Australian Physiotherapy Journal

Journal of Physiotherapy 60 (2014) 235 Journal of PHYSIOTHERAPY journal homepage: www.elsevier.com/locate/jphys Appraisal Critically Appraised Papers A caregiver-mediated home-based intervention improves physical functioning and social participation in people with chronic stroke Synopsis Summary of: Wang T-C, Tsai AC, Wang J-Y, Lin Y-T, Lin K-L, Impact Scale, Berg Balance Scale, 10-metre Walk Test, 6-minute Chen JJ, et al. Caregiver-mediated intervention can improve Walk Test, Barthel Index, and Caregiver Burden Scale. Results: All physical functional recovery of patients with chronic stroke: a 51 participants completed the study. At the end of the training, the randomized controlled trial. Neurorehabil Neural Repair 2014;doi: experimental group had significantly more improvement than 10.1177/1545968314532030. the control group in scores of the Stroke Impact Scale, including the composite physical (by 11.9 points, 95% CI 6.5 to 17.3), social Question: Does a caregiver-mediated home-based intervention participation (by 11.4 points, 95% CI 2.9 to 19.9), and general improve physical functioning and social participation in people recovery scores (by 17.2 points, 95% CI 10.0 to 24.4). The with chronic stroke? Design: Randomised controlled trial with experimental group also had significantly better outcomes than blinded outcome assessment. Setting: Home-based intervention in controls in the Berg Balance Scale (by 5.3 points, 95% CI 2.0 to 8.6), Taiwan. Participants: Key inclusion criteria were: diagnosis of free-walking velocity (by 8.9 cm/s, 95% CI 2.1 to 15.7), 6-minute hemispheric stroke; > 6 months post-onset; Brunnstrom stage III-V; Walk Test (95% CI 26.3 m, 95% CI 8.2 to 44.4), and Barthel Index (by home dwelling; and having family, friends or co-workers as 6.6 points, 95% CI 1.8 to 11.5). There were no significant differences caregivers. Key exclusion criteria were: recurrent stroke; dementia; between the groups on the Caregiver Burden Scale score. global or receptive aphasia; severe orthopaedic problems; and being Conclusion: The caregiver-mediated home-based intervention in medically unstable. Randomisation of 51 participants allocated 25 to this study improves physical functioning and social participation in the experimental group and 26 to the control group. Interventions: people with chronic stroke. The experimental group was given weekly, personalised caregiver- mediated home-based intervention training by a physiotherapist for Marco YC Pang 12 weeks. The caregiver was asked to encourage the patient to Department of Rehabilitation Sciences, perform the prescribed exercises and task-specific training at least The Hong Kong Polytechnic University, Hong Kong twice weekly. The control group received weekly visits or telephone calls from the therapist without exercise or task-specific training http://dx.doi.org/10.1016/j.jphys.2014.08.012 interventions during the same period. Outcome measures: Stroke Commentary the study is remarkable, and needs to be considered in the context of the flexible and cost-effective model of exercise delivery. A major component of rehabilitation after stroke is exercise therapy that serves to minimise the effects of brain cell damage In the current economic climate, it is imperative that healthcare and optimise motor re-learning.1 Physiotherapists have tradition- professionals identify similar interventions that can aid recovery ally been the mediators of post-stroke exercise therapy, but it has and optimise carer involvement following stroke, while being been suggested that the amount of therapy that is delivered is not mindful of available resources. This study responds to the clear optimal.2 While the evidence for novel exercise interventions to need for the provision of an evidence-based intervention that can promote motor recovery after stroke is growing, the involvement be delivered in the community setting and that is acceptable to of family and carers in a structured program of exercise delivery people with stroke and their carers. after stroke has remained largely unexamined.3 Rose Galvin The approach to exercise delivery by Wang and colleagues School of Physiotherapy, Royal College of Surgeons in Ireland, Ireland builds on the concept that people with stroke and their caregivers take responsibility for the bulk of exercise therapy.4 Wang and References colleague’s study has significant clinical and research implications, as it is the first randomised controlled trial to evaluate the delivery 1. Langhorne P, et al. Lancet Neurol. 2009;8:741–754. of exercise to people with chronic stroke by individuals who are 2. Galvin R, et al. Top Stroke Rehabil. 2008;15:365–377. not healthcare workers. Furthermore, the intervention demon- 3. Galvin R, et al. Stroke. 2011;42:681–686. strated significant gains to the person with stroke at the level of 4. De Weerdt W, et al. Lancet. 2002;359:182–183. impairment, activity limitation and participation restriction, with no additional burden on the carer. Finally, the lack of attrition from http://dx.doi.org/10.1016/j.jphys.2014.08.014 1836-9553/ß 2014 Australian Physiotherapy Association. Published by Elsevier B.V. All rights reserved.

Journal of Physiotherapy 60 (2014) 189–200 Journal of PHYSIOTHERAPY journal homepage: www.elsevier.com/locate/jphys Research Aerobic capacity and upper limb strength are reduced in women diagnosed with breast cancer: a systematic review Sarah E Neil-Sztramko, Amy A Kirkham, Stanley H Hung, Negin Niksirat, Kei Nishikawa, Kristin L Campbell University of British Colombia, Vancouver, Canada KEY WORDS ABSTRACT Meta-analysis Question: What are typical values of physical function for women diagnosed with breast cancer and how Breast neoplasms do these compare to normative data? Design: Systematic review with meta-analysis. Participants: Physical fitness Women diagnosed with breast cancer who were before, during or after treatment. Outcome measures: Physical endurance Physical function was divided into three categories: aerobic capacity, upper and lower extremity Muscle strength muscular fitness, and mobility. Measures of aerobic capacity included field tests (6-minute walk test, 12- minute walk tests, Rockport 1-mile test, and 2-km walk time) and submaximal/maximal exercise tests on a treadmill or cycle ergometer. Measures of upper and lower extremity muscular fitness included grip strength, one repetition maximum (bench, chest or leg press), muscle endurance tests, and chair stands. The only measure of mobility was the Timed Up and Go test. Results: Of the 1978 studies identified, 85 were eligible for inclusion. Wide ranges of values were reported, reflecting the range of ages, disease severity, treatment type and time since treatment of participants. Aerobic fitness values were generally below average, although 6-minute walk time was closer to population norms. Upper and lower extremity strength was lower than population norms for women who were currently receiving cancer treatment. Lower extremity strength was above population norms for women who had completed treatment. Conclusion: Aerobic capacity and upper extremity strength in women diagnosed with breast cancer are generally lower than population norms. Assessment of values for lower extremity strength is less conclusive. As more research is published, expected values for sub-groups by age, treatment, and co- morbidities should be developed. [Neil-Sztramko SE, Kirkham AA, Hung SH, Niksirat N, Nishikawa K Campbell KL (2014) Aerobic capacity and upper limb strength are reduced in women diagnosed with breast cancer: a systematic review. Journal of Physiotherapy 60: 189–200] ß 2014 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/3.0/). Introduction are commonly assessed in exercise oncology literature, and each has established objective outcome measures available for Worldwide, breast cancer remains the most commonly comparison. diagnosed cancer in women.1 Due to advancements in treatment approaches for breast cancer, the 5-year survival rate has Aerobic capacity improved dramatically, and in Canada is approximately 88%.2 Despite the efficacy of treatment in improving survival, women Declines in aerobic capacity have been observed during breast who have undergone treatment for breast cancer face both acute cancer treatment, which is likely a combination of the direct and and chronic impairments in various aspects of physical function indirect effects of the treatment itself, and associated reduction in as a result of their treatment, which may involve a combination physical activity leading to deconditioning.4 Maximal oxygen of surgery, chemotherapy, radiation therapy, hormonal therapy consumption (VO2max) – the upper limit to the rate of oxygen or other targeted biological therapies.3 Physiotherapists have utilisation, as measured by a cardiopulmonary exercise test – is the the potential to play an important role in cancer care by gold standard measurement of cardiorespiratory fitness and the identifying and monitoring changes in physical function during capacity for physical work.5 In clinical populations, VO2max may and following breast cancer treatment, and by prescribing not be achieved during a cardiopulmonary exercise test, so the interventions to address deficits in physical function. For the peak oxygen consumption (VO2peak) is used instead. VO2peak is purposes of the present review, three main aspects of physical associated with all-cause,6 cardiovascular disease-specific7,8 and function have been selected: aerobic capacity, muscular fitness breast cancer-specific9 mortality. A recent cross-sectional study of the upper and lower extremities, and mobility. These aspects reported that women diagnosed with breast cancer have a of physical function were selected because they represent VO2peak on average 27% lower than that expected for healthy clinically relevant areas of focus for physical therapists, they sedentary women.10 http://dx.doi.org/10.1016/j.jphys.2014.09.005 1836-9553/ß 2014 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/3.0/).

