who were eventually referred to a tematic exploration of the data while simultaneously maintaining an primary-care network physiotherapist, PCNP). Dose received was the effective and transparent examination path.31 In this study, an ‘analytical framework’ was constructed iteratively from the research aims, existing literature and the data derived from the focus groups.

300 Wiertsema et al: Process evaluation of the Transmural Trauma Care Model Table 1 Components of the Transmural Trauma Care Model (TTCM). TTCM component Description A multidisciplinary TTCM consultation hour at During the trauma patients’ outpatient visits, the trauma surgeon evaluated the bone and wound healing process and the outpatient clinic for trauma patients by a acted as the chief consultant. The hospital-based physiotherapist assessed physical function and acted as case manager trauma surgeon and a trauma-specialised throughout the rehabilitation process. During a shared decision-making process, the trauma surgeon, hospital-based hospital-based physiotherapist. physiotherapist and patient determined whether and when physiotherapy in primary care was required. Coordination and individual goal setting The hospital-based team coordinated the patients’ rehabilitation process in primary care by repeatedly defining individual for each patient by the multidisciplinary treatment goals in close cooperation with the patient. To support this process, 10 rehabilitation protocols were developed hospital-based team. for the most common fractures (eg, hip fractures, tibial plateau fractures). These protocols were customised for each individual patient by the hospital-based physiotherapist. A network of specialised primary care This Network Trauma Rehabilitation VUmc consisted of 40 physiotherapists, all of whom worked in a primary care practice physiotherapists. in the region of Amsterdam and were specifically trained to rehabilitate trauma patients (www.traumarevalidatie.nl). Secure email traffic between the hospital-based A secure email system (Zorgmail), developed for healthcare professionals, was linked to both the electronic patient physiotherapist and the primary care network records of the hospital-based physiotherapist and the primary care network physiotherapist, so that regular physiotherapist. communication was guaranteed throughout the rehabilitation process. For constructing the analytical framework, the ‘constellation culture and practice, a transition of the constellation has occurred. In perspective’, as described by Van Raak, was used as the theoretical general, the driving force of change is the sense of urgency for change framework.21 The constellation approach will be described briefly by ‘key actors’ within a constellation.32 These actors initiate and push below, followed by a stepwise description of the way the framework for change on the structural, cultural and practical levels.34 To achieve method was used for analysing the data. a transition, the relevant actors need to develop a collective sense of urgency to change and they need to develop new competences The constellation approach as a theoretical framework: The (knowledge, attitude and skills). Scaling up involves implementing constellation approach has its origins in organisational research and the results of niche experiments in the existing structure, culture and assumes that a healthcare system consists of so-called constellations, practice.35 defined as a set of interrelated practices and relevant, interrelated, structuring elements that together both define and fulfil a function in Stepwise procedure of the framework method to construct an the larger system. The needs of healthcare systems are diverse and analytical framework: This study iteratively constructed an therefore the system consists of a multitude of nested complement- ‘analytical framework’ from the literature and the focus group data. ing and competing constellations and (sub)constellations.32 Within a For building this analytical framework, the ‘constellation perspec- constellation there is a continuous interaction between the three tive’, as described above, was used as the ‘theoretical framework’. elements of the ‘structure, culture and practice triplet’, introduced by The first step of the framework method30 consisted of a verbatim Rotmans and Loorbach in 200933 and adapted by Van Raak (Appendix ‘transcription’ of the audiotaped focus groups, followed by the 1).21 Structure consists of the physical structures and resources, second step, which was ‘familiarisation’ with the data by listening enforced regulations and legal rights, economic resources and other and rereading the transcripts. The third step was ‘coding’ and was material elements that structure behaviour within a constellation. aimed at classifying the data in such a way that it could be sys- Culture refers to the paradigms, norms and values and other imma- tematically compared with other parts of the data set. For this terial elements that structure behaviour in practices. Practice involves purpose, all transcripts were manually coded line by line by the typical routines on the operational level, which are undertaken by applying a paraphrase or label to relevant parts of the text (the the actors within the constellation. Actors are the individuals (eg, ‘code’) using word processing softwareb. This started with open patients, physicians, managers) or groups (eg, insurance companies, coding, meaning that anything that could be possibly relevant was departments) who work or act in a certain constellation. Please notice coded independently by two of the researchers (SW and MD). that actors are not part of a constellation, but shape its culture and Subsequently, both researchers independently generated descrip- structure (and vice versa) through practice. tive themes and subthemes. The fourth step was the ‘development of an analytical framework’, in which codes were grouped into For the TTCM, several nested constellations can be recognised, for categories on the structural, cultural and practical level of the example: the outpatient clinic for trauma patients on the one hand theoretical framework (ie, the constellation approach). Subse- and the primary care network practices on the other hand. Moreover, quently, the final codes were developed through discussion be- both the hospital and the primary care network practices are part of a tween the two researchers. During these discussions, similar codes bigger constellation, in which insurers and policymakers act in a were grouped into main topics and subtopics in order to identify certain structure and culture. Dynamics, such as those created by the important themes (ie, selective coding), resulting in the initial implementation of the TTCM, provide an opportunity for change. analytical framework. Then, both researchers independently coded When the change process leads to a fundamental shift in structure, Table 2 Process evaluation components, definitions and scores. Assessment component Definition Score (%) 81 Reach The proportion of potentially eligible trauma patients who eventually participated in the TTCM during the clinical trial. 99 Dose delivered The proportion of intended multidisciplinary TTCM consultation hours that eventually took place at the outpatient clinic for trauma Dose received patients. 100 The proportion of included TTCM participants who were eventually referred to a primary care network physiotherapist. 95 Fidelity The proportion of included TTCM participants who eventually visited their scheduled multidisciplinary TTCM appointment at the outpatient clinic for trauma patients. 96 The proportion of included TTCM participants who eventually visited the primary care network physiotherapist they were referred to. 89 The use of the standardised referral form for the primary care network physiotherapist. 93 A secure email was sent from the hospital-based physiotherapist to the primary care network physiotherapist after each multidisciplinary TTCM visit. 76 A secure email was sent from the primary care network physiotherapist to the hospital-based physiotherapist prior to each multidisciplinary TTCM visit. 89 Individual functional goals were set for the patient by the multidisciplinary hospital-based team during each multidisciplinary TTCM visit. 66 Specific feedback from the primary care network physiotherapist to the hospital-based team whether the functional goals have been achieved or not (and why).

Research 301 all remaining transcripts of the focus groups using the initial Box 1. Recommendations for implementation and scaling up framework. Subsequently, they met again and, following discus- of the TTCM. sion, revised the initial framework to incorporate new and refined codes. The process of refining, applying and refining the analytical  Form a steering group with all stakeholders to take everyone’s framework was repeated until no new codes were generated. interests into account. Note that the process of developing the analytical framework was  Describe clear organisational structures for care providers at a combined deductive and inductive approach. On one hand, pre- the outpatient clinic and for primary care network selected themes and codes of Van Raaks’ theoretical framework physiotherapists (eg, communication pathways and templates were used (deductive), while on the other hand, themes and codes for standardised documentation). were generated from the data (inductive). The final framework con- sisted of 16 themes, clustered into six categories (facilitators and  Describe duties and responsibilities of the participating care barriers on the structural, cultural or practical level). In the fifth step, providers in a manual and organise training courses for the called ‘indexing’, both researchers systematically went through each primary care network physiotherapists. transcript again, highlighting each meaningful passage of text and selecting and attaching an appropriate code from the final analytical  Organise reflection meetings with stakeholders (homogeneous framework. At this stage, each code was assigned an abbreviation for as well as heterogeneous) per trauma centre and respect local easy identification (eg, FST1 = Facilitator Structural Theme 1). differences. Indexing involves the comparison of data within and between focus groups.  Arrange an appropriate and structural embedded reimbursement system for the hospital-based physiotherapist, The sixth step is called ‘charting’, in which a spreadsheet was used who acts as case manager within the TTCM. to generate a framework matrix. During this stage, data were sum- marised by category and subsequently categorised into the matrix, per participant was 4.7 (range 2 to 10). In total, 407 multidisciplinary followed by adding illustrative and interesting quotes from partici- TTCM appointments were scheduled, of which 387 visits eventually pants in the focus groups. During the seventh step, ‘interpretation of took place. Participants did not show up or cancelled their appoint- the data’, the framework matrix was used to interpret the data ments due to not having complaints anymore (n = 4), being ill (n = 10) together with some notes that were made during the focus groups and other reasons (n = 6). Thus, the dose received of the multidisci- and the coding process. This interpretation process was an iterative plinary TTCM visits was 95% (387 of 407). The proportion of partici- process and relied on a consultation between both researchers about pants who eventually visited the PCNP they were referred to was 80 the relevance and strength of a theme. The intensity, frequency, of 83, which made the dose received of this component of the TTCM persuasiveness and contrast with which statements were made by at 96%. the participants determined the value that was given to them. To ensure rigor and credibility of the findings, another researcher (JvD) Fidelity reviewed the generated matrix and checked whether the selected The extent to which the intervention protocol was followed by the quotes were of relevance to the themes. Disagreements were resolved by discussion. To guarantee quality of study reporting, the COnsoli- various care providers was expressed in terms of several fidelity dated criteria for REporting Qualitative research (COREQ) checklist scores, all of which are shown in Table 2. Fidelity scores ranged from was used.36 66% (ie, specific feedback from the PCNP to the HBP about whether functional goals were achieved) to 93% (ie, secure email was sent Results from the HBP to the PCNP after each visit). Quantitative results Qualitative results An overview of the quantitative results of the process evaluation Participants are presented in Table 2 and will be briefly discussed below. In total, 28 potential participants were purposively selected and Reach. Of the 1,993 trauma patients who were assessed for eligi- invited to take part in the focus groups, including six trauma patients, bility at the outpatient clinic for trauma patients between June 2014 six trauma surgeons, five HBPs and 11 PCNPs. Of them, two trauma and April 2015, 103 potentially eligible patients were identified. patients, two trauma surgeons, one HBP and one PCNP declined to Reasons for not being eligible included (amongst others) sprains, participate for various reasons (eg, unwilling, unavailable). Finally, orbital fractures, bone bruises or no need for physiotherapy. Of the five homogeneous focus groups (FGs) took place, consisting of four potentially eligible patients, 20 were excluded because they did not trauma patients (FG1), four trauma surgeons (FG2), four HBPs (FG3) provide informed consent (n = 9), had no internet access (n = 2), were and 10 PCNPs (FG4 and FG5). not willing to participate (n = 4) and had other reasons (n = 5). The remaining 83 patients were allocated to the TTCM (Box 1). The reach Barriers and facilitators associated with the implementation of the TTCM of the TTCM was therefore 81% (83 of 103). Various barriers and facilitators associated with the imple- Dose delivered mentation of the TTCM were identified (Table 3). Stakeholders During the intervention period, 544 multidisciplinary TTCM generally perceived the TTCM to be an improvement from usual care, enhancing both the quality and efficiency of care. However, differ- consultation hours at the outpatient clinic were scheduled. During ences were observed among stakeholders. Below, identified barriers four of these multidisciplinary TTCM consultation hours, the HBP was and facilitators will be discussed per level of the constellation absent due to personnel shortage (n = 1), illness of a care provider approach. First, similarities and differences between the various (n = 2) and scheduling problems (n = 1). Thus, the dose delivered of stakeholders will be described, followed by the within-group differ- the multidisciplinary TTCM consultation hours was 99% (540 of 544). ences per focus group. During the intervention period, all of the 83 TTCM participants were referred to a PCNP. Thus, the dose delivered of primary care network Structural level physiotherapy was 100%. On the structural level, six overarching themes were identified, Dose received which were categorised into facilitators and barriers (Table 3). During During the intervention period, all of the 83 TTCM participants all focus groups, the ‘communication structure of the TTCM’, including its use of a secure email system and standard referral forms, visited at least two of their scheduled multidisciplinary TTCM ap- was mentioned as an important improvement compared with usual pointments at the outpatient clinic. The mean number of actual visits care.