190 Neil-Sztramko et al: Physical function after breast cancer diagnosis Although VO2peak has a strong association with health strength and has been shown to be a predictor of postoperative outcomes, cardiopulmonary exercise testing requires expensive, complications, functional limitations, disability and mortality.22 specialised equipment and medical supervision for high-risk individuals, thereby limiting its feasibility. A submaximal exercise Mobility test, such as a progressive exercise test that is terminated at 85% of age-predicted maximal heart rate or 70% of heart rate reserve, is Mobility assessment is intended to be a functional measure that often a more feasible alternative in clinical practice because it poses is influenced by both muscular strength and agility. A common field less risk and can be done without collection of expired metabolic test, the Timed Up and Go (TUG) test, requires a participant to gases. VO2max can be estimated with a submaximal exercise test.11 perform a sequence of tasks that are all critical for independent mobility: rise from a chair, walk 3 metres, turn around, walk back to The 6-minute walk test (6MWT) is a widely used field test that the chair, and sit down.23 The test outcome is the total time required offers a more practical approach to quantifying physical capacity to complete the sequence. As such, the TUG test provides an overall than maximal or submaximal cardiopulmonary exercise tests. The assessment of mobility and does not identify problems with 6MWT measures the distance walked over a flat, hard surface particular tasks.23 This test is reliable and valid for quantifying in 6 minutes.12 The 6MWT distance correlates with VO2peak functional mobility and for assessing clinical change over time.24 (r = 0.59 to 0.73)12,13 and is more a measure of an individual’s Although intra-rater and inter-rater reliability of the test are high ability to perform daily activities than a surrogate measure of (ICC = 0.92 to 0.96), test-retest reliability is moderate (ICC = 0.56),25 aerobic capacity.12 Although there is concern regarding the need which is potentially due to a learning effect. Construct validity of this for a familiarisation trial to account for a potential learning effect, functional test has been supported by correlations with a number of the test-retest reliability of the 6MWT was recently reported for a functional measurements including: gait speed (r = 0.75), postural cancer population (ICC = 0.93, 95% CI 0.86 to 0.97), and the 6MWT sway (r = 0.48), step length (r = 0.74), stair test (r = 0.59) and step was significantly correlated with VO2peak (r = 0.67).14 Other field frequency (r = 0.59).25 Other assessments of mobility include tests assessing aerobic capacity without the need for expensive measuring gait speed, time to ascend or descend a certain number equipment include the Cooper 12-minute walk test (12MWT),12 of stairs, and the time it takes to get down and up from the floor. Rockport 1-mile test15 and 2-km walk time.16 In healthy populations, normative values of a variety of the tests Upper and lower extremity muscular fitness described above have been published. These values help phy- siotherapists and other health professionals interpret a patient’s Muscular fitness is a component of physical function that consists result on a specific test relative to others of similar age and gender of muscular strength, endurance and power.11 Following surgery for and may provide a goal for individuals and clinicians to attain. breast cancer, women may experience substantial impairment in Research to date has documented the decline in various aspects of upper extremity function. Functional limitations, including decline physical function during and following breast cancer treatment. In in strength and range of motion, may continue after acute recovery order to publish average values for this clinical population, a large from surgery is complete.17 Deconditioning during active cancer sample of participants is required. The aim of this review was to treatment (ie, chemotherapy and radiation) may also contribute to summarise the available data that have been published in studies declines in upper and lower extremity strength and endurance. that measured physical function in women who have been Aromatase inhibitors, commonly prescribed following the comple- diagnosed with breast cancer, to generate a resource for tion of chemotherapy and radiation therapy, are also associated with physiotherapists using the tests that are most commonly used musculoskeletal symptoms such as pain, which may also reduce in this field of research. The second aim is to compare reported participation in physical activity, further contribute to decondition- values to published normative data, where available. ing and, in turn, impact muscular fitness.18 Methods Muscular strength refers to the ability to exert force. The gold standard for assessment of muscle strength is the force exerted in a Identification and selection of studies maximum voluntary contraction with force output measured by a computerised dynamometer.19 This type of equipment is very Due to the wide range of assessment tools available, the review expensive and, thus, not commonly used outside of a research was limited to the most commonly used, objective and validated setting. In the field, strength is traditionally evaluated with a one tests reported in the exercise oncology literature that would also repetition maximum (1RM) or maximum voluntary contraction, be relevant to physiotherapists, as identified in a previous but four to 15 repetition tests to estimate 1RM have also been used literature review.26 A list of MeSH terms and key words related to assess strength.11 General upper extremity strength is typically to breast cancer, physical function, and the specific outcomes of assessed using a chest or bench press, while lower extremity interest were developed (see Appendix 1 in the eAddenda). strength is commonly assessed using leg press or leg extension.11 MEDLINE, Embase and CINAHL were searched using these terms up Alternatively, muscle strength can be measured objectively in a to and including 27 December, 2012. Included studies were clinical setting using a portable, tester-reliant tool called a hand- required to meet all inclusion criteria (Box 1). Case studies were held dynamometer. Inter-tester reliability coefficients for this tool excluded, as were studies including participants with other types range from –0.19 to 0.99, depending on the study, and appears to of cancer, unless values were reported separately by cancer type. be more reliable for upper than lower body strength measure- Studies that were limited to women with metastatic breast cancer ments.20 Muscular endurance refers to the ability to successively were also excluded; however, we did not otherwise exclude perform exertions of force and is evaluated via the maximum studies on the basis of individual study eligibility criteria. Lack of number of repetitions at a percentage of the 1RM or body weight, consensus about eligibility was resolved through discussion. often with the repetitions performed at a standard rate.11 The chair stand test is another commonly used field test of lower body Data extraction and synthesis strength and involves either the number of chair stands performed in 30 seconds or the amount of time required to perform a Relevant data were extracted from each identified paper, predetermined number of chair stands. The 30-second chair stand including demographic characteristics of the study participants, has moderately high test-retest reliability (ICC = 0.89) and details of the study design, name of the test used, specifics of the moderate construct validity as demonstrated by a correlation test protocol, and reported values of the selected physical function with the leg press (r = 0.77).21 tests. Data were extracted for the full study sample where available, and separate group data were pooled for simplicity.27 Finally, a commonly reported measure of global muscular strength is grip strength. Due to the internal consistency of strength measurements, grip strength may be used to characterise overall

ResearcFG]g$DiTue(r)1[_Ih 191 Box 1. Inclusion criteria. Design  Randomised trials  Non-randomised intervention studies  Observational studies Participants  Women diagnosed with breast cancer  Before, during or after treatment Intervention  Any intervention or no intervention Outcome measures  Aerobic capacity (maximal or submaximal exercise test, six or twelve minute walk test, Rockport 1-mile test, 2-km walk time)  Upper extremity strength and endurance (grip strength, bench/chest press)  Lower extremity strength and endurance (leg press, knee flexion/extension, chair stands)  Mobility (Timed Up and Go) A second author checked the data extraction. Where baseline values of outcomes of interest were not reported, authors were contacted for missing data. Of 13 authors contacted, data were received from three. Where necessary, data were converted to metric units. The selection of the age range for normative values reported was based on the average age and mean body weight of participants in the included studies. Data analysis Figure 1. Flow of studies through the review. For outcomes in which at least three different studies used a Aerobic capacity comparable protocol, a meta-analysis was conducted. Using methods described by Neyeloff et al27 for descriptive data analysis, The most common test used to assess aerobic capacity was a the pooled mean for each outcome was calculated using a random- maximal cardiopulmonary exercise test (n = 16) using either a effects model. Studies for which the mean and standard deviation cycle ergometer (n = 9) or treadmill (n = 8) protocol (see Table 3 in were not reported in the paper (eg, median and/or range were the eAddenda). Pooled relative VO2peak was a mean of 23.7 mL/kg/ reported instead) were not included in the meta-analysis. All studies min (95% CI 20.4 to 27.0) for women on treatment and 22.8 mL/kg/ reporting the specific outcome of interest were plotted on the same min (95% CI 20.7 to 24.9) for women off treatment (Figure 2). The forest plot, however pooled means were calculated separately for pooled absolute VO2peak was a mean of 1.65 L/min (95% CI: 1.59 to studies involving participants who were ‘on treatment’ and ‘off 1.72) from study groups on treatment and 1.60 L/min (95% CI treatment’. ‘On treatment’ was defined as measures taken prior to 1.48 to 1.72) from study groups off treatment (Figure 3). Compared the completion of surgery, chemotherapy or radiation therapy. ‘Off to published normative data, pooled means of VO2peak fell into the treatment’ was defined as studies in which authors report that ‘very poor’ category for women age 50 to 59 (Table 2).11 No participants had completed surgery, chemotherapy and/or radiation heterogeneity was identified (all I2 values < 30%). therapy, but may have still been taking hormonal therapies. Heterogeneity was assessed using I2 under the random-effects Submaximal exercise tests were used to predict VO2max in model using the methods described by Neyeloff et al.27 15 studies, more commonly using a treadmill (n = 12) than a cycle ergometer (n = 3) protocol. Predicted VO2max values tended to be Results higher than measured VO2peak. The pooled mean for predicted VO2max for women on and off treatment was 25.2 mL/kg/min (95% The search identified 1978 papers, of which 361 were retrieved CI 19.1 to 31.3) and 23.9 mL/kg/min (95% CI 22.5 to 25.4), and screened for eligibility and 85 met our inclusion criteria respectively (Figure 4). These mean values fall into the ‘very poor’ (Figure 1). A full list of included studies can be found in Appendix 2 category for women age 50 to 59 (Table 2).11 No heterogeneity was (in the eAddenda). The most common reasons for exclusion were identified (all I2 values < 30%). that the outcomes assessed did not meet the inclusion criteria, or the studies did not examine women diagnosed with breast cancer. The 6MWT was used as a measure of aerobic capacity in nine Study designs and relevant participant characteristics are listed in studies. The pooled mean value for distance walked was 523 m Table 1. Of the studies included, 42 were randomised trials, (95% CI 499 to 548) for women on treatment, and 500 m (95% CI 19 were non-randomised intervention studies, and 24 were 476 to 524) in women off treatment (Figure 5). These pooled means observational studies with no intervention. The majority of studies fall between the 25th and 50th percentiles of community-dwelling (n = 61) included women who were off treatment, while others adults aged 60 to 64 (Table 2).28 The 12MWT was used in included women following surgery but before chemotherapy/ 11 studies. The pooled mean value for distance walked was 1020 m radiation therapy (n = 20) and/or during chemotherapy/radiation (95% CI 982 to 1058) in women on treatment and 904 m (95% CI therapy (n = 9), and for the purposes of the present review were 831 to 976) in women off treatment (Figure 6). All I2 values classified as on treatment (n = 28). Some observational studies were < 30% except for the 6MWT in the off-treatment groups only, included assessments at multiple time points and were included in which had moderate heterogeneity. Other less commonly used both groups. Normative values for comparison are presented in tests include the 2-km walk time with values ranging from 16.9 to Table 2. 18.9 minutes, and Rockport 1-mile test (reported values of 17.45 and 17.65 minutes). There were no published norms identified for the 12MWT, 2-km walk test or Rockport 1-mile test.

Table 1 Included studies, study design and patient characteristics. References to included studies are available in Appendix 2 i Reference Study n Age (yr) BMI (kg/m2) Stage (% On treatment design mean (SD) mean (SD/range) Anderson 2012 RCT 104 54 (32 to 82) NR I: 49; II: 38; Beurskens 2007 RCT 30 55 (11) NR NR Campbell 2005 RCT 22 47 (5) NR I-III Courneya 2007 RCT 242 49 (25 to 78) 27 (6) I: 25; IIA: 41; IIIA: 15 Drouin 2005 RCT 21 50 (8) NR DCIS: 24; I: 14 III: 38 Haines 2010 RCT 89 55 (9) NR NR Haykowsky 2009 Pre-post 17 53 (7) BW (kg) 78 (21) I: 29; IIA: 35; III: 12 Johansson 2001 Obs 61 56 (10) NR NR Kaya 2010 Obs 67 54 (12) NR NR Kilgour 2008 RCT 27 50 (5) BW(kg) 75 (16) NR Kim 2006 RCT 41 50 (6) 29 (6) 0: 5; I: 41; II: 3 Kolden 2002 Pre-post 40 55 (8) BW (lb) 155 (25) I: 32; II: 55; Ligibel 2010 Pre-post 41 47 (7) NR I: 27; II: 37; Mock 1994 RCT 14 44 NR I: 14; II: 8 Mock 2001 50 26 (5) Mock 2005 Pre-post 119 48 (11) 26 (5) I: 54; II: 40; Morimoto 2003 RCT 72 52 (9) NR 0: 24; I: 43; II: Nikander 2007 28 28 (5) Postma 1995 Pre-post 50 (10) NR I/II RCT 7 52 (5) NR Obs NR 56 a (43 to 71) Reitman 2003 Obs 204 56 (12) NR I: 42; IIa: 41; Rietman 2004 189 I: 42; IIA: 42; Reitman 2006 Pre-post 181 56 (10) BW (kg) 68 (12) Schneider 2007 Pre-post 47 (35 to 57) NR NR Schwartz 2000a Pre-post 17 47 (27 to 71) II/III: 70 Schwartz 2000b Pre-post 27 23 (0) II: 8; III: 73; Schwartz 2001 71 47 (8) NR Schwartz 2007 RCT 61 48 (8) II: 54 Wang 2011 RCT 66 50 (10) BW (kg) 69 (2) I: 23; II: 58; 72 BW (kg) 55 (7) I: 22; II: 7

in the eAddenda. 192 Neil-Sztramko et al: Physical function after breast cancer diagnosis %) Patient characteristics Treatment type (%) Eligibility Time since treatment criteria mean (SD) No LD III: 12 0 to 2 wk Mast: 50; Lump: 46 Shoulder pain >1 2 wk Mast: 77; Lump: 23; ALND: 100 IIB: 20; NA Inactive 5 NA CT: 27; RT: 27; CT+RT: 45 – 4; II: 24; Lump: 59; Mast: 41; CT: 100 NA Inactive IIB: 24; Sx: 24; Sx+CT: 76; RT: 100 NA – 37; III: 17 Immediately before Sx: 100; CT: 36; RT: 92; HT: 39 HER1; 50% ejection III: 13 Mast: 53; Lump: 47; CT: 100 III: 34 Trastuzumab fraction 86 Pre-op, 6 mo, 1 yr, 2 yr Mast: 100; CT: 25; RT: 57; HT: 25 All ALND + Mast Mast: 87; Lump: 13; CT: 90; RT: 44 LD node removal ; III: 6 9 mo (2 to 67) 30; III: 3 3 d post Sx Mast: 100 Mast + ALND Sx CT: 49; RT: 34; CT+RT: 17 Inactive IIb: 17 ‘‘Shortly after diagnosis’’ Mast: 45; Lump: 60; CT: 65; – IIB: 16 83% 12 mo post diagnosis RT: 60; HT: 50 Inactive 0 NA Lump: 34; Mast: 56; CT: 98; RT: 54 – IV: 18 NA – III: 20 NA CT: 100; Sx+RT: 79; Mast: 21 78 NA Lump: 62; Mast: 38; RT: 64; CT: 36 Inactive NA – NA CT: 42; RT: 58 – NA Mast: 54; Lump: 46 CT: 25; HT: 21; CT+HT: 54 Anthracycline- 6 wk post Sx resistant; Paclitaxel 1 yr post Sx CT: 100 2 yr post Sx ALND Mast: 42; Lump: 58 NR RT: 67; CT: 34; HT: 38 – >21 d post Sx – >21 d post Sx Sx: 99; CT: 50; RT: 43 – >21 d post Sx Sx: 100 – Sx: 100 Inactive NA Chinese; BMI 30 NA Mast: 74 NR NA