302 Wiertsema et al: Process evaluation of the Transmural Trauma Care Model Table 3 Facilitators and barriers expressed by care-providers and patients regarding the implementation of the TTCM, related to structure, culture and practice. Quotes are from of trauma patients (P), trauma surgeons (T), hospital-based physiotherapists (HBP) and primary care network physiotherapists (PCNP). Level Theme Facilitator Barrier Illustrative quote Structure Communication  Use a secure email system from file to  Feedback from PCNP lingers with HBP  ‘There is a lot of regular email contact structure file between primary and secondary and does not reach trauma surgeon between the hospital physical thera- care (and vice versa) pist and the network physical thera- Infrastructure  The standardised email template is too pist but that does not always reach us’ and working  Use a standardised template for the standardised (T) agreements at secure email the outpatient  The occurrence of software failures  ‘Yes, the referral form has become a lot clinic  Use a standardised referral form  The incompatibility of electronic pa- more efficient. Which makes the care  The network practice receives an better. But certainly more efficient’ (T) Financial tient records in primary and secondary structures email from the HBP when a new care  ‘The next goal was very clear for trauma patient is referred everyone, for the surgeon, for the pa- Organisation of  The HBP sets functional goals for  The absence of reimbursement for the tient, for the physical therapist in the the primary trauma patients after each visit at the HBP at the outpatient clinic hospital and for us. If that succeeds, care network outpatient clinic we continue to the next goal and  Patients are not treated by the trained otherwise it will be evaluated and Training and  Trauma surgeon and HBP prepare the PCNP, due to inadequate logistic adjusted. This is a very clear structure, education consultation hour individually pathways within the network practice making the process very satisfying for everyone’ (PCNP) Guidelines  Trauma surgeon and HBP briefly  Guidelines are too detailed and do not discuss the patients prior to the apply in case of deviant course in  ‘I am usually 15 minutes/half an hour Culture Awareness of consultation hour fracture healing earlier to look at the difficult cases. It responsibilities would be nice if they (the HBP) were and leadership  Trauma surgeon and HBP regularly  There is ambiguity about ownership in already there’ (T) work together the consultation room  ‘For your own expertise it is good to  Let the network practices pay an  Care providers contradict each other work with multiple trauma surgeons, annual fee for education and but in the context of efficiency and accreditation work relationship it is better to work in regular couples’ (HBP)  Having an appropriate and up-to-date website  ‘What worries me even more is the uncertainty how the network will  Organise a two-day training for the survive without money. Because it PCNPs prior to joining the network takes a lot of time and a lot of effort to take good care of our network’ (HBP)  Organise regular meetings for PCNPs and the complete hospital team (eg,  ‘Yes, that’s partly our fault. I think we twice a year) have around 19 physical therapists working in our practice and some-  Request accreditation for PCNPs and times trauma patients are scheduled HBPs with a non-trauma physical therapist. That is a logistic problem’ (PCNP)  Availability of guidelines for the most common fractures  ‘I can imagine that we organise theme meetings with the network physical  Shared decision-making process at the therapists twice a year to discuss spe- outpatient clinic cific topics concerning our patient category, which also makes them more  Care providers share responsibility for involved’ (T) treatment options  ‘Yes, the protocols are so incredibly  Care providers at the outpatient clinic concrete that you can only use it for act as a team and are unambiguous one specific condition. You can no longer use it in case of a slightly different fracture or a deviate course of the recovery’ (PCNP)  ‘I really liked having a voice in formulating my own goals. During the visits there was time to think and talk about what is important to me, that I wanted to play tennis again. And whether it was actually achievable what I wanted. It really helped me to discuss these issues with the surgeon and the physical therapist’ (P)  ‘You really have to achieve that bal- ance, it is true that if you are very comfortable, you reinforce each other. But it is not good if the patient feels that we do not agree with each other’ (T)  ‘I also like that you can deliberate together, not out of uncertainty, but the fact that the hospital physical therapist is actively involved in the decision-making process positively affects the patient’ (T)  ‘You must be able to adapt to the sit- uation and to various trauma sur- geons’ (HBP)  ‘I think you should be able to express what you stand for at the outpatient clinic, what your vision is. of course well substantiated, but you should not be too anxious to say what you think.’ (HBP)

Research 303 Table 3 (Continued) Illustrative quote Level Theme Facilitator Barrier  ‘We are usually on the same line about normal content. That may vary, but we Awareness of  Care providers at the outpatient clinic  Care providers go across the bound- do manage that. But it is a problem if professional (trauma surgeon and HBP) take pro- aries of their profession they come outside their domain’ (T) boundaries fessional boundaries into account  ‘Discomfort arises when the hospital Job satisfaction  Increased job satisfaction for all care physical therapist makes a statement providers about non physical therapy topics’ (T) Contact between HBP and PCNP  The possibility of low- threshold con-  ‘Once, the physical therapist was not tact between HBP and PCNP via email present during my visit. The advice the and telephone, in addition to the surgeon gave me about walking with structural forms of communication crutches was not really clear for me at that point’ (P)  The PCNPs feel that they are a sub- stantial part of the care chain  ‘It’s just really nice to work this way’ (HBP) Patients’ experience  Large acceptance of care outcome by  Receiving conflicting statements the patients due to clear expectation regarding prognosis by doctors who  ‘Yes, I am really happy that the management do not work according to the TTCM network physical therapist can easily (eg, emergency department) contact the hospital physical therapist’  Patients experience a clear treatment (P) plan and strategy  The primary care physical therapist  Patients experience it as positive to be finds us quicker and easier than before treated by a specialised PCNP, having a in case they have an acute problem high level of knowledge and skills with a patient’ (HBP)  Patients are satisfied with care  ‘Yes, you are now being encouraged to contact the hospital, the threshold has Practice Practical concerns  Availability of a separate consultation  The HBP who acts as case manager is been lowered enormously’ (PCNP) at the outpatient room for the HBP poorly accessible by telephone for clinic PCNP  ‘Yes, patients receive better care. At  Trauma surgeon and HBP work over- least, that is what we usually hear; lapping at outpatient clinic that they are satisfied with the care’ (T) Practical concerns  In total there is a higher number of  The presence of quality differences at the primary referrals for members of the network between PCNPs (both in terms of  ‘Yes, there is more focus on the care network knowledge and equipment) everyday things that patients have to  Due to the specialised primary care deal with’ (T) Quality and network, some patients can rehabili-  The absence of a social media platform efficiency tate at home instead of in a clinical for HBPs and PCNPs  ‘The doctor at the emergency depart- of care rehabilitation setting ment told me very crudely that I  The lack of guarantee on a high num- would never regain full function again’ Workload  Patients with simple fractures receive ber of referrals (P) compact advice from HBP and do not Knowledge need a referral to a PCNP  Absence of awareness of the TTCM in  ‘Nobody told me that such a simple other relevant departments in the fracture could affect my daily live in  Lower administrative workload for hospital (eg, emergency department) such an enormous way’ (P) trauma surgeon  High administrative workload for HBP  ‘I really didn’t have to wait long, it was my turn quickly’ (P)  Increased level of knowledge and skills  PCNPs have increased expertise in  ‘Usually, the hospital physical thera- pist continues to explain exercises and trauma rehabilitation because they then I start up with the next patient’ treat more trauma patients (T)  Trauma surgeon and HBP at outpatient clinic learn from each other’s field  ‘The surgeon spends less time with a patient, he can proceed with the next patient, while I give some extra advice’ (HBP)  ‘For example, I see a patient who comes for wound control without the hospital physical therapist. She does something behind the computer or already examines a new patient with a knee distortion, and then I walk in later’ (T)  ‘I think the quality of the connected network practices still varies, and that is regrettable, because the patients expect a lot from such a practice’ (HBP)  ‘Well, we have to realise that we are doing something very special. This is the future of healthcare’ (PCNP)  ‘That means that you have to prepare well, and that preparation takes quite a lot of time. So the TTCM takes more time than just being present at the outpatient clinic’ (HBP)  ‘Yes, I have seen a lot of ankle fractures lately and I noticed that I now have a better view of the course and whether it deviates or not. I recognise certain patterns. I used to have more diffi- culties with that before’ (PCNP)

304 Wiertsema et al: Process evaluation of the Transmural Trauma Care Model Yes, the referral form has become a lot more efficient, which makes it to be important that the healthcare providers who are present the care better. But certainly more efficient (trauma surgeon). during the outpatient consultations are aware of the boundaries of their own discipline. They sometimes found it hard to strike a balance The most frequently mentioned barrier on the structural level in co-working with the physiotherapist at the outpatient clinic. After was ‘the absence of reimbursement for the HBP at the outpatient an adequate balance was achieved, trauma surgeons were of the clinic’. This was identified as an important barrier because it seri- opinion that the quality of care and patient satisfaction increased ously hampers broader implementation of the TTCM. Another significantly, and working closely with a HBP became one of the most facilitator that was mentioned by all focus groups on the structural important assets of the TTCM. level was the ‘availability of guidelines for the most common fractures’. Some participants, however, thought that ‘these guide- ‘I also like that you can deliberate together, not out of uncertainty, lines are too detailed and do not apply in case of a deviant course in but the fact that the hospital physical therapist is actively involved in fracture healing’. the decision-making process positively affects the patient (trauma surgeon). Different structural aspects of the TTCM were considered to be more or less important by the various stakeholders. Trauma surgeons, The HBP also perceived the ‘awareness of responsibilities and for example, were pleased with the fact that there now was ‘a clear leadership’ to be important. For them, it was at times complicated to infrastructure and clear working agreements at the outpatient clinic’. adapt to their new role and position within the existing hierarchal They liked, for example, that they could briefly discuss the list of culture of the hospital. Despite these challenges, the most important patients prior to the multidisciplinary TTCM consultation hour with asset of the TTCM according to the HBPs was the fact that ‘care the HBP. However, they did mention that feedback from the PCNPs providers at the outpatient clinic now act as a team and are sometimes lingered with the HBP and did not reach them. For the unambiguous’. HBPs, the ‘absence of reimbursement for the HBP at the outpatient clinic’ was the most important barrier, and was also mentioned as a You must be able to adapt to the situation and to various trauma barrier by most of the other stakeholders. Another frequently surgeons (HBP). mentioned barrier by the HBPs was ‘the occurrence of software fail- ures’. PCNPs indicated that they were very satisfied with the ‘use of a PCNPs most frequently indicated that they now felt like ‘a sub- standardised referral form’ and with the fact that ‘the network stantial part of the care chain’. That is, they now perceived them- practice receives an email from the HBP when a new trauma patient selves as a member of the trauma patients’ treatment team instead of is referred’. Furthermore, they highly appreciated the ‘functional working alone, which was the case before implementation of the goals they received from the HBP for trauma patients after each visit TTCM. Another facilitator that was frequently mentioned by the at the outpatient clinic’. PCNPs was ‘the possibility of low-threshold contact between HBPs and PCNPs via email and telephone, in addition to the structural The next goal was very clear for everyone, for the surgeon, for the forms of communication’. patient, for the physical therapist in the hospital and for us. If that succeeds, we continue to the next goal and otherwise it will be Trauma patients were very pleased with the existence of a ‘shared evaluated and adjusted. This is a very clear structure, making the decision-making process at the outpatient clinic’. For them, the process very satisfying for everyone (PCNP). experience of being involved in the decision-making process and having a voice in formulating their own functional goals was of great Patients were most satisfied with the fact that ‘the HBP sets importance. This is evidenced by the following quote of a partici- functional goals for trauma patients after each visit at the outpatient pating patient: clinic’. This functional goal setting provided the trauma patients with clear expectations on their recovery and their expected I really liked having a voice in formulating my own goals. During the outcome. visits there was time to think and talk about what is important to me, that I wanted to play tennis again. And whether it was actually The within-group differences on the structural level were negli- achievable what I wanted. It really helped me to discuss these issues gible, meaning that the participants of one homogeneous focus group with the surgeon and the physical therapist (patient). agreed on most themes and subthemes. However, some of the trauma patients indicated that they had Cultural level ‘received conflicting statements regarding prognosis by doctors who On the cultural level, five overarching themes were identified, do not work according to the TTCM’, including those working at the emergency department or trauma ward of the hospital. This was which were further specified in subthemes, categorised into facili- therefore considered to be an important barrier to the implementa- tators and barriers (Table 3). During all focus groups, the ‘shared tion of the TTCM. decision-making process at the outpatient clinic’ was mentioned as an important facilitator for the implementation of the TTCM. Another Within the focus groups there were minor differences among theme that was frequently mentioned during all focus groups was the stakeholders. For example, some HBPs indicated that they preferred ‘contact between the HBP and PCNP’ with ‘the possibility of low- working with the same trauma surgeon every week, while others threshold contact between HBP and PCNP via email and telephone, preferred to work with various trauma surgeons. The same applied to in addition to the structural forms of communication’ as the most the trauma surgeons. mentioned facilitator. Yes, you are now being encouraged to contact the hospital, the Practical level threshold has been lowered enormously (PCNP). On the practical level, five overarching themes were identified, The most mentioned barrier by all focus groups was that some- which were further specified in subthemes and categorised into fa- times ‘care providers contradict each other’. cilitators and barriers (Table 3). All healthcare providers indicated that they liked their ‘increased level of knowledge and skills’ resulting You really have to achieve that balance, it is true that if you are very from working with the TTCM. That is, many of them repeatedly stated comfortable, you reinforce each other. But it is not good if the patient that they learned a lot from the other healthcare providers they feels that we do not agree with each other (trauma surgeon). collaborated with. Some differences between the focus groups were noteworthy. Stakeholders differed in terms of the practical aspects of the TTCM Trauma surgeons, for example, emphasised the importance of the that they considered to be important. Trauma surgeons and HBPs ‘awareness of professional boundaries’, meaning that they perceived were of the opinion that the ‘availability of a separate consultation room for the HBPs’ would improve their way of working. Then, the physiotherapist could examine patients (eg, for function-control or

Research 305 instructions), while the trauma surgeon could proceed to the next expressed in several facilitators (eg, the ‘use of a secure email sys- patient. tem’). The ‘absence of reimbursement for the HBP at the outpatient clinic’ was identified as a main barrier at the structural level. At the For example, I see a patient who comes for wound control without cultural level, the existence of a ‘shared decision-making process at the hospital physical therapist. She does something behind the the outpatient clinic’ was found to be an important facilitator, and the computer or already examines a new patient with a knee distortion, fact that ‘care providers sometimes contradict each other’ to be a and then I walk in later (trauma surgeon). barrier. At the practical level, the ‘increased level of knowledge and skills’ was an important facilitator and the ‘absence of awareness of Trauma surgeons also indicated that they had a ‘lower adminis- the TTCM in other relevant departments’ was recognised as a barrier. trative workload’ due to the TTCM, as the HBP was now responsible In general, stakeholders were of the opinion that if the barriers were for the communication with the PCNPs. HBPs, on the other hand, overcome, the quality of care and patient satisfaction were likely to experienced a ‘higher administrative workload at the outpatient significantly improve after implementing the TTCM. clinic’. That is, all HBPs indicated that their workload increased due to their new role as case manager, but that working according to the In trauma surgery and trauma rehabilitation, process evaluations TTCM also gave them energy because they perceived it to be are rare, and therefore an appropriate substantive comparison with inspiring. the literature is difficult to perform. However, process evaluations have been described in adjoining fields. For example, a mixed- That means that you have to prepare well, and that preparation takes method study of older patients with fragility fractures was found quite a lot of time. So the TTCM takes more time than just being that assessed the relationship between patient-related factors and present at the outpatient clinic (HBP). adherence to ‘healthy bone advice’ (ie, taking osteoporosis medica- tion and participating in physical activity). The qualitative interviews The PCNPs also indicated that they gained an ‘increased level of in this study suggested that feedback from case managers helped knowledge and skills’ and ‘increased expertise in trauma rehabilita- participants understand the underlying cause of their fragility frac- tion’ due to their involvement in the TTCM. As a consequence, they tures and helped them to adhere to the advice.19 We found similar really enjoyed working according to the TTCM. results regarding the role of the HBP, who acted as case managers. Next to other components of the TTCM, having an appropriate case Yes, I have seen a lot of ankle fractures lately and I noticed that I now manager was found to be a crucial factor for successful imple- have a better view of the course and whether it deviates or not. I mentation of the TTCM. Another process evaluation, which was recognise certain patterns. I used to have more difficulties with that conducted alongside a randomised controlled trial, evaluated the before (PCNP). implementation of RESPOND.37 This is a telephone-based falls pre- vention program including person-centred education and goal However, for them, ‘the lack of guarantee of a high number of setting, designed for older patients visiting an emergency department referrals’ was an important barrier because they preferred a contin- after a fall, but not necessarily with a fracture. The results from this uous number of new referrals, perceived from a business perspective. process evaluation, in which focus groups were held with participants For trauma patients, an important barrier was the ‘absence of and interviews were conducted with clinicians, provided detailed awareness of the TTCM at other relevant departments in the hospital information to guide future implementation of RESPOND. One of the (eg, emergency department)’. As a consequence, they sometimes main findings was that implementation of the intervention was received conflicting information regarding their treatment and facilitated by the use of ‘positive and personally relevant health prognosis from physicians from other departments. messages’.37 Parts of the RESPOND intervention program are com- parable with the TTCM (eg, personal goal setting), whereas the scope The doctor at the emergency department told me very crudely that I of the TTCM differed from RESPOND (ie, trauma rehabilitation versus would never regain full function again (patient). prevention). Furthermore, a recently published focus group study among trauma patients, aiming to describe their perceived quality of The within-group differences were small for the trauma surgeons life at least one year after trauma, found that inadequate aftercare and HBPs. For the PCNPs, within-group differences were also small, negatively influenced the trauma patients’ perceived quality of life.20 but depending of the number of new referrals they received during However, in contrast to the present study, this focus group study was the intervention period, they were more or less satisfied with the of a descriptive nature and was not aimed at identifying facilitators TTCM. The within-group differences for trauma patients were and barriers of an intervention. While the aforementioned process negligible. evaluations are meaningful and important in their own field, they differ in terms of their design, population and intervention and are Discussion therefore not entirely comparable. However, they all confirm or suggest that various elements of an intervention such as the TTCM, This paper describes the results of a process evaluation exploring aiming to improve rehabilitation and outcome after (major) trauma, the extent to which the TTCM, an advanced rehabilitation model for are of great importance and that its implementation should be trauma patients, was implemented as intended, and identifies bar- evaluated quantitatively as well as qualitatively, as was done in this riers and facilitators associated with its implementation. study. The results showed that the TTCM was largely implemented as This study had several strengths. First, it is believed that it is one intended, with a moderate reach (81%), a high dose delivered (99% of the first studies to apply qualitative research methods in the field and 100%) and high dose received (95% and 96%) for the multidisci- of trauma rehabilitation. The use of a mixed-methods approach plinary TTCM consultation hours at the outpatient clinic and the enabled assessment of both the implementation of the TTCM and its primary care network physiotherapists, respectively. Moderate to associated barriers and facilitators.38 Second, five homogeneous focus high fidelity scores were found (66 to 93%), indicating the extent to groups were chosen, including a broad range of stakeholders, which which the intervention protocol was followed by the care providers. had several advantages. That is, according to the literature, five is the The fidelity scores regarding the secure email traffic from the PCNPs optimal number of focus groups for analysis.39 It is important for a to the HBP provided the most room for improvement. That is, no broad range of stakeholders to have a voice in focus groups, in order secure email was sent to the hospital in 24% of the cases and it was to obtain the maximum amount of information necessary to optimise not clearly reported whether functional goals of the patient were the possible implementation of the TTCM. Moreover, the use of ho- achieved or not in 34% of cases. mogenous focus groups created a safe environment, in which par- ticipants were most likely to speak freely and openly.28 Third, data Focus groups indicated that on the structural level, the ‘commu- derived from the focus groups were analysed systematically, using a nication structure of the TTCM’ was found to be an important theme, well-founded theoretical model (ie, the framework method).29,30 This