Table 1 (Continued ) Study n Age (yr) BMI (kg/m2) Stage (% Reference design mean mean (SD) II/III Off treatment RCT 50 (SD/range) 27 (5) Ahmed Omar 2011 RCT 46 DCIS: 4; I: 28; RCT 60 54 (3) 27 (7) III: 11 Ahmed 2006 RCT 18 Pre-post 14 52 (1) 29 (8) DCIS: 22; I: 28; Basen-Engquist 2006 RCT 67 18; IV: 3; unkn Obs 95 55 (8) 27 (4) I-IIIA Brdareski 2012 Pre-post 27 30 (4) I: 7; II: 64; I Campbell 2012 RCT 52 52 (8) RCT 108 55 (8) NR I: 24; II: 58; I Cantarero- Pre-post 29 Villanueva 2012a Obs 50 49 (8) NR I: 30; II: 53; I Obs 12 Cantarero- NR BW (kg) 71 (12) DCIS: 7; I: 22; Villanueva 2012b Obs 7 III: 33 RCT 42 58 (7) 29 (7) Cheema 2006 I: 40; IIA: 33; RCT 87 59 (6) 28 (0) IIIA: 6 Courneya 2003 NR Pre-post 11 51 (10) NR Daley 2007 Obs 214 27 (5) I: 45; II: 5 60 (7) 26 (6) I/II Damush 2006 RCT 16 59 (10) Dawes, 2008 Pre-post 75 26 (4) I: 42; IIA: 50; Dolan 2012 Pre-post 96 55 (6) NR 25 II: 43; III: Evans 2009 Obs 49 54 (7) NR NR Eyigor 2010 RCT 61 49 (6) Obs 26 (6) 0: 7; I: 44; II: 3 Fillion 2008 52 (10) NR I/II Garner 2008 51 (6) 25 (4) I: 27; II: 31; Hayes 2005 53 (10) 26 (9) BW(lb) 168 (39) I: 44; II: 5 Herrero 2006 50 (8) NR Hokken 2009 49 (9) NR NR Hsieh 2008 58 (10) 27 (7) Hughes 2008 50 (8) I/II: 60; III: 4 Hutnick 2005 50 (7) NR I/II: 90 Johansson 2001 56 (10) NR

%) Patient characteristics Treatment type (%) Eligibility Time since treatment criteria mean (SD) LD ; II: 57; 40 (9) mo Mast: 82; Lump: 18; RT+CT: 38; Research II: 25; III: RT+HT: 14; CT+RT+HT: 48 Inactive, stable BW nown: 3 IG: 13 (5 to 32); CG: III: 14 13 (4 to 37) mo RT: 74; CT: 87; Sx: 100; HT: 87 Inactive IIIA: 18 IIIA: 17 IG: 39.2 (16.7); CG: Lump: 40; Mast 60; None: 12; RT: < 65 yr old ; II: 22; 37.1 (14.1) mo post diagnosis 22; CT: 22; RT+CT: 45 BMI 25 to 35; Inactive IIB: 21; IG1: 5 (4); IG2: 3 (3) yr Sx: 100; CT: 72; RT: 72; HT: 50 Functional problems 55 24 (22) mo Sx: 7; Sx+RT: 14; Sx+CT: 7; ; IIIA: 8 Sx+RT+CT: 71 Functional problems <12 mo: 76%; >12 mo: 24% 57 Lump: 63; Mast: 37; RT: 3; CT: Dragon boat team 6 mo 9; RT+CT: 88 post season 35; III: 15 5 (5) yr Lump: 68; Mast: 32; HT: 78; Age 50 to 69 yr, PostM III: 32 Trastuzumab: 12 56 14 (6) mo Inactive Lump: 59; Mast: 48; RT: 67; CT: 4; IV: 4 IG: 18 (7); ExCG: 18 (7); 52; HT: 22  50 yr 0 CG: 17 (6) mo LD symptom Mast: 54; Lump: 46; RT: 71; CT: 3 yr post diagnosis 40; HT: 46 – 193 NR Mast: 53; Lump: 47; CT: 74; RT: Inactive <5 yr: 50%; 5 to 10 yr: 79; HT: 73 Inactive 33%; >10 yr: 17% post Mast: 59; CT: 62 – diagnosis Sx: 100% 6 mo Inactive, PostM Mast: 83; Lump: 17; CT: 83; RT: – IG: 39 (40); CG: 38 (52) 75; HT: 58 mo post diagnosis Inactive, ALND, PostM NR – 257 (107) d post diagnosis Mast: 100 – 4 (3) yr post diagnosis Mast: 14; Lump: 86; CT: 56; RT: Hispanic only 6 mo post diagnosis 100; HT: 74 – NR 36 mo All ALND + Mast AC: 23 (13); FEC: 17 (9) wk Mast: 28; Lump: 72; RT: 71; CT: 41; HT: 42 Immediately 61 (35) mo post diagnosis Lump: 56; Mast: 44; CT: 100 CT: 100; RT: 73 2 wk to 2 mo Pre-op, 6 mo, 1 yr, 2 yr Mast: 69; Lump: 28, CT: 57; RT: 44 NR Lump: 53; Mast: 45; RT+CT: 71 CT: 25; RT: 57; HT: 25

Table 1 (Continued ) Study n Age (yr) BMI (kg/m2) Stage (% Reference design mean mean (SD) NR Jones 2007b Obs 47 (SD/range) 28 (5) I/II: 65 Jones 2007a Obs 26 57 (7) 29 (6) Kaltsatou 2011 48 (9) Lane 2005 Linterman 2011 RCT 27 57 (5) NR NR Merchant 2008 Pre-post 16 52 (7) 24 (3) I: 44; II: 38; Mulero Portela 2008 Musanti 2012 Obs 33 AI: 64a (51 to 74); AI: 24a (18 to 45); NR Mustian 2006 Tam: 61a (54 to 68) Tam: 23a (21 to 26) Mutrie 2007 Obs 40 NR Neil 2012 57 (12) NR Nikander 2012 RCT 34 I: 15; II: 29; III: Nuri 2012 RCT 55 51 (6) 30 (6) I: 45; II: 44; O&Neill 2006 RCT 21 51 (8) BW (lb) 169 (36) 0/IIIb Pinto 2005 52 (9) Rietman 2004b RCT 201 26 (5) 0/III Reitman 2006 52 (10) Rietman 2004a Obs 27 27 (6) I/IIIa Rogers 2009 RCT 67 53 (10) I: 15; II: 60; Rogers 2013 RCT 29 53 (8) 25 (5) Saarto 2012 Pre-post 17 58 (6) BW(kg) 72 (16) I/IIIb NR Schmitz 2010 RCT 86 64 (45 to 76) 28 (5) Schneider 2007 NR 0: 16; I: 37; Scott 2013 Obs 189 53 (9) 181 28 (9) I: 42; IIA: 42; Obs 56 (12) RCT 55 NR 0: 4; I: 47; IIa: 3 RCT 41 57 (13) I: 29; II: 51; RCT 28 53 (9) BW (kg) 74 (17) I: 54; II: 32; 498 31 (6) NR RCT 56 (11) 32 (5) 154 IG, PreM: 26 (0) DCIS: 1; I: 56; II: Pre-post 46 (36 to 54); PostM: RCT 96 58 (48 to 68); CG, PreM: 28 (0) NR 90 46 (35 to 57); PostM: I/II/III 58 (46 to 68) BW (kg) A: 76 (19) 30 (0) 55 (12) 56 (10) 56 (9)

%) Patient characteristics Treatment type (%) Eligibility 194 Neil-Sztramko et al: Physical function after breast cancer diagnosis 5 Time since treatment criteria mean (SD) Lump: 28; Mast: 72; RT: 98; CT: 100 Mast: 52; CT: 100; RT: 65; PostM 3 yr HT: 62 Node positive or 20 (10) mo high-risk node negative; NR operable HER2/neu III: 19 2.2 yr Lump: 75; Mast: 25; Lump+Mast: >6 mo – 26; IV: 3 25; RT: 94 – III: 11 NR CT: 12 PostM III: 25 29 (21) mo Mast: 40; Lump: 60; CT: 10; RT: 43; CT+RT: 30; HT: 53 Unilateral Sx > 6 mo prior II: 47 >2 mo Lump: 53; Mast: 47 IIB: 16 6 wk to 2 yr CT: 87; RT: 73; HT: 56 – 35; IIb: 15 1 wk to 30 mo Inactive III: 20 Lump: 61; Mast: 39; CT: 84; Inactive III: 14 162 (74) d RT: 61; HT: 56 Inactive : 9; III: 34 F: 21 (18); CG: 23 (19) mo CT: 8; RT: 28; CT+RT: 64; Mast: < 4 mo 40; Lump: 60 RT NR – CT+RT: 78; RT: 22; HT: 52 3 to 28 yr CT: 90; RT: 78; HT: 88 Inactive, Stable BW, PostM NR LD IG: 2 (1); CG: 2 (1) yr Mast: 100; RT: 29; CT: 12; Inactive; Diagnosis <5 yr 1 yr post Sx RT+CT: 47 2 yr post Sx ALND Lump: 72; Mast: 28; RT: 69; CT: 3 (1) yr 56; HT: 62 Mast + ALND 33 (36) mo Current HT; Inactive 74 (71) mo post diagnosis RT: 67; CT: 34; HT: 38 IG, PreM: 33.5 (8.2); PostM: Inactive 33.9 (9.3); CG, PreM: RT: 43; CT: 20 – 31.3 (9.2); PostM: Sx: 100; CT: 83; RT: 17; HT: 100 33.6 (8.7) wk 2 nodes removed; No LD Sx: 100; CT: 75; RT: 79; HT: 50 IG: 39 (15); CG: 42 (16) m Sx: 100; CT: 91; RT: 78; HT: 84 – o post diagnosis BMI >25; Inactive NR Sx: 100; CT: 71; RT: 76; HT: 35 3 to 18 mo Sx: 99; CT: 50; RT: 43 Mast: 41; Lump: 59; CT: 56; RT: 83; HT: 78