306 Wiertsema et al: Process evaluation of the Transmural Trauma Care Model method enabled systematic exploration of the data, while simulta- Footnotes: aSPSS software, SPSS, Chicago, USA. bWord software, neously maintaining an effective and transparent examination path.31 Microsoft, Redmond, WA, USA. Finally, to optimise the implementation of the TTCM, reflection meetings for the HBPs were held during the implementation phase of eAddenda: Appendix 1 can be found online at https://doi.org/10. the TTCM. These meetings were valuable in gaining insight to their 1016/j.jphys.2021.08.017. new role and in matching professional responsibilities and borders. Ethics approval: The Medical Ethics Committee of the Amsterdam The study also had some limitations. Participants in the focus UMC, location VUmc, assessed the present study and decided that the groups were purposively selected and participated voluntarily, which Dutch Medical Research Involving Human Subjects Act (WMO) was may have resulted in participants being more content with the TTCM not applicable (registered under number 2013.454). All participants than the average care provider and/or trauma patient. This could have gave informed consent. The trial is registered at the Dutch Trial resulted in an overestimation of positive opinions regarding the Register (NTR5474). TTCM, especially in the focus group with trauma patients (FG5). Another limitation was the absence of healthcare decision-makers Competing interests: Nil. and insurers in the focus groups; therefore, there was a lack of Source(s) of support: This work was supported by Zilveren Kruis input from a relevant group of stakeholders regarding the theme Health Insurer (grant number Z516). ‘financial structures’ on the structural level. Furthermore, more Acknowledgements: We thank all trauma patients and care pro- detailed information could probably have been obtained if interviews viders who participated in the study. Furthermore, we thank Vincent were conducted in addition to the focus groups, since in-depth in- van Vliet for transcription of the audiotaped data and for collecting terviews can provide more detailed information on specific topics.25 the quantitative data as a part of his graduation thesis. Finally, we thank Dr. Gogem Topcu, an experienced qualitative researcher from Information derived from the current process evaluation can be the School of Medicine, University of Nottingham, for giving very used to further improve the TTCM and to enable the transition of valuable advice in applying the framework method on the data. research evidence into clinical practice.17,18 The TTCM seems feasible Adopting this method consolidated the rigor of the analysis and the in practice and was implemented as intended for nearly all partici- credibility of the findings. pants (ie, appropriate reach, dose delivered, dose received and fi- Provenance: Not invited. Peer reviewed. delity). Important needs for a successful implementation of the TTCM Correspondence: Suzanne H Wiertsema, Amsterdam UMC, Vrije were ‘having an appropriate communication structure’ and ‘reim- Universiteit Amsterdam, Department of Rehabilitation Medicine, bursement for the HBP at the outpatient clinic’ on the structural level; Amsterdam Movement Sciences, Amsterdam, Netherlands. Email: the presence of a ‘shared decision-making process at the outpatient [email protected] clinic’ on the cultural level; and an ‘increased level of knowledge and skills’ on the practical level. Additionally, it is known from the liter- References ature that other important needs for successfully scaling up and deepening of a new practice include: the establishment of coalitions 1. World Health Organisation. The World Health Report 2002 - Reducing Risks, Pro- among strategically chosen parties; transparent organisational moting Healthy Life. Geneva: WHO; 2002. structures; a clear division of responsibilities; a change in mind set; and an appropriate legal and financial framework.35,40 When these 2. Lozano R, Naghavi M, Foreman K, Lim S, Shibuya K, Aboyans V, et al. Global and needs are specified, complemented with the results of the current regional mortality from 235 causes of death for 20 age groups in 1990 and 2010: a process evaluation, the recommendations for implementation and systematic analysis for the Global Burden of Disease Study 2010. Lancet. scaling up of the TTCM shown in Box 1 can be made. 2012;380:2095–2128. As mentioned above, an important limitation of the current study 3. World Health Organization. Injuries and Violence, the Facts. Geneva: WHO; 2014. was the lack of input from healthcare decision-makers and insurers. 4. Holbrook TL, Anderson JP, Sieber WJ, Browner D, Hoyt DB. Outcome after major Their input is important because a structurally embedded reim- bursement system for the HBP is required for successful imple- trauma: 12-month and 18-month follow-up results from the Trauma Recovery mentation of the TTCM. Consequently, a final recommendation for Project. J Trauma. 1999;46:765–771. discussion 771–763. future research is to include these stakeholders in the focus groups, or 5. Trauma TAAftSo. Cost of Injury. The American Association for the Surgery of to conduct semi-structured interviews with them to obtain a complete Trauma; 2017. http://www.aast.org/GeneralInformation/costofinjury.aspx. overview of facilitators and barriers for implementation of the TTCM. Accessed 2021. 6. Lansink KW, Leenen LP. Do designated trauma systems improve outcome? Curr This process evaluation showed that the TTCM was largely Opin Crit Care. 2007;13:686–690. implemented as intended. Various barriers and facilitators were found 7. MacKenzie EJ, Rivara FP, Jurkovich GJ, Nathens AB, Frey KP, Egleston BL, et al. to be associated with the implementation of the TTCM. Moreover, A national evaluation of the effect of trauma-center care on mortality. N Engl J Med. some differences were found among stakeholders, but they were 2006;354:366–378. generally of the opinion that if the barriers were overcome and a good 8. Haas B, Jurkovich GJ, Wang J, Rivara FP, Mackenzie EJ, Nathens AB. Survival working balance was achieved, the quality of care and patient satis- advantage in trauma centers: expeditious intervention or experience? J Am Coll faction would significantly improve after implementing the TTCM. Surg. 2009;208:28–36. 9. Celso B, Tepas J, Langland-Orban B, Pracht E, Papa L, Lottenberg L, et al. What was already known on this topic: Traumatic injury A systematic review and meta-analysis comparing outcome of severely injured commonly results in long-term functional limitations. The patients treated in trauma centers following the establishment of trauma systems. Transmural Trauma Care Model involves a multidisciplinary J Trauma. 2006;60:371–378. discussion 378. hospital-based team supervising a network of primary care 10. Wiertsema SH, van Dongen JM, Geleijn E, Schothorst M, Bloemers FW, de physiotherapists during the rehabilitation process of trauma pa- Groot V, et al. Evaluation of a new Transmural Trauma Care Model (TTCM) for the tients. This cost-effective model is associated with better patient rehabilitation of trauma patients: a study protocol. BMC Health Serv Res. outcomes. 2017;17:99. What this study adds: The Transmural Trauma Care Model can 11. Wiertsema SH, van Dongen JM, Geleijn E, Huijsmans RJ, Bloemers FW, de Groot V, be implemented with good fidelity. Stakeholders recommend et al. Cost-Effectiveness of the Transmural Trauma Care Model (TTCM) for the that greater fidelity could be achieved with a steering group, Rehabilitation of Trauma Patients. Int J Technol Assess Health Care. 2019:1–10. clear communication pathways, clear description of re- 12. Wiertsema SH, van Dongen JM, Geleijn E, Beckerman H, Bloemers FW, sponsibilities, reflection meetings and appropriate Ostelo RWJG, et al. The Transmural Trauma Care Model (TTCM) for the reha- reimbursement. bilitation of trauma patients is effective in improving patient related outcome measures: a non-randomized controlled trial. BMC Health Serv Res. 2019; 19:819. 13. Saunders RP, Evans MH, Joshi P. Developing a process-evaluation plan for assessing health promotion program implementation: a how-to guide. Health Promot Pract. 2005;6:134–147. 14. Craig P, Dieppe P, Macintyre S, Michie S, Nazareth I, Petticrew M. Developing and evaluating complex interventions: the new Medical Research Council guidance. Int J Nurs Stud. 2013;50:587–592. 15. Hulscher ME, Laurant MG, Grol RP. Process evaluation on quality improvement interventions. Qual Saf Health Care. 2003;12:40–46. 16. Oakley A, Strange V, Bonell C, Allen E, Stephenson J, Team RS. Process evaluation in randomised controlled trials of complex interventions. BMJ. 2006;332:413–416. 17. Grol R, Grimshaw J. From best evidence to best practice: effective implementation of change in patients’ care. Lancet. 2003;362:1225–1230.

Research 307 18. Grol R, Baker R, Moss F. Quality improvement research: understanding the science 30. Gale NK, Heath G, Cameron E, Rashid S, Redwood S. Using the framework method of change in health care. Qual Saf Health Care. 2002;11:110–111. for the analysis of qualitative data in multi-disciplinary health research. BMC Med Res Methodol. 2013;13:117. 19. Luc M, Corriveau H, Boire G, Filiatrault J, Beaulieu MC, Gaboury I. Patient-related factors associated with adherence to recommendations made by a fracture liaison 31. Thomas SA, Drummond AE, Lincoln NB, Palmer RL, das Nair R, Latimer NR, et al. service: A mixed-method prospective study. Int J Environ Res Public Health. Behavioural activation therapy for post-stroke depression: the BEADS feasibility 2018;15:944. RCT. Health Technol Assess. 2019;23:1–176. 20. Kruithof N, Traa MJ, Karabatzakis M, Polinder S, de Vries J, de Jongh MAC. Perceived 32. Essink DR. Sustainable Health Systems: the Role of Change Agents in Health System changes in quality of life in trauma patients: a focus group study. J Trauma Nurs. Innovation. Amsterdam: VU University; 2012. 2018;25:177–186. 33. Rotmans J, Loorbach DA. Complexity and transition management. J Ind Ecol. 21. van Raak R. The transition (management) perspective on long-term changes 2009;13:184–196. in healthcare. In: Broerse JEW, Bunders JFG, eds. Transitions in health systems: dealing with persistent problems. Amsterdam: VU University Press; 2010: 34. Achterbergh R, Lakeman P, Stemerding D, Moors EH, Cornel MC. Implementation of 49–86. preconceptional carrier screening for cystic fibrosis and haemoglobinopathies: a sociotechnical analysis. Health Policy. 2007;83:277–286. 22. Higgins JTP, Green S. Cochrane handbook for systematic reviews of interventions. Version 5.1.0. [Updated March 2011]. In: The Cochrane Collaboration; 2011. http:// 35. van den Bosch S. Transition Experiments: Exploring Societal Changes Towards Sus- handbook.cochrane.org/. Accessed 2011. tainability. Rotterdam: Erasmus University; 2010. 23. Linan L, Steckler A. Process Evaluation for Public Helath Intervention and Research; 36. Tong A, Sainsbury P, Craig J. Consolidated criteria for reporting qualitative research an overview. In: Process Evaluation for Public Health Intervention and Research. San (COREQ): a 32-item checklist for interviews and focus groups. Int J Qual Health Francisco, USA: Jossey-Bass Inc; 2004:1–23. Care. 2007;19:349–357. 24. Grol R, Wensing M. What drives change? Barriers to and incentives for achieving 37. Morris RL, Hill KD, Ackerman IN, Ayton D, Arendts G, Brand C, et al. A mixed evidence-based practice. Med J Aust. 2004;180(6 Suppl):S57–S60. methods process evaluation of a person-centred falls prevention program. BMC Health Serv Res. 2019;19:906. 25. Merton RK. The focussed interview and focus groups: continuities and disconti- nuities. Public Opin Quart. 1987;51:550–566. 38. Richards DA, Bazeley P, Borglin G, Craig P, Emsley R, Frost J, et al. Integrating quantitative and qualitative data and findings when undertaking randomised 26. Stewart DW, Shamdasani PM. Focus Groups: Theory and Practice. Newbury Park: controlled trials. BMJ Open. 2019;9:e032081. Sage Publishers; 1990. 39. Fern EF. Advanced Focus Group Research. London: Sage Publishers; 27. Green J, Thorogood N. Qualitative Methods for Health Research. London: Sage 2001. Publishers; 2018. 40. Rigter T, Henneman L, Broerse JE, Shepherd M, Blanco I, Kristoffersson U, et al. 28. Morgan DL. Planning Focus Groups. Vol 2. Thousand Oaks, CA: Sage Publishers; 1997. Developing a framework for implementation of genetic services: learning from 29. Ritchie J, Lewis J. Qualitative Research Practice: A Guide for Social Science Students examples of testing for monogenic forms of common diseases. J Community Genet. 2014;5:337–347. and Researchers. London: Sage; 2003.

Journal of Physiotherapy 67 (2021) 284–290 j o u r n a l h o m e p a g e : w w w. e l s ev i e r. c o m / l o c a t e / j p hy s Research Twelve weeks of water-based circuit training exercise improves fitness, body fat and leg strength in people with stable coronary heart disease: a randomised trial Anna Scheer a, Amit Shah b, Beatriz Ito Ramos de Oliveira a, Ignacio Moreno-Suarez a,c, Angela Jacques a, Daniel Green d, Andrew Maiorana a,b,e a School of Allied Health, Curtin University, Perth, Australia; b Advanced Heart Failure and Cardiac Transplant Service, Fiona Stanley Hospital, Perth, Australia; c The Hong Kong Polytechnic University, Department of Rehabilitation Sciences, Hong Kong; d School of Human Sciences, The University of Western Australia, Perth, Australia; e Allied Health Department, Fiona Stanley Hospital, Perth, Australia KEY WORDS ABSTRACT Hydrotherapy Question: In people with stable coronary heart disease, what are the effects of water-based circuit training Coronary artery disease exercise on aerobic capacity, strength and body composition? How do these effects compare with those of Circuit-based exercise gym-based exercise? Design: Parallel group, randomised controlled trial with concealed allocation and Muscle strength intention-to-treat analysis. Participants: Fifty-two participants with stable coronary heart disease. Cardiorespiratory fitness Interventions: Twelve weeks of: three 1-hour sessions per week of moderate-intensity water-based circuit training exercise with alternating aerobic and resistance stations (WEX); three 1-hour sessions per week of moderate-intensity gym-based circuit training exercise (GEX); or continuing usual activities (control). Outcome measures: Aerobic capacity (VO2peak), upper and lower limb one repetition maximum strength (biceps curl, latissimus dorsi pulldown, hamstring curl and leg press), anthropometry (weight, body mass index and girth) and dual energy x-ray absorptiometry. Results: Forty-five participants completed the study (WEX n = 15, GEX n = 18, control n = 12). Both training groups significantly improved VO2peak compared with control: WEX by 2.5 ml/kg/min (95% CI 0.6 to 4.4) and GEX by 2.3 ml/kg/min (95% CI 0.6 to 4.0). WEX and GEX improved hamstring strength compared with control: WEX by 6.3 kg (95% CI 1.2 to 11.3) and GEX by 7.6 kg (95% CI 2.9 to 12.2). Compared with control, GEX increased leg press strength by 15.5 kg (95% CI 5.7 to 25.3), whereas the effect of WEX was less clear (MD 7.1 kg, 95% CI –3.5 to 17.7). Only GEX improved latissimus dorsi pulldown strength. Compared with control, total body fat was reduced with WEX (–1.1 kg, 95% CI –2.3 to 0.0) and GEX (–1.2 kg, 95% CI –2.3 to –0.1). There were negligible between-group differences in weight or waist circumference. Conclusion: WEX was well tolerated and improved aerobic capacity, leg strength and body fat to a similar degree as GEX in people with coronary heart disease. These findings suggest that WEX is an effective exercise training alternative to GEX for people with coronary heart disease. Trial registration: ANZCTR12616000102471. [Scheer A, Shah A, Ito Ramos de Oliveira B, Moreno-Suarez I, Jacques A, Green D, Maiorana A (2021) Twelve weeks of water-based circuit training exercise improves fitness, body fat and leg strength in people with stable coronary heart disease: a randomised trial. Journal of Phys- iotherapy 67:284–290] © 2021 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 and approximately two-thirds of people with CHD are overweight or obese.10 The high prevalence of these comorbidities suggests that Exercise participation and higher levels of aerobic fitness have been associated with reduced mortality and morbidity in people with low-impact exercise strategies may be beneficial for many people coronary heart disease (CHD).1,2 As such, exercise training has with CHD.11 Water-based exercise presents one such option due to become an important component of CHD management.3,4 Guidelines recommend accumulating 150 to 300 minutes per week of moderate- the effects of buoyancy on reducing the weight bearing load on the intensity physical activity, 75 to 150 minutes of vigorous-intensity lower limbs and spine.12 physical activity, or an equivalent combination, along with muscle strengthening exercises on  2 days per week.5 Despite the estab- It was recently reported that an aquatic exercise training circuit lished benefits of exercise in people with CHD, many do not under- take sufficient physical activity to meet guidelines.6,7 with integrated aerobic and resistance exercises improved strength and aerobic fitness in people with type 2 diabetes.13 However, there Physical activity is particularly low in people with CHD who have are currently very few randomised controlled studies with a duration comorbid conditions such as arthritis or obesity.8,9 This is concerning, given that arthritis is experienced by over half of the CHD population9 of . 3 weeks comparing water-based exercise with gym-based ex- ercise in people with CHD. The only study with a duration . 3 weeks comparing water-based and gym-based exercise found promising improvements in body weight, skinfolds, exercise test time and maximum strength in people with CHD following water-based https://doi.org/10.1016/j.jphys.2021.08.012 1836-9553/© 2021 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 285 exercise that were similar to those observed with gym exercise.14 conditioning phase involved a circuit of alternating aerobic and However, this study was limited by the lack of gold-standard resistance exercise stations (45 seconds work, 15 seconds active re- outcome measures such as peak oxygen uptake and imaging- covery). Sessions progressed over the first 3 weeks from one to three derived body composition measures. circuits per session. Aerobic exercise intensity commenced at 50 to 65% of measured heart rate maximum in weeks 1 and 2, and This study aimed to investigate the effect of 12 weeks of water- increased to 60 to 65% in weeks 3 and 4, 60 to 70% in weeks 5 and 6, based circuit training exercise (WEX) training compared with 12 70 to 80% in weeks 7 and 8, and 80% in weeks 9 to 12. Heart rate was weeks of gym-based circuit training exercise (GEX) training at a monitored using wrist-worn monitors with chest strapsa. Rating of similar intensity, to determine if water-based exercise is an effective perceived exertion (RPE) was also used to guide exercise prescription alternative to the widely prescribed approach of gym-based exercise and was progressed from 11 to 14 over the course of training.15 for people with CHD. Resistance exercises were matched for muscle group between WEX and GEX, and the range of motion of the arm exercises for both Therefore, the research questions for this randomised controlled groups was limited to the range allowed by the water level of the trial were: WEX group. Resistance exercise RPE targets were 12 to 15. 1. In people with stable coronary heart disease, what are the effects The WEX group trained in a graded-depth hydrotherapy pool of water-based circuit training exercise on aerobic capacity, (34.5 C) and were submersed to the xiphoid process. Aerobic exer- strength and body composition? cises included walking/jogging and high knee lifts. Strength exercises included unilateral knee flexion/extension, unilateral hip flexion/ 2. How do the effects compare with those of gym-based exercise? extension, unilateral hip abduction/adduction (with the side of this exercise alternated at each circuit), bilateral elbow flexion/extension, Method and bilateral shoulder abduction/adduction (limited by water depth). More details are presented in Appendix 1 on the eAddenda. The Design resistance was provided by custom-designed paddles, using acrylic sheets either side of the limb, fastened by hook and loop fasteners or This was a three-arm, parallel group, randomised controlled trial buckles, ranging from 15 x 20 cm to 30 x 35 cm (see Appendix 1 on with concealed allocation and intention-to-treat analysis. Participants the eAddenda). Resistance was progressed by increasing the speed of underwent baseline assessments prior to group allocation. The movement guided by waterproof metronomes worn by participants blocked random allocation list was weighted towards the two on headbandsb whilst maintaining the same range of movement. experimental groups and concealed in opaque and sealed envelopes. Exercises were progressed to include larger paddle sizes once par- Participants were allocated to continue their allocated intervention ticipants were able to maintain the set pace per repetition with good (WEX or GEX three times per week, or continuing usual activities for technique and Borg RPE fell below 13. the control group) for 12 weeks, before reassessment within 2 weeks of the final training session. The nature of the exercises prevented The GEX group exercised in a cardiac rehabilitation outpatient blinding of participants. Recruitment occurred from December 2016 gymnasium. Aerobic exercises involved stationary cycling and to March 2019. Data collection was completed in July 2019. treadmill walking. Strength exercises involved bilateral knee exten- sion, bilateral latissimus pulldown, bilateral knee flexion, unilateral Participants, therapists, centre hip flexion/extension, bilateral tricepsc, unilateral hip abduction/ adduction with ankle weightsd, bilateral biceps and bilateral shoulder Participants were recruited through hospital databases and com- abduction with dumbbells (limited to the height of the xiphoid pro- munity advertising (Figure 1). The inclusion criteria were: a diagnosis cess). Resistance exercises were initially prescribed at approximately of CHD (based on a previous myocardial infarction, percutaneous 50% one repetition maximum (1RM). Participants commenced with coronary intervention, diagnostic imaging showing  50% occlusion 10 repetitions and progressed to the next weight once they could of  1 coronary artery, or coronary artery bypass graft surgery); stable achieve 12 to 15 repetitions in 45 seconds, with good technique and medication for  1 month; and  6 months after any coronary event at an RPE , 13. Due to the concentric-only nature of the WEX or surgery. Participants were excluded if they: had an ejection frac- training, biceps and triceps and shoulder abduction and latissimus tion , 45%: had severe musculoskeletal, respiratory or neurological pulldown were alternated each circuit in the GEX program, so that impairment that would limit exercise training; had received current after two consecutive sessions the time training each muscle group or recent (within 6 months) chemotherapy or radiotherapy for cancer was matched between the WEX and GEX groups. treatment; had displayed an adverse response on baseline exercise testing (ie, ischaemic signs or symptoms at a workload of , 4 METs, Control group new-onset left bundle branch block or ventricular tachyarrhythmia); had baseline blood test results suggesting significant other health Participants in the control group were instructed to maintain their issues; had participated in a supervised exercise program in the past usual activities throughout the study and were offered an optional 3 months; were current smokers; or were using insulin. gym-based exercise program or home exercise advice at the completion of the study. Assessments and exercise training sessions were conducted in a tertiary hospital outpatient setting in Perth, Australia, by a physio- Outcome measures therapist or an accredited exercise physiologist. Therapists were not blinded for the assessments. All water-based exercise sessions were Primary outcome supervised by at least two people (at least one physiotherapist or The primary outcome was aerobic exercise capacity (VO2peak). exercise physiologist). Secondary outcomes Intervention Aerobic exercise capacity (VO2peak) was assessed by indirect cal- Regardless of which intervention group they were randomised to, orimetrye using a modified chronotropic protocol on a treadmillf with participants were asked to maintain their usual diet throughout the increases in speed and incline every 3 minutes. Participants were study. Any medication changes during the study were documented. monitored with a 12-lead ECGg. The metabolic cart was calibrated according to standard procedures prior to each assessment. The test Exercise training programs continued until volitional exhaustion, unless terminated by the su- The WEX and GEX participants trained 3 days per week for pervising doctor. approximately 1 hour per session. All sessions commenced and Muscular strength was assessed with 1RM assessments for bicep concluded with 5 minutes of light aerobic activity and stretching. The curl using a dumbbellh and for latissimus pulldown, leg press and