Table 1 (Continued ) Study n Age (yr) BMI (kg/m2) Stage (% Reference design mean (SD) NR mean Smoot 2010 Obs 144 (SD/range) 26.1 (5) 56 (12) Sprod 2005 RCT 12 53 (3) BW (lb) 179 (17) NR Sprod 2010 Pre-post 114 59 (0) BW (lb) 169 (11) NR Taylor 2010 RCT 260 55 (9) 31 (5) IA: 42; IB: 3; I IIB: 2; IIIA: 23 Tolentino 2010 Obs 22 49 (9) 27 (4) 0.1; DCIS: Tosti 2011 Obs 7 51 (3) 29 (1) Turner 2004 Pre-post 47 (8) IIA: 18; IIB: 41; Twiss 2009 RCT 10 59 (8) NR I: 14; II: 43; Wampler 2007 Obs 223 50 (9) 27 (4) NR Winters-Stone 2008 Obs 68 (7) BW (kg) 68 (9) 0/I/II Winters-Stone 2011 Obs 20 59 (10) NR Winters-Stone 2012 RCT 47 62 (10) NR 59 28 (7) 0: 3; I: 37; II: 2 Yuen 2007 RCT 106 54 (12) IG: 30 (6) 0: 5; I: 29; II: 3 CG: 30 (6) 22 0: 6; I: 40; II: 4 NR NR AI = Aromotase inhibitors, AET = aerobic exercise training, BW = body weight, BMI = Body Mass Index, CG = control group group, Lump = lumpectomy, LD = lymphoedema, Mast = mastectomy, NR = not reported, Obs = observational study, Pre-pos trial, RET = resistance exercise group, RT = radiation therapy, SE = standard error, Sx = surgery, Tam = Tamoxifen. a Median reported instead of mean.

%) Patient characteristics Treatment type (%) Eligibility Time since treatment criteria mean (SD) No LD: 5 (4) Lump: 57; Mast: 43; CT: – LD: 7 (6) yr post diagnosis 70; RT: 74 IG: 20 (5); CG: 26 (5) mo Sx: 100; Mast: 92; CT: – 83; RT: 58 IG3: 13 (1); IG6: 29 (1); CT: 60; RT: 40 – CG: 35 (5) mo IIA: 20; 5 (3) yr post diagnosis Sx: 10; Sx+CT: 13; Sx+ RT: BMI > 25.0 3; IIIB: 26; Sx+ CT+RT: 49 :6 ; IIIA: 41 <14 mo post diagnosis Sx+CT+RT: 100 – III: 43 < 6 mo Sx: 100; CT: 71 RT: 86; HT: 86 – 17 a (4 to 60) mo Sx+CT+RT: 100 – 6 (6) yr Sx: 98; RT: 45; CT: 68; HT: 50 BMD < –1.0; Inactive PostM < 30 d CT: 100 Taxane-based CT 22; III: 7 8 (7) yr post diagnosis RT: 58; CT: 37; HT: 48 >60 yr old 39; III: 19 < 2 yr CT: 29; HT: 32; CT+HT: 39 < 70 yr, PostM Research 42; III: 6 IG: 57 (40); CG: 65 (35) CT: 60; RT: 88; HT: 57 > 50 yr at diagnosis; mo post diagnosis Inactive, PostM 9 d to 35 mo Sx: 100; Mast: 41; Lump: 55; Inactive; Moderate fatigue CT: 82; RT: 77 p, CT = chemotherapy, HER-2/neu = human epidermal growth factor receptor 2, HT = hormonal therapy, IG = intervention st = pre-post non-randomised intervention, PreM = premenopausal, PostM = postmenopausal, RCT = randomised controlled 195

]Gre_$FTI)[ugi2(D196 Neil-Sztramko et al: Physical function after breast cancer diagnosis Table 2 Courneya 2007 (CG) Normative values for common tests of physical function. Aerobic capacity Courneya 2007 (RET) Cardiopulmonary exercise test, VO2max (mL/kg/min); females only 11 Courneya 2007 (AET) Age 40 to 49 Age 50 to 59 Age 60 to 69 very poor 22.2 to 28.2 20.1 to 25.8 19.5 to 23.9 Haykowsky 2009 poor 29.4 to 32.3 26.8 to 29.4 24.6 to 26.6 Cheema 2006 (IG) fair 32.8 to 35.2 29.9 to 32.3 27.3 to 29.4 good 35.9 to 38.6 32.6 to 35.2 29.7 to 32.3 Courneya 2003 (IG) excellent 39.6 to 43.1 36.7 to 38.8 32.7 to 35.9 superior 45.3 to 51.1 41.0 to 46.1 37.8 to 42.4 Courneya 2003 (CG) 6-min walk distance (m) in community dwelling adults, reported as mean (SD) 28 Dolan 2012 (Cycle) 10th percentile Age 60 to 64 Age 65 to 69 Age 70 to 74 Dolan 2012 (Treadmill) 25th percentile 452.6 402.3 384.0 50th percentile 498.3 457.2 438.9 Garner 2008 553.2 521.2 502.9 75th percentile 603.5 580.6 562.4 Herrero 2006 (IG) 90th percentile 649.2 635.5 617.2 Upper extremity strength Herrero 2006 (CG) Grip strength (kg) 29 Jones 2007a 5th percentile Age 40 to 49 Age 50 to 59 Age 60 to 69 Jones 2007b median 21.8 20.4 18.1 28.6 27.7 25.4 Neil 2012 (Fatigue) Bench press one repetition max (weight pushed/body weight) 11 Age 40 to 49 Age 50 to 59 Age 60+ Neil 2012 (CG) very poor <0.35 to 0.42 < 0.31 to 0.38 < 0.26 to 0.36 Tolentino 2010 poor 0.43 to 0.48 0.39 to 0.43 0.38 to 0.41 fair 0.50 to 0.53 0.44 to 0.47 0.43 to 0.46 Pooled, on treatment Pooled, off treatment good 0.54 to 0.60 0.48 to 053 0.47 to 0.53 16 excellent 0.62 to 0.71 0.55 to 0.61 0.54 to 0.64 superior > 0.77 > 0.68 > 0.72 20 24 28 32 Elbow flexion (kg) obtained using hand-held dynamometry, reported as VO2peak (mL/kg/min) mean (95% CI) mean (SD) 42 Age 50 to 59 Age 60 to 69 Age 70 to 79 dominant 17.0 (2.9) 16.0 (3.0) 14.1 (2.7) Figure 2. Forest plot of weighted mean (95% CI) VO2peak corrected for body weight from a maximal exercise test. References to included studies are available in non-dominant 16.3 (2.7) 15.4 (2.7) 14.4 (2.4) Appendix 2 in the eAddenda. AET = aerobic exercise training, CG = control group, IG = intervention group, RET = Lower extremity strength resistance exercise training. Leg press one repetition max (weight pushed/body weight) 11 Age 40 to 49 Age 50 to 59 Age 60+ 0.72 to 0.85 well below average 0.94 to 1.02 0.78 to 0.88 0.88 to 0.93 reported in weight pushed per kg of body weight, but for a woman 0.99 to 1.04 weighing 70 kg, these pooled values fall into the ‘very poor’ below average 1.08 to 1.13 0.95 to 0.99 1.13 to 1.18 category across all age groups (Table 2).11 Other methods of assessing upper extremity strength include a bench press 6RM, average 1.18 to 1.23 1.05 to 1.10 > 1.32 bench press endurance with various protocols, and elbow flexion. above average 1.29 to 1.37 1.17 to 1.25 Age 70 to 74 12.8 (3.1) well above average > 1.48 > 1.37 Chair stands (number in 30s), reported as mean (SD) 28 Age 60 to 64 Age 65 to 69 13.3 (3.6) 13.7 (3.5) Lower extremity muscular fitness Knee flexion (kg) obtained using hand-held dynamometry, reported as The most commonly reported test of lower extremity strength mean (SD) 42 was the 1RM for leg press, estimated in three studies and measured in five studies (see Table 4 in the eAddenda). The pooled mean for Age 50 to 59 Age 60 to 69 Age 70 to 79 1RM was 67.6 kg (95% CI 61.2 to 73.8) for women on treatment and 95.8 kg (95% CI 88.3 to 103.4) for women off treatment (Figure 9). dominant 17.2 (4.1) 16.0 (2.8) 14.0 (3.5) Heterogeneity was found to be substantial for women off treatment only (I2 = 69%). Reported normative values are reported non-dominant 17.3 (4.7) 15.6 (3.0) 14.4 (3.9) in weight pushed per kg of body weight, but for a woman weighing 70 kg, values for women on treatment fall into the ‘below average’ Mobility category for women aged 50 to 59, while values for women off treatment fall into the ‘above average’ category for women aged Timed Up and Go (s), reported as mean (SD) 50 to 59 (Table 2).11 A leg-press protocol was also used to measure maximum isometric contraction and muscle endurance. Other 3-m course 43 Age 40 to 49 Age 50 to 59 Age 60 to 69 protocols requiring resistance-training equipment include knee 7.24 (0.17) flexion and knee extension machines. Chair stands were also used 8-foot course 28 6.24 (0.67) 6.44 (0.17) Age 70 to 74 as a functional measure of lower extremity function (n = 7), Age 60 to 64 Age 65 to 69 although pooled analysis was not possible due to the heterogeneity 6.0 (1.3) of protocols used. 5.4 (1.2) 5.6 (1.0) Upper extremity muscular fitness Mobility Grip strength was the most commonly used upper extremity The TUG test was used to evaluate functional mobility in two function test; it was used in 26 studies (see Table 3 in the included studies (see Table 5 in the eAddenda). However, the eAddenda). The mean of the grip strength data that could be pooled results from the two are not directly comparable as they used two was 24.6 kg (95% CI 23.7 to 25.5) in women on treatment and different protocols: one used an 8-foot course and the other a 3- 22.8 kg (95% CI 20.6 to 25.1) in women off treatment (Figure 7). metre course. In both studies, reported times were slower than These values fall below the median reported values of 27.7 kg for population normative values for similar age groups (Table 2). healthy adults aged 50 to 59 (Table 2).29 No heterogeneity was identified (I2 values < 20%). 1RM using a bench or chest press protocol was estimated in four studies and measured directly in four studies. The pooled mean for bilateral bench press 1RM was 20.9 kg (95% CI 17.0 to 24.7) in women on treatment and 23.9 kg (95% CI 21.0 to 26.8) in women off treatment (Figure 8). Moderate heterogeneity was identified (I2 = 36%) for women off treatment. Normative values for 1RM are