286 Scheer et al: Water-based exercise for coronary heart disease Preliminary medical record screen Response to advertisements (n = 4,700) (n = 298) Not meeting criteria or significant Not meeting criteria or declined to geographical distance (n = 4,373) participate (n = 235) Invitation letter sent (n = 327) Declined to participate (n = 276) Phone screened for eligibility (n = 114) Excluded (n = 62) • ineligible (n = 44) • declined (n = 10) • medical exclusion at baseline testing (n = 7) • withdrew after VO2peak assessment (n = 1) Measured aerobic capacity, muscular strength, anthropometry and body composition Week 0 Randomised (n = 52) (n = 20) (n = 20) (n = 12) WEX group GEX group Con group • water-based exercise • gym-based exercise • continue usual exercise • 60-min sessions • 60-min sessions • 12 weeks • 3/week • 3/week • 12 weeks • 12 weeks Lost to follow-up (n = 0) Lost to follow-up (n = 5) Lost to follow-up (n = 2) • unrelated illness (n = 2) • family reasons (n = 1) • work reasons (n = 2) • arrhythmia (n = 1) • family reasons (n = 1) Measured aerobic capacity, muscular strength, anthropometry and body composition Week 12 (n = 15) (n = 18) (n = 12) Figure 1. Design and flow of participants through the study. hamstring curl using seated weight stack machinesi and the machine x-ray absorptiometry (DXA) was used to measure body mass, body fat settings at baseline were used for follow-up assessments for repro- and bone mineral density. The DXA scan was performed on a body ducibility. Participants were instructed to avoid a Valsalva manoeuvre scannerk after  6-hour fast.16 Standard procedure of omitting the throughout strength testing. Participants initially conducted six rep- right arm and duplicating the left arm was used if a participant was etitions of a light weight for a muscle-specific warm-up, followed by too large to fit on the DXA bed. 1-minute rest, then two repetitions of a moderate weight followed by a 1-minute rest. Subsequently, progressively heavier weights were Data analysis lifted once with 2 minutes rest between repetitions, with the last successful repetition that could be lifted through a full range of Sample size was based on data from Tokmakidis et al,17 who found motion recorded as the 1RM. that water-based exercise training over 4 months increased VO2peak in people with CHD from 26.2 (SD 4.0) to 28.4 (SD 4.9) ml/kg/min.17 The Anthropometric measures included weight, body mass index calculated Cohen’s d was 0.48,18 with an effect size of 0.24.19 (BMI), waist circumference and hip circumference. Body weight was Assuming 90% power and a 5% level of significance, and a conserva- measured by digital scalesj. Waist and hip girth measurements were tive estimated correlation among repeated measures of 0.75 (based taken in triplicate, with the median used for analysis. Dual energy

Research 287 Table 1 Baseline characteristics of participants. Characteristic Completed (n = 45) Lost to follow-up (n = 7) WEX (n = 15) GEX (n = 18) Con (n = 12) WEX (n = 5) GEX (n = 2) Age (yr), mean (SD) 66 (8) 67 (8) 71 (5) 66 (5) 66 (7) Gender, n males (%) 12 (80) 16 (89) 9 (75) 3 (60) 2 (100) Time since diagnosis (yr) 3.3 2.3 5.5 4.0 11.2 median 0.8 to 15 0.5 to 29 0.5 to 18 1.3 to 9 2.4 to 20 range Medical history, n (%) 1 (7) 5 (28) 2 (17) 1 (20) 0 (0) cancer (other than skin) 1 (7) 1 (6) 3 (25) 0 (0) 0 (0) type 2 diabetes 1 (7) 3 (17) 0 (0) 0 (0) 1 (50) pre-diabetes 1 (7) 0 (0) 2 (17) 0 (0) 0 (0) stroke 0 (0) 1 (6) 1 (8) 0 (0) 0 (0) transient ischaemic attack 1 (7) 0 (0) 0 (0) 0 (0) 0 (0) peripheral vascular disease 5 (33) 4 (22) 3 (25) 2 (40) 0 (0) arthritis 6 (40) 4 (22) 6 (50) 3 (60) 2 (100) non-specified joint pain 2 (13) 4 (22) 3 (25) 2 (40) 0 (0) respiratory condition Cardiac history, n (%) 3 (20) 2 (11) 2 (17) 0 (0) 0 (0) silent 4 (27) 7 (39) 5 (42) 1 (20) 0 (0) angina only 8 (53) 9 (50) 5 (42) 4 (80) 2 (100) myocardial infarction Treatment, n (%) 2 (13) 7 (39) 6 (50) 0 (0) 1 (50) CABG 11 (73) 14 (78) 7 (58) 5 (100) 1 (50) PCI 3 (17) 2 (17) 0 (0) CABG and PCI 0 (0) 2 (17) 0 (0) 0 (0) medication only 2 (13) 0 (0) 0 (0) Medication, n (%) 9 (75) 1 (50) beta blockers 5 (33) 7(39) 11 (92) 1 (20) 2 (100) lipid lowering 14 (93) 17 (94) 3 (25) 4 (80) glucose lowering 12 (100) 0 (0) 0 (0) antiplatelet/anticoagulant 0 (0) 1 (6) 4 (33) 5 (100) 2 (100) ACE inhibitors 15 (100) 17 (94) 5 (42) 1 (20) 2 (100) angiotensin II receptor blockers 10 (56) 3 (25) 2 (40) calcium channel blockers 5 (33) 4 (22) 3 (60) 0 (0) diuretic 4 (27) 3 (17) 1 (8) 1 (20) 0 (0) proton pump inhibitors 1 (7) 5 (42) 2 (40) 0 (0) 0 (0) 1 (6) 1 (50) 5 (33) 4 (22) ACE = angiotensin converting enzyme, CABG = coronary artery bypass graft surgery, Con = control, GEX = gym-based exercise, PCI = percutaneous coronary intervention, Silent = picked up on investigation without prior cardiac symptoms, WEX = water-based exercise. on our previous study of water-based exercise for people with type 2 one control participant due to illness. DXA data were incomplete in diabetes),13 the estimated total sample size was 33 (ie, 11 per group one WEX participant due to artefact. for a repeated measures ANOVA (mixed model) with three groups over two timepoints)l. To allow for training program withdrawals, Research question 1 recruitment targets were set at 20 per group for the training groups. Water-based exercise improved the primary outcome – aerobic Results were analysed with commercial softwarem. Generalised capacity – by a mean of 2.5 ml/kg/min (95% CI 0.6 to 4.4) (Table 2). linear mixed models with appropriate links were used for within- Water-based exercise also improved a measure of leg strength (mean group pre-post assessments and group-time interactions. Mixed ef- difference in 1RM hamstring curl of 6.3 kg, 95% CI 1.2 to 11.3) fects Tobit models were used for leg press and hamstring data due to (Table 2). Water-based exercise also reduced total body fat by a mean ceiling effects in those outcomes. of 1.1 kg (95% CI 0.0 to 2.3) (Table 3). Water-based exercise induced no clear changes in upper body strength or anthropometric measures Results (Tables 2 and 3). Flow of participants and therapists through the study Research question 2 Participant flow through the study is presented in Figure 1. Both modes of exercise training improved exercise capacity Participant characteristics are described in Table 1. There was one (VO2peak) to a similar extent and GEX increased exercise time (Table 2). adverse event during the study: a GEX participant with a history of Both WEX and GEX increased leg strength, but only GEX significantly supraventricular tachycardia experienced an episode of it during improved latissimus pulldown strength (Table 2). Body fat decreased training in week 8 of his program. He received immediate medical with both types of exercise training over time, with a negligible dif- attention and recovered fully that day, although he withdrew from ference between the effects of the two types of training (Table 3). the study to undergo an elective ablation procedure. The results from all participants who completed baseline testing Aerobic capacity assessments were incomplete for three WEX are available in Appendix 2 (aerobic capacity and strength testing) participants (one due to illness, one injured outside the study and one and Appendix 3 (body composition) on the eAddenda. Individual unable to tolerate a mask or mouthpiece). Muscular strength as- participant data are presented in Table 4 on the eAddenda. sessments were incomplete for one WEX and one control participant due to illness. Due to musculoskeletal limitations, paired data were Discussion unavailable for leg press in three WEX, two GEX and two control participants; hamstring curl in two WEX and one control participant; The estimates of treatment effects generated by this study show biceps curl in one WEX, two GEX and one control participant; and that water-based circuit training resulted in similar benefits in latissimus pulldown in four WEX, three GEX and three control par- aerobic capacity, leg strength and total body fat as traditionally pre- ticipants. DXA scans were not available for one WEX, one GEX and scribed gym-based circuit training. These findings highlight water- one control participant due to machine servicing/unavailability, and

Table 2 Mean (SD) of groups, mean (SD) difference within groups, and mean (95% CI) difference between groups for paired aerobic Outcome Groups Week 0 Week 12 WEX GEX Con WEX GEX Co VO2peak (ml/kg/min) 29.1 (8.9) 26.3 (5.5) 24.6 (6.7) 30.9 (9.1) 27.9 (5.8) 23.9 VO2peak (ml/min) 2,398 (616) 2,188 (383) 1,970 (594) 2,502 (571) 2,302 (384) 1,915 Percent of predicted VO2 maximum 104.2 (23.5) 97.4 (21.5) 97.8 (20.4) 110.6 (24.9) 103.3 (21.7) 95.0 ( Exercise duration (s) VCO2peak (ml/min) 942 (184) 843 (116) 831 (233) 998 (220) 946 (120) 867 ( VE/VCO2 slope 2,679 (703) 2,392 (526) 2,117 (683) 2,836 (654) 2,620 (581) 2,090 RER RPE 28.9 (3.9) 27.6 (4.1) 29.6 (3.5) 29.0 (3.4) 27.8 (4.0) 28.7 Bicep curl 1RM (kg) 1.12 (0.07) 1.09 (0.08) 1.07 (0.08) 1.14 (0.08) 1.13 (0.10) 1.10 ( Latissimus pulldown 1RM (kg) Hamstring curl 1RM (kg) 17 (2) 18 (2) 17 (1) 17 (2) 18 (2) 17 Leg press 1RM (kg) 11.4 (4.4) 10.1 (3.5) 9.0 (2.7) 11.2 (4.2) 11.1 (3.6) 9.2 ( 45.0 (12.7) 41.3 (10.9) 38.8 (9.5) 47.0 (10.9) 47.0 (12.1) 40.0 49.9 (12.5) 50.6 (12.4) 44.6 (9.8) 55.1 (16.0) 57.3 (12.1) 43.4 124.0 (37.0) 121.5 (33.6) 122.5 (44.8) 132.1 (39.9) 138.2 (31.4) 123.5 Con = control group, GEX = gym-based exercise training group, RER = respiratory exchange ratio, RPE = rating of perceived e 1RM = one repetition maximum strength. a Difference between estimated margins of means using general linear mixed model analysis. b Metobit analysis used. Table 3 Mean (SD) of groups, mean (SD) difference within groups, and mean (95% CI) difference between groups for paired anthrop Outcome Groups Week 0 Week 12 WEX GEX Con WEX GEX Con Anthropometry Weight (kg) 85.2 (13.9) 84.7 (12.9) 80.3 (10.4) 84.4 (14.1) 83.8 (12.1) 80.1 (10.7) BMI (kg/cm2) 29.1 (4.2) 28.4 (3.9) 28.0 (3.7) 28.8 (4.3) 28.1 (3.7) 27.9 (3.5) Waist girth (cm) 101.5 (8.6) 100.2 (11.8) 101.2 (9.9) 101.6 (9.9) Hip girth (cm) 100.6 (12.5) 102.8 (10.7) 107.1 (8.7) 107.9 (7.4) 105.6 (6.2) 104.5 (6.9) 0.94 (0.07) 104.4 (5.8) 103.9 (7.5) Waist:hip ratio 0.93 (0.09) 0.97 (0.07) 0.97 (0.06) 0.97 (0.06) 0.98 (0.08) DXA data Total mass (kg) 85.3 (14.7) 83.8 (13) 79.7 (11.4) 83.9 (14.9) 83.2 (12.4) 80.1 (11.9) – Total fat (kg) 29.5 (12.0) 28.6 (7.0) 27.1 (8.0) 28.5 (12.1) 27.6 (6.7) 27.3 (8.2) Tissue fat (%) 35.0 (11.4) 35.2 (6.1) 35.0 (7.7) 34.4 (12.1) 34.3 (6.5) 35.0 (7.5) Total lean (kg) 52.3 (8.6) 49.7 (8.4) 52.4 (9.9) 52.6 (9.0) 50.0 (8.2) BMD (g/cm2) 52.9 (9.7) 1.28 (0.13) 1.27 (0.15) 1.28 (0.12) 1.26 (0.18) 1.28 (0.15) 1.27 (0.19) BMD = bone mineral density, BMI = body mass index, Con = control group, DXA = dual energy x-ray absorptiometry, GEX = a Difference between estimated margins of means using general linear mixed model analysis.