TD$)(G]3_erugiFI[ Researc4Fi$gDTure(_[]G)Ih 197 Courneya 2007 (CG) Kolden 2002 Ligibel 2010 Courneya 2007 (RET) Schneider 2007 Brdareski 2012 (IG1)ª Courneya 2007 (AET) Brdareski 2012 (IG2)ª Tosti 2011ª Haykowsky 2009 Turner 2004ª Daley 2007 (IG) Kim 2006 (IG) Daley 2007 (ExCG) Daley 2007 (CG) Kim 2006 (CG) Evans 2009 Fillion 2008 (IG) Campbell 2012 Fillion 2008 (CG) Hsieh 2008 (Sx) Cheema 2006 Hsieh 2008 (Sx+CT) Hsieh 2008 (Sx+RT) Courneya 2003 (IG) Hsieh 2008 (Sx+CT+RT) Hughes 2008 Courneya 2003 (CG) Mustanti 2012 (CG) Mustanti 2012 (AET) Dolan 2012 (Cycle) Mustanti 2012 (RET) Mustanti 2012 (AET+RET) Dolan 2012 (Treadmill) Rogers 2009 (IG) Rogers 2009 (CG) Hokken 2009 (AC) Rogers 2013 (IG) Rogers 2013 (CG) Hokken 2009 (FC) Schneider 2007 (After) Scott 2013 (IG) Hutnick 2005 (IG) Scott 2013 (CG) Hutnick 2005 (CG) Pooled, on treatment 36 Jones 2007a Pooled, off treatment Jones 2007b 12 16 20 24 28 32 Estimated VO2max (mL/kg/min) Neil 2012 (Fatigue) mean (95% CI) Neil 2012 (CG) Figure 4. Forest plot of weighted mean (95% CI) VO2max estimated from a submaximal exercise test. References to included studies are available in Appendix Pooled, on treatment 2 in the eAddenda. AET = aerobic exercise training, CG = control group, CT = chemotherapy, Pooled, off treatment ExCG = exercise control group, IG = intervention group, RET = resistance exercise training, RT = radiotherapy, Sx = surgery. a Cycle ergometer. 1.0 1.2 1.4 1.6 1.8 2.0 2.2 VO2peak (L/min) mean reported VO2peak values were 22 to 30% lower than mean (95% CI) published norms for those aged 50 to 59. Figure 3. Forest plot of weighted mean (95% CI) VO2peak from a maximal exercise An important consideration when comparing results across test. References to included studies are available in Appendix 2 in the eAddenda. studies is the age range of the participants. While mean ages were AC = four cycles of adjuvant chemotherapy with doxorubicin/cyclophosphamide, extracted from the papers included, individual level data would be AET = aerobic exercise training, CG = control group, FC = five cycles of 5-fluorouracil/ needed in order to compare values of physical function amongst epirubicin/cyclophosphamide, IG = intervention group, RET = resistance exercise different age groups. For example, aerobic capacity has been training. shown to decline by approximately 9% per decade after the age of 50, so comparisons of mean VO2peak values across a wide range of Discussion ages may not be appropriate.30 Monitoring physical function during and after cancer treatment In the present meta-analysis, pooled values of all measures of may help physiotherapists and other health professionals to aerobic capacity and grip strength were lower for women who identify declines in physical function, and prescribe interventions were off treatment than women who were on treatment. The to mitigate these declines and improve functional outcomes. We opposite was observed for bench press and leg press 1RM values. aimed to summarise the published values in the literature to date Findings from 1RM should be interpreted with caution, due to its in order to provide clinicians with expected values in this substantial heterogeneity among women off treatment. The 1RM population for the tests of physical function most commonly data were a combination of estimated and objectively measured reported in the literature and to inform clinicians and researchers values. It is possible that the predictive equations used to estimate of testing options. A longer-term goal of the research is greater standardisation of testing in both clinical and research settings. We also aimed to compare the values that are currently being reported in women who have been diagnosed with breast cancer to normative values that have been published in healthy populations, with the goal of contextualising the physical function deficits experienced by women with breast cancer. Reported values of aerobic capacity, upper extremity strength and mobility were generally lower than reported normative values in similar age groups. This was not surprising given the various side effects of cancer treatment and fatigue leading to decline in overall physical activity. Jones and colleagues compared VO2peak between women with breast cancer at various stages of the disease and expected values for healthy sedentary women.10 Similar to the findings of the present review, VO2peak was much lower in women diagnosed with breast cancer than would be expected. Women in the Jones study who were 50 years old and diagnosed with breast cancer were on average 30% less aerobically fit, which is similar to the present review’s finding that pooled

1]GIF$DT)5_erugi([ 98 F$Di(T)G7_erug[]INeil-Sztramko et al: Physical function after breast cancer diagnosis Anderson 2012 (IG) Beurskens 2007 (IG) Beurskens 2007 (CG) Anderson 2012 (CG) Haines 2010 (IG) Haines 2010 (IG) Haines 2010 (CG) Kilgour 2008 (CG) Haines 2010 (CG) Kilgour 2008 (IG) Wang 2011 Morimoto 2003 Ahmed Omar 2011 (IG) Basen-Engquist 2006 Ahmed Omar 2011 (CG) Eyigor 2010 (IG) Dawes 2008 (Affected) Dawes 2008 (Unaffected) Eyigor 2010 (CG) Hutnick 2005 (IG, L) Kaltsatou 2011 (IG) Hutnick 2005 (IG, R) Hutnick 2005 (CG, L) Kaltsatou 2011 (CG) Hutnick 2005 (CG, R) Kaltsatou 2011 (IG, L) Yuen 2007 (CG) Kaltsatou 2011 (IG, R) Kaltsatou 2011 (CG, L) Yuen 2007 (RET) Kaltsatou 2011 (CG, R) Yuen 2007 (AET) Merchant 2008 (Affected) Merchant 2008 (Unaffected) Pooled, on treatment 475 525 575 625 675 Pooled, off treatment 6-minute walk distance (m) Nikander 2012 (IG) Nikander 2012 (CG) 375 425 mean (95% CI) O’Neill 2006 (Affected) O’Neill 2006 (Unaffected) Figure 5. Forest plot of weighted mean (95% CI) 6-minute walk distance. References Reitman 2004 (Affected) to included studies are available in Appendix 2 in the eAddenda. Reitman 2004 (Unaffected) AET = aerobic exercise training, CG = control group, IG = intervention group, Rogers 2009 (IG, R) RET = resistance exercise training. Rogers 2009 (CG, R) Rogers 2009 (IG, L) [(Figure_6)TD$IG] Rogers 2009 (CG, L) Smoot 2010 (NoLO Unaffected) Campbell 2005 (IG) 1300 Smoot 2010 (NoLO Affected) Campbell 2005 (CG) Smoot 2010 (LO Unaffected) Smoot 2010 (LO Affected) Mock 1994 (IG) Winters-Stone 2012 (IG, L) Mock 1994 (CG) Winters-Stone 2012 (IG, R) Mock 2001 (Low) Winters-Stone 2012 (CG, L) Mock 2001 (High) Winters-Stone 2012 (CG, R) Mock 2005 Pooled, on treatment Schwartz 2000a Pooled, off treatment Schwartz 2000b (IG) Schwartz 2000b (CG) 8 12 16 20 24 28 32 36 Grip strength (kg) Schwartz 2001 mean (95% CI) Schwartz 2007 (AET) Schwartz 2007 (RET) Figure 7. Forest plot of weighted mean (95% CI) grip strength. References to Schwartz 2007 (CG) included studies are available in Appendix 2 in the eAddenda. Mulero-Portela 2008 (CG) CG = control group, IG = intervention group, L = left, LO = lymphoedema, NoLO = no Mulero-Portela 2008 (Gym) lymphoedema, R = right. Mulero-Portela 2008 (Home) during chemotherapy or radiation; however, the majority were Mutrie 2007 (CG) assessed at the beginning of chemotherapy treatment. Women Mutrie 2007 (IG) classified as off treatment ranged from a few months to many years Winters-Stone 2008 after treatment. Future observational studies repeating measures of physical function before, during, and after treatment are needed Pooled, on treatment to more accurately determine the expected pattern of change in Pooled, off treatment physical function throughout the cancer trajectory. 700 800 900 1000 1100 1200 Another source of variation between studies was the specific 12-minute Walk Distance (m) testing protocol used. Submaximal and maximal exercise tests mean (95% CI) may be performed on either a cycle ergometer or a treadmill and may use a ramp or incremental protocol with a number of Figure 6. Forest plot of weighted mean (95% CI) 12-minute walk distance. possibilities in length of test stage and workload increment per References to included studies are available in Appendix 2 in the eAddenda. stage. Values for VO2peak have been shown to be higher using a AET = aerobic exercise training, CG = control group, IG = intervention group, treadmill than cycle ergometer protocol in women diagnosed with RET = resistance exercise training. breast cancer.31 Values for upper and lower extremity strength, such as grip strength, maximal contraction for leg press, or knee 1RM overestimated the true value. The timing of measurement flexion/extension, may be reported as average of three trials or also varied between studies, which should be kept in mind when maximum value obtained. There was also variation in the protocols comparing groups on and off treatment. Women classified as on used for assessing muscular endurance and the chair stand test, treatment may have completed surgery, or been at any point which prevented pooling of the results together. This highlights the importance of reporting full details of the testing protocol in order to determine whether comparisons can be made between studies. Overall, 56 (66%) studies included some measure of aerobic capacity, indicating recognition of the importance of this compo- nent of health-related physical fitness. The most common method of measurement used was the gold-standard, maximal, cardiopul- monary exercise test, followed by a submaximal exercise test

u[e_8)TDgFIG]i$(r ResearcFig]urT[D$)9(Ge_Ih 199 Courneya 2007 (CG) Kolden 2002 Courneya 2007 (RET) Schwartz 2007 (AET) Courneya 2007 (AET) Schwartz 2007 (RET) Kolden 2002 Schwartz 2007 (CG) Cheema 2006 Cheema 2006 Lane 2005 Hokken 2009 (AC) Ahmed 2006 (IG) Hokken 2009 (FC) Ahmed 2006 (CG) Ahmed 2006 (IG) Schmitz 2010 (IG) Ahmed 2006 (CG) Schmitz 2010 (CG) Schmitz 2010 (IG) Winters-Stone 2012 (IG) Schmitz 2010 (CG) Winters-Stone 2012 (CG) Winters-Stone 2011 (Fall) Pooled, on treatment Winters-Stone 2011 (No fall) Pooled, off treatment Winters-Stone 2012 (IG) 10 20 30 40 50 60 70 Winters-Stone 2012 (CG) Bench press 1RM (kg) mean (95% CI) Pooled, on treatment Figure 8. Forest plot of weighted mean (95% CI) one-repetition maximum (1RM) for Pooled, off treatment bench press. References to included studies are available in Appendix 2 in the eAddenda. 40 60 80 100 120 140 160 AET = aerobic exercise training, CG = control group, IG = intervention group, Leg press 1RM (kg) RET = resistance exercise training. mean (95% CI) terminated at a specified percentage of age-predicted heart rate Figure 9. Forest plot of weighted mean (95% CI) one-repetition maximum (1RM) for reserve or maximal heart rate. Although formal, large-scale leg press. References to included studies are available in Appendix 2 in the assessment of the safety of the cardiopulmonary exercise testing eAddenda. procedure in individuals with cancer has not been performed, it AC = four cycles of adjuvant chemotherapy with doxorubicin/cyclophosphamide, does appear to be relatively safe with appropriate screening and AET = aerobic exercise training, CG = control group, FC = five cycles of 5-fluorouracil/ monitoring during the test.32 Submaximal exercise testing is epirubicin/cyclophosphamide, IG = intervention group, RET = resistance exercise considered to be a safer option, and may not require medical training. supervision, but is not as accurate for quantifying VO2peak.11 Finally, walking tests (6MWT and 12MWT) were commonly One limitation of this review is the likely presence of selection reported. Research is needed to determine if the 12MWT is a more bias in the individuals included in the research studies, limiting the appropriate test for capturing physical function in women with generalisability of these results to all women diagnosed with breast cancer than the 6MWT. It may be that women diagnosed breast cancer. Due to the nature of the outcome measures of with breast cancer have greater physical capacity than individuals interest in this review, many of the studies included were physical in cardiac and pulmonary rehabilitation where the 6MWT is activity interventions. While some studies did restrict eligibility to commonly used, and therefore may experience a ceiling effect with women who were sedentary or not currently exercising routinely, the 6MWT.12 due to the nature of the intervention, these studies likely recruited a select group who were the most healthy or health-conscious. Grip strength was the most commonly used measure of Other studies specifically limited their study populations to strength in this review and has been recommended as an women who experienced functional limitations36–40 or women assessment of muscle function for oncology rehabilitation.33 Grip with lymphoedema.8,41 In these cases, values below those reported strength dynamometry is an attractive measure of strength in all for the average woman diagnosed with breast cancer can be populations due to its ease of use, reliability, generalisability to expected. Other studies excluded women with functional pro- overall strength, and availability of published age and gender blems that may be worsened by exercise, such as shoulder pain. norms.22 Additionally, grip strength is reported to be a significant Therefore, we decided to include all relevant papers with the predictor of health-related quality of life in breast cancer caveat that results from individual studies reported may be more survivors.34 While 1RM testing may be more sensitive and specific relevant to different subgroups of women diagnosed with breast for strength training interventions, the small number of studies cancer, and the pooled meta-analysis may not be applicable to all performing 1RM testing for upper body testing could be attributed women. As more research becomes available, future work should to fear of musculoskeletal injury in a population likely to be na¨ıve aim to analyse physical function in these groups of women to strength training, and concern regarding risk of precipitating separately. lymphoedema. However, guidelines from the American College of Sports Medicine published in 2010 advocate that 1RM testing is One strength of this review is the inclusion of objective gold- safe in women with breast cancer, even those with or at higher risk standard tests of physical function, such as measured VO2peak and for lymphoedema.35 1RM testing for muscular strength. While these tests may provide the best assessment of physical function, they require the use of Only two studies included measurements of mobility. This may specialised, often expensive equipment and individuals who have be because the TUG test and other mobility tests have been been trained to conduct the testing. Therefore, submaximal and developed for and validated in older adults,25 and thus may not be field tests to estimate maximal values are invaluable in clinical sufficiently sensitive to capture impairment experienced following practice, and may also be quite useful in some research settings. A breast cancer treatment. An alternative explanation is that second strength is the meta-analysis used to combine data from mobility impairments following breast cancer and its treatment multiple studies, which provides a general estimate of expected have not been widely recognised in the literature, and as a values in this population. result few studies have measured this. Thus the utility of mobility testing in this population requires further investigation. This review summarises the values that have been reported in the literature to date for various components of physical function,

200 Neil-Sztramko et al: Physical function after breast cancer diagnosis namely aerobic capacity, upper and lower extremity strength and 11. Pescatello LS, Arena R, Riebe D, Thompson PD, eds. In: ACSM’s Guidelines for Exercise mobility in women diagnosed with breast cancer. Values for Testing and Prescription 9 ed. Philadelphia: Wolters Kluwer Lippincott Williams & aerobic capacity and upper extremity strength are generally lower Wilkins; 2014. than published normative values in similar age groups. Lower extremity strength does not appear to follow this pattern, with 12. ATS Committee on Proficiency Standards for Clinical Pulmonary Function Labora- values higher than population norms. This review also highlights tories. ATS statement: guidelines for the six-minute walk test. Am J Respir Crit Care the variety of tests used in the literature to assess physical function Med 2002; 166:111-117. and the variations in testing protocols that may potentially contribute to the heterogeneity in values reported. Objective 13. Ross RM, Murthy JN, Wollak ID, Jackson AS. The six minute walk test accurately assessments of various aspects of physical function are important estimates mean peak oxygen uptake. BMC Pulm Med. 2010;10:31. for documenting deficits in physical function and reporting change in response to specific interventions and monitoring individual 14. Schmidt K, Vogt L, Thiel C, Jager E, Banzer W. Validity of the six-minute walk test in progress in physiotherapy practice and research settings. As more cancer patients. Int J Sports Med. 2013;34:631–636. research becomes available, expected values for sub-populations of different ages, stages of treatment and with various co- 15. Kline GM, Porcari JP, Hintermeister R, Freedson PS, Ward A, McCarron RF, et al. morbidities will be useful for both researchers and clinicians Estimation of VO2max from a one-mile track walk, gender, age, and body weight. working with women after a breast cancer diagnosis. Med Sci Sports Exerc. 1987;19:253–259. What is already known on this topic: Breast cancer and its 16. Oja P, Laukkanen R, Pasanen M, Tyry T, Vuori I. A 2-km walking test for assessing treatment can cause impairment in physical function in wom- the cardiorespiratory fitness of healthy adults. Int J Sports Med. 1991;12:356–362. en. What this study adds: Compared to normative data, women 17. Hayes S, Battistutta D, Newman B. Objective and subjective upper body function during and after treatment for breast cancer had reduced six months following diagnosis of breast cancer. Breast Cancer Res Treat. aerobic fitness. Upper and lower extremity strength was also 2005;94:1–10. reduced for women who were currently receiving cancer treatment. Lower extremity strength was above population 18. Brown JC, Mao JJ, Stricker C, Hwang WT, Tan KS, Schmitz KH. Aromatase inhibitor norms for women who had completed treatment. associated musculoskeletal symptoms are associated with reduced physical activ- ity among breast cancer survivors. Breast J. 2014;20:22–28. eAddenda: Tables 3, 4, 5 and 6, and Appendix 1 and 2 can be found online at doi:10.1016/j.jphys.2014.09.005 19. Gaines JM, Talbot LA. Isokinetic strength testing in research and practice. Biol Res Nurs. 1999;1:57–64. Ethics approval: N/A Competing interests: Nil. 20. Bohannon RW. Intertester reliability of hand-held dynamometry: a concise sum- Source(s) of support: SENS and AAK are supported by doctoral mary of published research. Percept Mot Skills. 1999;88:899–902. student awards from the Canadian Institute for Health Research. Acknowledgements: We wish to acknowledge Jonathan Chu, 21. Jones CJ, Rikli RE, Beam WC. A 30-s chair-stand test as a measure of lower body Jackson Lam, Kenneth Lo, and Vincent Sy, members of the 2012 MPT strength in community-residing older adults. Res Q Exerc Sport. 1999;70:113–119. class at the University of British Columbia for their work on developing the search strategy for an earlier version of this review. 22. Bohannon RW. Dynamometer measurements of hand-grip strength predict mul- Correspondence: Kristin L Campbell, Department of Physical tiple outcomes. Percept Mot Skills. 2001;93:323–328. Therapy, University of British Columbia, Vancouver, Canada. Email: [email protected] 23. Wall JC, Bell C, Campbell S, Davis J. The Timed Get-up-and-Go test revisited: measurement of the component tasks. J Rehabil Res Dev. 2000;37:109–113. References 24. Podsiadlo D, Richardson S. The timed ‘‘Up & Go’’: a test of basic functional mobility 1. IARC. GLOBOCAN 2008: Cancer Incidence and Mortality Worldwide. Lyon, France: for frail elderly persons. J Am Geriatr Soc. 1991;39:142–148. International Agency for Research on Cancer; 2010. 25. Steffen TM, Hacker TA, Mollinger L. Age- and gender-related test performance in 2. Canadian Cancer Society. Canadian Cancer Statistics. Toronto, ON: Canadian Cancer community-dwelling elderly people: Six-Minute Walk Test, Berg Balance Scale, Society; 2013. Timed Up & Go Test, and gait speeds. Phys Ther. 2002;82:128–137. 3. Campbell KL, Pusic AL, Zucker DS, McNeely ML, Binkley JM, Cheville AL, et al. A 26. Nishikawa K, Lo K, Lam J, Sy V, Chu J. Values of Physical Function in Breast Cancer prospective model of care for breast cancer rehabilitation: function. Cancer. Survivors: A systematic review [Systematic Review]. Vancouver: Department of 2012;118:2300–2311. Physical Therapy, University of British Columbia; 2012. 4. Courneya KS, Segal RJ, Mackey JR, Gelmon K, Reid RD, Friedenreich CM, et al. Effects 27. Neyeloff JL, Fuchs SC, Moreira LB. Meta-analyses and Forest plots using a microsoft of aerobic and resistance exercise in breast cancer patients receiving adjuvant excel spreadsheet: step-by-step guide focusing on descriptive data analysis. BMC chemotherapy: a multicenter randomized controlled trial. J Clin Oncol. Res Notes. 2012;5:52. 2007;25:4396–4404. 28. Rikli RE, Jones CJ. Functional fitness normative scores for community-residing 5. Jones LW, Eves ND, Haykowsky M, Freedland SJ, Mackey JR. Exercise intolerance in older adults, ages 60-94. J Aging Phys Activity. 1999;7:162–181. cancer and the role of exercise therapy to reverse dysfunction. Lancet Oncol. 2009;10:598–605. 29. Peters MJ, van Nes SI, Vanhoutte EK, Bakkers M, van Doorn PA, Merkies IS, et al. Revised normative values for grip strength with the Jamar dynamometer. J Peripher 6. Myers J, Prakash M, Froelicher V, Do D, Partington S, Atwood JE. Exercise capacity Nerv Syst. 2011;16:47–50. and mortality among men referred for exercise testing. N Engl J Med. 2002;346:793–801. 30. Tanaka H, Desouza CA, Jones PP, Stevenson ET, Davy KP, Seals DR. Greater rate of decline in maximal aerobic capacity with age in physically active vs. sedentary 7. Keteyian SJ, Brawner CA, Savage PD, Ehrman JK, Schairer J, Divine G, et al. Peak healthy women. J Appl Physiol. 1997;83:1947–1953. aerobic capacity predicts prognosis in patients with coronary heart disease. Am Heart J. 2008;156:292–300. 31. Dolan LB, Lane K, McKenzie DC. Optimal mode for maximal aerobic exercise testing in breast cancer survivors. Integr Cancer Ther. 2012;11:321–326. 8. O’Neill JO, Young JB, Pothier CE, Lauer MS. Peak oxygen consumption as a predictor of death in patients with heart failure receiving beta-blockers. Circulation. 32. Jones LW, Eves ND, Haykowsky M, Joy AA, Douglas PS. Cardiorespiratory exercise 2005;111:2313–2318. testing in clinical oncology research: systematic review and practice recommen- dations. Lancet Oncol. 2008;9:757–765. 9. Peel JB, Sui X, Adams SA, Hebert JR, Hardin JW, Blair SN. A prospective study of cardiorespiratory fitness and breast cancer mortality. Med Sci Sports Exerc. 33. Gilchrist LS, Galantino ML, Wampler M, Marchese VG, Morris GS, Ness KK. A 2009;41:742–748. framework for assessment in oncology rehabilitation. Phys Ther. 2009;89:286–306. 10. Jones LW, Courneya KS, Mackey JR, Muss HB, Pituskin EN, Scott JM, et al. Cardio- 34. Rietman JS, Dijkstra PU, Debreczeni R, Geertzen JH, Robinson DP, De Vries J. pulmonary function and age-related decline across the breast cancer survivorship Impairments, disabilities and health related quality of life after treatment for continuum. J Clin Oncol. 2012;30:2530–2537. breast cancer: a follow-up study 2.7 years after surgery. Disabil Rehabil. 2004;26: 78–84. 35. Schmitz KH, Courneya KS, Matthews C, Demark-Wahnefried W, Galvao DA, Pinto BM, et al. American College of Sports Medicine roundtable on exercise guidelines for cancer survivors. Med Sci Sports Exerc. 2010;42:1409–1426. 36. Beurskens CH, van Uden CJ, Strobbe LJ, Oostendorp RA, Wobbes T. The efficacy of physiotherapy upon shoulder function following axillary dissection in breast cancer, a randomized controlled study. BMC Cancer. 2007;7:166. 37. Cantarero-Villanueva I, Fernandez-Lao C, Del Moral-Avila R, Fernandez-de-Las- Penas C, Feriche-Fernandez-Castanys MB, Arroyo-Morales M. Effectiveness of core stability exercises and recovery myofascial release massage on fatigue in breast cancer survivors: a randomized controlled clinical trial. Evid Based Complement Alternat Med. 2012;2012:620619. 38. Cantarero-Villanueva I, Fernandez-Lao C, Diaz-Rodriguez L, Fernandez-de-Las- Penas C, Ruiz JR, Arroyo-Morales M. The handgrip strength test as a measure of function in breast cancer survivors: relationship to cancer-related symptoms and physical and physiologic parameters. Am J Phys Med Rehabil. 2012;91:774–782. 39. Haykowsky MJ, Mackey JR, Thompson RB, Jones LW, Paterson DI. Adjuvant trastuzumab induces ventricular remodeling despite aerobic exercise training. Clin Cancer Res. 2009;15:4963–4967. 40. Yuen HK, Sword D. Home-based exercise to alleviate fatigue and improve func- tional capacity among breast cancer survivors. J Allied Health. 2007;36:e257–e275. 41. Ahmed Omar MT, Abd-El-Gayed Ebid A, El Morsy AM. Treatment of post-mastec- tomy lymphedema with laser therapy: double blind placebo control randomized study. J Surg Res. 2011;165:82–90. 42. Andrews AW, Thomas MW, Bohannon RW. Normative values for isometric muscle force measurements obtained with hand-held dynamometers. Phys Ther. 1996;76:248–259. 43. Isles RC, Choy NL, Steer M, Nitz JC. Normal values of balance tests in women aged 20-80. J Am Geriatr Soc. 2004;52:1367–1372.