capacity and muscular strength data. 288 Scheer et al: Water-based exercise for coronary heart disease Difference within groups Difference between groupsa Week 12 minus Week 0 Week 12 minus Week 0 on WEX GEX Con WEX minus Con GEX minus Con WEX minus GEX (5.9) 1.8 (2.6) 1.6 (1.9) –0.7 (2.6) 2.5 (0.6 to 4.4) 2.3 (0.6 to 4.0) 0.2 (–1.5 to 1.9) (576) 104 (242) 114 (128) –55 (239) 166 (16 to 317) 167 (30 to 305) –1 (–138 to 136) (15.4) –2.8 (7.4) 9.5 (2.5 to 16.5) 8.7 (2.3 to 15.1) 0.8 (–5.5 to 7.2) (174) 6.4 (9.9) 5.9 (7.2) 22 (–35 to 79) (624) 56 (83) 103 (39) 36 (84) 195 (9 to 381) 68 (16 to 121) –47 (–98 to 5) (3.4) 157 (277) 228 (185) –27 (267) 1.1 (–0.8 to 3.0) 253 (83 to 424) –58 (–228 to 112) (0.06) 0.1 (1.6) 0.2 (1.4) –0.9 (1.5) –0.01 (–0.05 to 0.04) 1.1 (–0.7 to 2.8) –0.0 (–1.7 to 1.7) (2) 0.02 (0.03) 0.04 (0.03) 0.03 (0.03) 0.01 (–0.03 to 0.05) –0.02 (–0.06 to 0.02) (3.0) –1 (–2 to 0) (8.5) 0 (1) 0 (1) 0 (1) –0.4 (–1.4 to 0.7) 0 (–1 to 1) –1 (–2 to 0) (11.1) –0.2 (1.7) 1 (1.3) 0.2 (1.3) 0.6 (–3.7 to 4.8) 0.7 (–0.3 to 1.7) –1.0 (–2.0 to –0.1) (49.2) 5.7 (4.2) 1.2 (4.5) 6.3b (1.2 to 11.3) –3.6 (–7.3 to 0.0) 2 (5.3) 6.7 (4.1) –1.2 (11.1) 7.1 b (–3.5 to 17.7) 4.2 (0.3 to 8.1) –1.3 b (–5.8 to 3.2) 5.2 (5.9) 16.7 (11.5) 1 (22.2) 7.6 b (2.9 to 12.2) –8.4 b (–17.6 to 0.8) 8.1 (16.4) 15.5 b (5.7 to 25.3) exertion, VCO2peak = peak carbon dioxide output, VO2peak = peak oxygen uptake, WEX = water-based exercise training group, pometry and body composition data. Difference within groups Difference between groupsa Week 12 minus Week 0 Week 12 minus Week 0 GEX minus Con WEX GEX Con WEX minus Con WEX minus GEX –0.8 (5.1) –0.9 (4.2) –0.2 (4.3) –0.6 (–2.0 to 0.8) –0.7 (–2.1 to 0.6) 0.1 (–1.1 to 1.4) –0.3 (1.6) –0.3 (1.3) –0.1 (1.5) –0.2 (–0.6 to 0.3) –0.2 (–0.7 to 0.3) 0.0 (–0.4 to 0.5) –0.4 (4.4) –1.6 (3.4) 0.1 (3.9) –0.4 (–2.8 to 1.9) –1.7 (–3.9 to 0.6) 1.2 (–0.8 to 3.3) –0.8 (2.9) –0.6 (3.1) –0.1 (–2.1 to 1.9) –0.6 (–2.5 to 1.4) 0.4 (–1.3 to 2.2) 0.01 (0.03) –1.2 (2) 0.01 (0.03) –0.00 (–0.03 to 0.02) –0.01 (–0.04 to 0.01) 0.01 (–0.02 to 0.03) 0.00 (0.02) –1.4 (5.8) –0.6 (4.4) 0.4 (5.2) –1.8 (–3.3 to –0.4) –1.2 (–2.5 to 0.2) –0.7 (–1.9 to 0.6) –0.9 (4.7) –1.0 (2.4) 0.2 (3.6) –1.1 (–2.3 to 0.0) –1.2 (–2.3 to –0.1) 0.1 (–0.9 to 1.1) –0.6 (4.6) –0.9 (2.2) 0.0 (3.4) –0.6 (–1.6 to 0.5) –0.8 (–1.8 to 0.2) 0.2 (–0.7 to 1.2) –0.4 (3.8) 0.3 (3.0) 0.3 (3.7) –0.7 (–1.5 to 0.0) –0.01 (0.06) 0.00 (0.05) –0.01 (0.09) 0.0 (–0.7 to 0.7) –0.7 (–1.4 to –0.1) 0.01 (–0.01 to 0.02) 0.01 (–0.01 to 0.02) 0.00 (–0.01 to 0.02) = gym-based exercise training group, WEX = water-based exercise training group.

Research 289 based exercise as an effective alternative to gym-based training for response is universal in response to WEX. Encouragingly, improve- ments in leg strength alone have been associated with reduced people with CHD. all-cause and cardiovascular mortality in people with CHD,30 sug- gesting that the strength changes seen in the current study are of It is believed that this study is the first parallel-group controlled clinical value, despite the lack of upper limb improvement. trial to investigate the effect of water-based versus gym-based circuit With improved treatment for acute coronary events, increasing training on VO2peak in people with stable CHD. It found that WEX numbers of people are living with chronic CHD. To reduce recurrent resulted in a similar magnitude improvement in VO2peak to GEX, coronary events in this group, it is important that the exercise pre- increasing by 1.8 ml/kg/min and 1.6 ml/kg/min, respectively. This is scription paradigm shifts from merely a focus on time-limited reha- bilitation to long-term secondary prevention. The findings from this similar to DeSchutter et al’s finding of a 1.9 ml/kg/min mean study suggest that water-based exercise should be encouraged as one improvement in a large cohort trial in centre-based exercise reha- of the suite of exercise options to help people with chronic CHD be bilitation.2 Whilst not a randomised controlled trial, Tokmakidis et al sufficiently active to achieve health and fitness benefits. Importantly, WEX was well tolerated in the cohort of participants with stable CHD, examined two 4-month blocks of water-based exercise training, with no adverse events occurring in this group. The three sessions of separated by 4 months of detraining in people with CHD.17 They re- WEX per week prescribed in the study is consistent with exercise training guidelines.5 In the context of community exercise partici- ported comparable increases in VO2peak of 2.2 ml/kg/min and 1.8 ml/ pation, people may wish to undertake multiple WEX sessions or kg/min.17 There is increasing recognition that aerobic capacity is an combine WEX with other modes of exercise, depending on their preference and capabilities. For example, WEX could be used as an important prognostic indicator across a range of chronic condi- initial entry into exercise programs for deconditioned patients or tions.2,20 In people with CHD, a 1 ml/kg/min higher VO2peak at base- those with musculoskeletal comorbidities. Alternatively, WEX could line or improvement over time has been associated with a 10 to 16% be used as an adjunct with GEX and/or walking programs to increase improvement in cardiac and all-cause mortality,2,20,21 highlighting exercise variety. Clinicians wishing to prescribe similar aquatic exer- cise programs to the WEX program investigated in this study could the importance of even modest improvements in VO2peak in this easily and cheaply replicate the equipment used in this study (acrylic population. The mean change in VO2peak for both WEX and GEX plastic sheets, with hook and loop or buckle fastenings, as pictured in exceeded 1 ml/kg/min more than the mean change in the control Appendix 1) or use commercially available aquatic resistance equip- ment. Importantly, once individual tolerance to the program is group, and there was no important difference in the change in established and correct techniques have been taught, many patients would be able to continue the program independently. VO2peak for the two training types, suggesting that these forms of exercise training are equally effective. This study excluded people with CHD if they also had left ven- tricular dysfunction, type 1 diabetes, treatment with insulin, or Despite similar changes in VO2peak with both types of exercise serious respiratory, neurological or musculoskeletal pathology, so the training, only GEX significantly increased exercise test time compared generalisation of the safety and efficacy for these sub-populations with the control group. A similar effect was observed in a study of 24 remains unknown. Additionally, participants were recruited  6 weeks of water-based and land-based walking in older adults,22 which months after any myocardial infarction, coronary artery bypass graft surgery or percutaneous coronary intervention, so the results may found a significant increase in exercise test time with land-based not be applicable to the initial stages of cardiac rehabilitation. It is walking only, despite similar VO2peak changes. This may have been recommended that future research should examine the effects of due to the specificity of treadmill training and land-based walking to water-based exercise in sub-acute outpatient cardiac rehabilitation. It the outcome of treadmill test duration,23 in contrast to a more gen- should be noted that the resistance machines used for strength assessment of latissimus pulldowns, biceps curl and hamstring curl eralisable effect on aerobic capacity, as measured by change in VO2peak. were the same as those employed during training. This specificity Both training groups experienced a similar mean reduction in fat may have influenced the findings for the GEX group. mass. These reductions in fat mass occurred in the absence of dietary This is the first outpatient-based, parallel-group, randomised controlled trial of combined aerobic and resistance exercise modifications. Previous studies have examined other markers of body comparing water-based and gym-based exercise to examine the fat, such as sum of skinfolds, and found favourable effects of aquatic effect on VO2peak and DXA-derived body composition. The study exercise on reducing body fat,14,17 which appear similar to gym ex- found that water-based circuit training was well tolerated and ercise,14 whilst the effect on body fat measured with bioelectric effective for improving aerobic capacity, leg strength and fat mass, impedance did not reach significance in another study.24 However, it similar to gym-based exercise in people with stable CHD. This is believed that this is the first study comparing DXA data in response supports the expansion of exercise prescription options for people to water-based exercise training compared with gym-based training with stable CHD to include water-based exercise, which may be useful for adding a low joint-impact exercise option to facilitate in people with CHD. There were no significant pre-post differences exercise engagement. for lean tissue data in any group; however, there was a difference in What was already known on this topic: With improved the change in lean mass observed between GEX and WEX. The in- treatment for acute coronary events, increasing numbers of people are living with coronary heart disease. To reduce recurrent dividual group differences were , 450 g, which fall within the margin coronary events in this group, it is important that the exercise of accuracy of 0.61 to 0.86 kg for lean tissue changes for the scannerk prescription includes a focus on long-term secondary prevention. proposed by several studies,25–27 suggesting that clarification with a What this study adds: Water-based circuit training that larger sample is required to determine the effects on lean tissue mass. included both aerobic and resistance stations was well tolerated and effective for improving aerobic capacity, leg strength and fat Additionally, research in older adults found that 24 weeks of water- mass in people with stable coronary heart disease. The benefits were similar in magnitude to the benefits of gym-based exercise walking significantly increased lower limb lean tissue mass in this population. compared to a control group,28 suggesting a longer duration program may be required to induce substantial lean tissue changes. The mean improvements in leg strength with GEX equated to a 13% improvement in hamstring curl and a 14% improvement in leg press. The mean improvements with WEX equated to 10% for hamstring curl and 7% for leg press, with some uncertainty in the latter estimate (Table 2). For upper limb strength (latissimus pulldown), there were greater changes with GEX than WEX. This may reflect the limited range of motion for upper limb exercises during WEX due to the water depth during training.29 Buoyancy-resisted exercises may have a greater impact on upper limb strengthening in the pool and would be an interesting concept for future research. Alternatively, it may be necessary to supplement aquatic exercises with some gym-based or free weight exercises to address upper-body strength. Other studies of WEX in CHD have reported overall strength gains of between 12 and 13%14,17 and found similar changes between WEX and GEX.14 However, individual breakdown of muscle groups was not reported in these studies, so it is unknown if the lack of upper limb