Journal of Physiotherapy 60 (2014) 237 Journal of PHYSIOTHERAPY journal homepage: www.elsevier.com/locate/jphys Appraisal Critically Appraised Papers A program of neck exercises can prevent neck pain in office workers Synopsis Summary of: Sihawong R, Janwantanakul P, Jiamjararangsi diary. The usual care control group did not receive any prescribed W. Effects of an exercise programme on preventing neck pain exercise. Outcome measures: The primary outcome was pain among office workers: a 12-month cluster-randomized controlled greater than 30 mm on a 100 mm visual analogue scale during the trial. Occup Environ Med 2014;71:63–70. 12-month intervention period. Secondary outcome measures were neck flexor range and neck flexor endurance on a randomly Question: Does an exercise program that is focused on selected sub-group of participants (n = 40). Results: A total of stretching and endurance training reduce the 12-month incidence 534 participants (94%) completed the study. Over the 12 months, of neck pain in office workers? Design: Cluster randomised 32 of 267 (12%) participants in the intervention group, and 72 of controlled trial with concealed allocation of sites. Setting: Twelve 270 (27%) participants in the control group reported neck pain. offices in Thailand. Participants: People aged 18 to 55 years, There was a reduced risk of neck pain in the intervention group of working full time, with reduced neck flexion range and neck about half (Hazard rate ratio = 0.45, 95% CI 0.28 to 0.71). Exercise muscle endurance were included. Key exclusion criteria were: adherence (percentage of prescribed sessions completed) for the symptoms in the spine in the previous 6 months, pregnancy, a intervention group was 30 to 34% for stretching, and 57% for history of trauma to the spine in the previous 12 months, and endurance exercise. The sub-groups did not significantly differ for performing regular exercise. Randomisation of 567 participants change in neck range or neck flexor endurance, as interaction allocated 285 to the intervention group and 282 to the control effects were non-significant. Conclusion: A 12-month program of group. Interventions: The intervention group was prescribed an neck exercises prescribed for office workers with reduced range exercise program of stretching and endurance exercises. The and poor muscle endurance led to a reduced incidence of neck pain. exercise program consisted of stretching exercises for four neck muscles, each held for 30 seconds, twice each working day. Nicholas Taylor Endurance training consisted of 10 repetitions of sustained holds of Section Editor, Journal of Physiotherapy neck flexor muscles, twice each week for 12 months. Exercise participants received a reminder via mobile phone every working http://dx.doi.org/10.1016/j.jphys.2014.08.003 day for the first 3 months, and recorded exercise completion in a Commentary that neck pain experienced by participants did not have much impact on daily function. Alternatively, perhaps the NDI did not Finding ways to prevent neck pain is important because a capture significant upper limb-associated activity limitations that significant proportion of office workers who have neck pain report are often related to neck pain.3,4 Future research should measure associated activity limitations and often experience recurrent additional outcomes related to activity limitation, productivity, symptoms.1 Although neck pain most likely develops because of medical usage, etc, so that workers, clinicians, and other stake- exposure to multiple physical and psychosocial factors over time, it holders can make well-informed decisions about implementing has been suggested that new episodes of neck pain might be exercise-related strategies that may help to prevent neck pain. prevented by improving the physical capabilities of the neck and shoulder muscles.1 However, previous clinical trials on the effective- Robert J Nee ness of this prevention strategy have shown conflicting results.2 School of Physical Therapy, Pacific University, Hillsboro, USA Judging the potential value of a preventive intervention requires References weighing the resources required to implement the intervention against the relative severity of the event to be prevented. Therefore, 1. Cote P, et al. Spine. 2008;33(4S):S60–S74. the definition of neck pain (pain > 30 mm on a 100-mm visual 2. Sihawong R, et al. J Manipulative Physiol Ther. 2011;34:62–71. analogue scale that lasted > 24 hours) should be considered when 3. Mehta S, et al. Spine. 2010;35:2150–2156. interpreting the results from Sihawong et al because it does not 4. Olson W, et al. Man Ther. 2013;18:492–497. communicate the level of activity limitation associated with the pain experience. Regardless of group assignment, participants who http://dx.doi.org/10.1016/j.jphys.2014.08.015 reported neck pain had an average Neck Disability Index (NDI) score of only 7 to 8 points (0 to 100 scale). These low scores suggest 1836-9553/ß 2014 Australian Physiotherapy Association. Published by Elsevier B.V. All rights reserved.

Journal of Physiotherapy 60 (2014) 179–180 Journal of PHYSIOTHERAPY journal homepage: www.elsevier.com/locate/jphys Editorial A seven-day physiotherapy service Nicholas F Taylor a,b, Nora Shields a,c a Department of Physiotherapy, La Trobe University; b Allied Health Clinical Research Office, Eastern Health; c Department of Allied Health, Northern Health, Melbourne, Australia Hospitals and primary healthcare services operate around the better quicker, with benefits in functional independence and clock, 7 days a week. Traditionally, physiotherapy services have health-related quality of life sustained at 6 months after operated within business hours from Monday to Friday or, if an discharge.7 A recent study with comparison to a historical control out-of-hours service has been provided, it has been a reduced also found that implementing a multidisciplinary rehabilitation service. However, the health problems of some of our patients can service on a Saturday in Australia improved functional indepen- deteriorate if not addressed immediately. In addition, many people dence.8 A retrospective study in the United States found that a with less urgent problems may find it difficult to attend 7-day rehabilitation service including physiotherapy reduced length physiotherapy appointments during business hours due to their of stay by 1 day, compared to a 5-day service.9 Studies have also own commitments or work. Consistent with the principles of reported a reduction in pulmonary complications for patients with patient-centred and family-centred care,1 we have an obligation to acute spinal injury,10 and the elderly after surgery,11 in an intensive provide care for people when they need it and when they are care unit with additional out-of-hours physiotherapy. available. This situation, together with the fact that other services and professions in the healthcare system provide care 7 days a In other areas of practice, however, the evidence for out-of- week, provides a rationale for a discussion on providing a 7-day hours physiotherapy services is, to date, less convincing. A physiotherapy service. retrospective study found that introducing a 7-day service after lower-limb joint replacement in an Australian regional hospital did Some areas of physiotherapy practice have a long tradition of not decrease hospital length of stay.12 A review found that there providing services outside of business hours, for example, was no strong evidence to support the provision of out-of-hours weekend physiotherapy services for patients with high acuity physiotherapy services in the areas of orthopaedics, neurology and such as in intensive care units. In the United Kingdom, 97% of cardiac surgery.13 One of the difficulties in evaluating the evidence intensive care units provide 24-hour access to physiotherapy,2 and is that so few studies in this area have been randomised controlled in Canada, 97% of intensive care units have weekend physiotherapy trials. The lack of controlled trials is a problem because apart from services.3 A recent Australian survey found that 80% of acute there being an increased risk of bias in the results, other factors wards provided physiotherapy on a Saturday.4 Also, physiothera- that could influence outcomes, such as the amount of physiother- pists working in private practice, often with a focus on treating apy, may not be controlled or accounted for. musculoskeletal problems, have long provided, at least in Australia, services outside of business hours including weekends. A key issue in evaluating the effectiveness of out-of-hours Although we were not able to locate data about the extent of the physiotherapy services is determining whether the services out-of-hours services provided by private practitioners, informa- provided are additional services, or whether they are redistrib- tion about the number of hours worked by physiotherapists in uted from existing Monday-to-Friday services.3 There is strong excess of 40 hours a week suggests that these services may be evidence that providing additional physiotherapy across a range widespread.5 of health conditions and across acute hospital and rehabilitation settings can improve patient outcomes and reduce length of In other areas of physiotherapy practice, out-of-hours services stay.14 Out-of-hours services are one way of increasing the are either much reduced or absent. For example, only 30% of amount of physiotherapy provided to patients. In the context of rehabilitation services in Australia,4 and approximately 69% of providing additional physiotherapy services, it has also been community hospitals in Canada,6 provide physiotherapy services reported that rehabilitation inpatients had a different attitude to at weekends. Although 97% of tertiary care hospitals in Canada treatment when services were provided at the weekend; they provide physiotherapy services at weekends, the service is 88% less considered that they were there to work, whereas the attitude of than during the week, suggesting that only a skeleton staff is patients receiving a 5-day service was that rest was more employed to address the most urgent cases.3 Furthermore, in some important at the weekend.15 Perhaps the key benefit of an out- centres, night rosters are covered by the most junior staff, who of-hours physiotherapy service is that it provides an opportunity have the least experience at dealing with unexpected or complex to increase the intensity of therapy provided.7 This benefit may changes in a patient’s clinical condition. not manifest if the overall amount of physiotherapy is not increased by the redistribution of a 5-day service over 7 days. The case for advocating increased out-of-hours physiotherapy services would be more compelling if its provision was supported As a member of a multidisciplinary team, it may be a problem if by evidence. Such evidence is starting to emerge. A randomised the physiotherapist is providing out-of-hours service, but the other controlled trial from Australia, for example, found that the members of the team are not. For example, in a retrospective study provision of additional Saturday physiotherapy and occupational where only the physiotherapy service was increased at the therapy helped adults receiving inpatient rehabilitation to get weekend, the physiotherapy length of stay decreased but the http://dx.doi.org/10.1016/j.jphys.2014.08.004 1836-9553/ß 2014 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/3.0/).