290 Scheer et al: Water-based exercise for coronary heart disease Footnotes: aA300, FT4 and FT7 watches, Polar Electronics, 6. Evenson KR, Butler EN, Rosamond WD. Prevalence of physical activity and Guangzhou, China. bTempo Trainer Pro, FINIS, Tracy, USA. cBravo sedentary behavior among adults with cardiovascular disease in the United States. Pulley Machine, Cybex International, Medway, USA. dAnkle cuff J Cardiopulm Rehabil Prev. 2014;34:406–419. weight, Fortress Fitness, Sydney, Australia. eVmax Encore Metabolic Cart, Vyntus, Vyaire Medical, Illinois, USA. fTMX428220, Trackmaster, 7. Guiraud T, Granger R, Grémeaux V, Bousquet M, Richard L, Soukarie L, et al. Newton, USA. gPC ECG 1200, Norav Medical, Wiesbaden, Germany. Accelerometer as a tool to assess sedentarity and adherence to physical activity hAustralian Barbell Company, Australia. iCybex, Cybex International, recommendations after cardiac rehabilitation program. Ann Phys Rehabil Med. Medway, USA. jSeca 676 wireless scales, Seca gmbh & co, Hamburg, 2012;55:312–321. Germany. kGE Lunar iDXA, GE Healthcare, Wisconsin, USA. lG*Power 3, Universität Düsseldorf, Düsseldorf, Germany. mSTATA V16 for 8. Anderson BK, Phillips MM, Gerald L, DiLillo V, Bittner V. Factors associated with the Windows, Microsoft, Redmond, USA. failure of patients to complete cardiac rehabilitation for medical and nonmedical reasons. J Cardiopulm Rehabil. 2003;23:281–289. eAddenda: Table 4 and Appendices 1 to 3 can be found online at DOI: https://doi.org/10.1016/j.jphys.2021.08.012. 9. Centers for Disease Control and Prevention. Arthritis as a potential barrier to physical activity among adults with heart disease–United States, 2005 and 2007. Ethics approval: The Royal Perth Hospital Human Research Ethics Morb Mortal Wkly Rep. 2009;58:165–169. Committee approved this study (ref: RGS0000002071), with recip- rocal approval from Curtin University (ref: HR227/2015) and The 10. Angerås O, Albertsson P, Karason K, Råmunddal T, Matejka G, James S, et al. Evi- University of Western Australia’s (Ref: RA/4/1/8382) Human Research dence for obesity paradox in patients with acute coronary syndromes: a report Ethics Committees. All participants gave written informed consent from the Swedish Coronary Angiography and Angioplasty Registry. Eur Heart J. before data collection began. 2012;34:345–353. Competing interests: Nil. 11. Pavy B, Iliou M-C, Vergès-Patois B, Brion R, Monpère C, Carré F, et al. French Society Source(s) of support: This work was supported by the National of Cardiology guidelines for cardiac rehabilitation in adults. Arch Cardiovasc Dis. Heart Foundation of Australia [grant number 101489] and the Spin- 2012;105:309–328. naker Health Research Foundation. Winthrop Professor Green is an NHMRC Principal Research Fellow [APP1080914]. Anna Scheer was 12. Becker B. Aquatic therapy: Scientific foundations and clinical rehabilitation ap- supported by a Curtin University Research Scholarship and an plications. Am J Phys Med Rehabil. 2009;1:859–872. Australian Government Research Training Program Scholarship. Ignacio Moreno-Suarez was supported by a Curtin University 13. Scheer AS, Naylor LH, Gan SK, Charlesworth J, Benjanuvatra N, Green DJ, et al. The Research Scholarship and the Heart and Lung Transplant Foundation effects of water-based exercise training in people with type 2 diabetes. Med Sci of Western Australia. Sports and Exerc. 2020;52:417–424. Acknowledgements: We would like to thank Mr Eric Watson for his assistance with the pool-based training sessions and Ms Leanne 14. Volaklis K, Spassis A, Tomakadis S. Land versus water exercise in patients with Campbell for her assistance with the VO2peak assessments and coronary artery disease: effects on body composition, blood lipids, and physical training sessions. fitness. Am Heart J. 2007;154:560.e561–560.e566. Provenance: Not invited. Peer reviewed. Correspondence: Professor Andrew Maiorana, Curtin University, 15. Borg G. Perceived exertion as an indicator of somatic stress. Scand J Rehabil Med. School of Allied Health, GPO Box U1987, Perth, Western Australia, 1970;2:92–98. Australia. Email: [email protected] 16. Carver TE, Christou NV, Andersen RE. In vivo precision of the GE iDXA for the References assessment of total body composition and fat distribution in severely obese pa- tients. Obesity. 2013;21:1367–1369. 1. Stewart RAH, Held C, Hadziosmanovic N, Armstrong PW, Cannon CP, Granger CB, et al. Physical activity and mortality in patients with stable coronary heart disease. 17. Tokmakidis SP, Spassis AT, Volaklis KA. Training, detraining and retraining effects J Am Coll Cardiol. 2017;70:1689–1700. after a water-based exercise program in patients with coronary artery disease. Cardiology. 2008;111:257–264. 2. De Schutter A, Kachur S, Lavie CJ, Menezes A, Shum KK, Bangalore S, et al. Cardiac rehabilitation fitness changes and subsequent survival. Eur Heart J Qual Care Clin 18. Lakens D. Calculating and reporting effect sizes to facilitate cumulative science: a Outcomes. 2018;4:173–179. practical primer for t-tests and ANOVAs. Front Psychol. 2013;4:863. 3. Thompson PD, Buchner D, Piña IL, Balady GJ, Williams MA, Marcus BH, et al. Ex- 19. Cohen J. Statistical Power Analysis for the Behavioral Sciences. 2nd ed. Hoboken, USA: ercise and physical activity in the prevention and treatment of atherosclerotic Taylor and Francis; 2013. cardiovascular disease: a statement from the Council on Clinical Cardiology (Subcommittee on Exercise, Rehabilitation, and Prevention) and the Council on 20. Keteyian SJ, Brawner CA, Savage PD, Ehrman JK, Schairer J, Divine G, et al. Peak Nutrition, Physical Activity, and Metabolism (Subcommittee on Physical Activity). aerobic capacity predicts prognosis in patients with coronary heart disease. Am Circulation. 2003;107(24):3109–3116. Heart J. 2008;156:292–300. 4. Woodruffe S, Neubeck L, Clark RA, Gray K, Ferry C, Finan J, et al. Australian Car- 21. Kavanagh T, Mertens DJ, Hamm LF, Beyene J, Kennedy J, Corey P, et al. Peak oxygen diovascular Health and Rehabilitation Association (ACRA) core components of intake and cardiac mortality in women referred for cardiac rehabilitation. J Am Coll cardiovascular disease secondary prevention and cardiac rehabilitation 2014. Heart Cardiol. 2003;42:2139–2143. Lung Circ. 2015;24:430–441. 22. Haynes A, Naylor LH, Carter HH, Spence AL, Robey E, Cox KL, et al. Land-walking vs. 5. Australian Government Department of Health. Australia’s Physical Activity and water-walking interventions in older adults: Effects on aerobic fitness. J Sport Sedentary Behaviour Guidelines and the Australian 24-Hour Movement Guide- Health Sci. 2020;9:274–282. lines; 2019. https://www1.health.gov.au/internet/main/publishing.nsf/Content/ health-pubhlth-strateg-phys-act-guidelines. Accessed 13 January, 2020. 23. Myers J. Applications of cardiopulmonary exercise testing in the management of cardiovascular and pulmonary disease. Int J Sports Med. 2005;26(Suppl 1):S49–S55. 24. Fiogbé E, Ferreira R, Sindorf MAG, Tavares SA, de Souza KP, de Castro Cesar M, et al. Water exercise in coronary artery disease patients, effects on heart rate variability, and body composition: a randomized controlled trial. Physiother Res Int. 2018;23:e1713. 25. Rothney MP, Martin F-P, Xia Y, Beaumont M, Davis C, Ergun D, et al. Precision of GE Lunar iDXA for the measurement of total and regional body composition in non- obese adults. J Clin Densitom. 2012;15:399–404. 26. Hind K, Oldroyd B, Truscott JG. In vivo precision of the GE Lunar iDXA densitometer for the measurement of total body composition and fat distribution in adults. Eur J Clin Nutr. 2011;65:140–142. 27. Dordevic AL, Bonham M, Ghasem-Zadeh A, Evans A, Barber E, Day K, et al. Reli- ability of compartmental body composition measures in weight-stable adults using GE iDXA: implications for research and practice. Nutrients. 2018;10:1484. 28. Naylor LH, Maslen BA, Cox KL, Spence AL, Robey E, Haynes A, et al. Land- versus water-walking interventions in older adults: Effects on body composition. J Sci Med Sport. 2020;23:164–170. 29. Pinto RS, Gomes N, Radaelli R, Botton CE, Brown LE, Bottaro M. Effect of range of motion on muscle strength and thickness. J Strength Cond Res. 2012;26:2140– 2145. 30. Kamiya K, Masuda T, Tanaka S, Hamazaki N, Matsue Y, Mezzani A, et al. Quadriceps strength as a predictor of mortality in coronary artery disease. Am J Med. 2015;128:1212–1219.