180 Editorial hospital length of stay did not.14 The main issue identified for this service. It may also help to structure work schedules to include a day discrepancy was that other parts of the health service were not off at the weekend, which can be important in helping health ready for patient discharge. Consistent with this, other allied professionals to recover from work stress.22 health professions such as social work and occupational therapy, which are essential to patient management and discharge In conclusion, a 7-day physiotherapy service in some form and planning, typically have much lower weekend coverage than in some areas has long been a part of practice. There is now physiotherapy.6 This issue of recognising that one area of the emerging evidence that providing additional out-of-hours phys- health service cannot function effectively at the weekend without iotherapy services (including at the weekends) can help to improve having access to other areas of the health service has been more patient outcomes and be cost-effective. As health professionals broadly recognised in a discussion about providing a 7-day service providing an important service in the health system, it seems that in the National Health Service in the United Kingdom.16 physiotherapists should be working when other members of the healthcare team are working and at a time that provides care when Another issue is whether the efficacy of a particular physio- patients need it. The challenge is to provide evidence in areas of therapy intervention has been established with 5-day or 7-day practice where evidence remains scant, and to change the culture input. For example, all four trials of inspiratory muscle training to and embed the notion that providing additional physiotherapy facilitate weaning from artificial ventilation in the intensive care through a 7-day service can be a routine, beneficial and desirable unit have provided the physiotherapist-administered training on a part of practice. 7-day basis.17,18 If physiotherapists only administer this on a 5-day basis, the gains in respiratory muscle endurance may be reduced or Ethics approval: N/A lost during the 60 hours of continuous mechanical ventilation from Competing interests: N/A the last Friday treatment to the first Monday treatment. Source(s) of support: N/A Acknowledgements: N/A Even if providing additional out-of-hours physiotherapy Correspondence: Nicholas F Taylor, Department of Physiother- services is effective, the issue of who pays remains.19 Are apy, La Trobe University, Australia. Email: [email protected] additional physiotherapy services worth the cost? Several studies have investigated the cost-effectiveness of providing additional References physiotherapy at weekends. A review of the health economics of providing rehabilitation concluded that it was cost-effective to 1. Australian Physiotherapy Association. Standard for Physiotherapy Practices 8th provide additional rehabilitation therapy for people with stroke or edition 2014. Viewed 2 Sept 2014, from https://www.physiotherapy.asn.au/ orthopaedic diagnoses.20 Recently, a health economic analysis DocumentsFolder/Resources_Private_Practice_Standards_for_physiotherapy_ alongside a randomised controlled trial found that there were practices_2011.pdf. likely cost savings in providing additional Saturday rehabilitation to a mixed cohort of inpatients.21 Primarily through a reduction in 2. Jones AYM, et al. Physiother Theory Prac. 1992;8:39–47. length of stay, costs to the health service were reduced, even 3. Campbell L, et al. Physiother Can. 2010;62:347–354. though there was the added expense of employing physiothera- 4. Shaw KD, et al. Physiother Res Int. 2013;18:115–123. pists and occupational therapists at the weekends. One of the 5. New South Wales Health: Physiotherapy Labour Force Profile NSW – 2008. Viewed challenges is that the part of the health system that accrues the savings may not be the same part that provides the immediate 2 Sept 2014, from http://www0.health.nsw.gov.au/pubs/2010/pdf/physiotherapy_ budget for staffing the additional services. 08.pdf. 6. Ottensmeyer CA, et al. Physiother Can. 2012;64:178–187. A barrier to providing a 7-day physiotherapy service may be the 7. Peiris CL, et al. BMC Med. 2013;11:198. attitudes of physiotherapists and the perceived stress of working out 8. Hakkenes S, et al. Disabil Rehabil. 2014;early online:1–7. of regular hours. Physiotherapists who are used to working Monday 9. DiSotto-Monastero M, et al. Arch Phys Med Rehabil. 2012;93:2165–2169. to Friday may be less willing to work at weekends or in the evenings. 10. Berney S, et al. Physiother Res Int. 2002;7:14–22. However, it was found in our trial that there was no difficulty in 11. Ntoumenopoulos G, et al. Aust J Physiother. 1996;42:279–303. staffing a Saturday rehabilitation service.7,20 Part of the issue may be 12. Maidment ZL, et al. Aust Hlth Rev. 2014;38:265–270. in expectations established during training. Including out-of-hours 13. Brusco NK, et al. Physiother Theory Prac. 2006;22:291–307. clinical placements during training, similar to nurses and doctors, 14. Peiris CL, et al. Arch Phys Med Rehabil. 2011;92:1490–1500. may lead to positive attitudes and acceptance of working in a 7-day 15. Peiris CL, et al. J Physiother. 2012;58:261–268. 16. Keogh B. BMJ. 2013;346:f621. 17. Moodie L, et al. J Physiother. 2011;57:213–221. 18. Condessa RL, et al. J Physiother. 2013;59:101–107. 19. Flynn P. BMJ. 2013;346:f622. 20. Brusco NK, et al. Arch Phys Med Rehabil. 2014;95:94–116. 21. Brusco NK, et al. BMC Med. 2014;12:89. 22. Drach-Zahavy A, et al. J Adv Nurs. 2013;69:578–589.

Journal of Physiotherapy 60 (2014) 233 Journal of PHYSIOTHERAPY journal homepage: www.elsevier.com/locate/jphys Appraisal Trial Protocol A workplace exercise versus health promotion intervention to prevent and reduce the economic and personal burden of non-specific neck pain in office personnel: protocol of a cluster-randomised controlled trial V Johnston a, S O’Leary b,c, T Comans d,e, L Straker f, M Melloh g,h, A Khan a, G Sjøgaard i a School of Health and Rehabilitation Sciences, The University of Queensland, Brisbane; b Department of Physiotherapy, Royal Brisbane and Women’s Hospital, Queensland Health, Brisbane; c NHMRC Centre for Clinical Research Excellence in Spinal Pain, Injury and Health, The University of Queensland, Brisbane; d Centre for Applied Health Economics, School of Medicine, Griffith University, Brisbane; e Population and Social Health Research Program, Griffith Health Institute, Logan; f School of Physiotherapy and Exercise Science, Curtin University, Perth; g Centre for Medical Research, The University of Western Australia, Australia; h Centre for Health Sciences, Zurich University of Applied Sciences, Winterthur, Switzerland; i Institute of Sport Sciences and Clinical Biomechanics, Faculty of Health Sciences, University of Southern Denmark, Odense, Denmark Abstract Introduction: Non-specific neck pain is a major burden to office personnel will be randomly allocated to either an intervention industry, yet the impact of introducing a workplace ergonomics or control arm in work group clusters. Analysis: Analysis will be on and exercise intervention on work productivity and severity of an ‘intent-to-treat’ basis and per protocol. Multilevel, generalised neck pain in a population of office personnel is unknown. Research linear models will be used to examine the effect of the intervention question: Does a combined workplace-based best practice on reducing the productivity loss in dollar units (AUD), and severity ergonomic and neck exercise program reduce productivity losses of neck pain and disability. Discussion: The findings of this study and risk of developing neck pain in asymptomatic workers, or will have a direct impact on policies that underpin the prevention decrease severity of neck pain in symptomatic workers, compared and management of neck pain in office personnel. to a best practice ergonomic and general health promotion program? Design: Prospective cluster randomised controlled trial. Trial registration: Australian New Zealand Clinical Trials Regis- Participants and setting: Office personnel aged over 18 years, try (ANZCTR). Registration number: ACTRN12612001154897. Was and who work > 30 hours/week. Intervention: Individualised this trial prospectively registered: Yes. Funded by: National best practice ergonomic intervention plus 3 Â 20 minute weekly, Health and Medical Research Council Project Grant. Funder progressive neck/shoulder girdle exercise group sessions for approval number: APP1042508. Anticipated completion: 2016. 12 weeks. Control: Individualised best practice ergonomic Correspondence: Dr Venerina Johnston, School of Health and Reha- intervention plus 1-hour weekly health information sessions bilitation Sciences, The University of Queensland, St Lucia, Australia. for 12 weeks. Measurements: Primary (productivity loss) and Email: [email protected] secondary (neck pain and disability, muscle performance, and quality of life) outcome measures will be collected using Full protocol: Available on the eAddenda at doi:10.1016/j.jphys. validated scales at baseline, immediate post-intervention and 2014.08.007 12 months after commencement. Procedure: 640 volunteering http://dx.doi.org/10.1016/j.jphys.2014.08.007 Commentary Recent data highlights the rising problem of musculoskeletal exercise programs or health information sessions, this comparison is conditions with neck pain now the fourth most burdensome particularly relevant. condition globally in terms of years lived with disability.1 With the combination of more people in sedentary office work and A key strength of this protocol is the plan to both assess health intensification of work practices, it is critical to effectively address outcomes for individuals and to conduct a cost-benefit analysis for musculoskeletal conditions in the workplace. productivity. For health interventions to be meaningfully translated into policy and practice, they must be economically viable. Given the The protocol by Johnston and colleagues can be seen as a intensive nature of the combined ergonomic and exercise interven- significant development on previous research. They propose to tion to be studied, weighing the cost of intervention against the investigate a combination of best practice ergonomic interventions productivity of employees will be pivotal in deciding whether such with exercise for the upper quadrant, compared with a combina- interventions should be recommended in the workplace. tion of ergonomic intervention and health promotion sessions. Niamh Moloney The design of the study is impressive. The exercise programs are Discipline of Physiotherapy, University of Sydney, Australia individualised and progressive, following best practice guidelines in exercise prescription.2 Importantly, the comparison group in this References study will receive an intervention equal in time to the exercise- training group. The comparison group, receiving health promotion 1. Vos T, et al. Lancet. 2012;380:2163–2196. information sessions, may gain health and productivity benefits 2. Kraemer WJ, et al. Med Sci Sports Exerc. 2002;34:364–380. from these sessions (eg, on stress management and keeping active; 3. Hush JM, et al. Eur Spine J. 2009;18:1532–1540. both of which have been identified as influential in neck pain).3 In deciding whether to opt for expensive and intensive supervised http://dx.doi.org/10.1016/j.jphys.2014.09.002 1836-9553/ß 2014 Australian Physiotherapy Association. Published by Elsevier B.V. All rights reserved.

Journal of Physiotherapy 60 (2014) 241–244 Journal of PHYSIOTHERAPY journal homepage: www.elsevier.com/locate/jphys Appraisal Research Note Cohort studies of aetiology and prognosis: they’re different Introduction and Robbins6,7 provide a lucid presentation of contemporary statistical methods for studying causation, including the use of Cohort designs are widely used in epidemiological research. The causal modelling in longitudinal studies. A group of leading key features of cohort studies are that: (a) a sample of participants methodologists recently published a series of articles that provide at risk of a particular outcome (the ‘cohort’) is identified; (b) data an overview of the design and analysis of prognostic studies.8–11 on individual cohort members’ exposures to certain risk factors Steyerberg11 has written a very accessible textbook on statistical and their subsequent outcomes are obtained; and (c) associations methods for prognostic research. between exposures and outcomes are quantified. Differences between cohort studies of aetiology and prognosis Here are two examples of cohort studies. Paul and colleagues1 conducted a range of physical and Objectives cognitive tests, including balance tests and the Mini-Mental State Examination, on 205 community-dwelling people who had A pre-requisite for investigating either causation or prediction Parkinson’s disease. Subsequently, the participants used falls is to identify associations between exposures and outcomes. diaries to document any falls that occurred over the next 6 months. However, the researcher who is concerned with aetiology is The analysis examined associations between physical and cogni- ultimately interested in identifying associations that are causal, tive risk factors and the incidence rate of falls. whereas the researcher who is concerned with prediction does not Moseley and colleagues2 studied 1549 patients presenting to need to differentiate between causal and non-causal associations; hospital with an acute wrist fracture. A number of clinical variables in the latter case, any association, causal or non-causal, can fulfill such as pain and swelling were assessed in the first week after the role of predicting outcomes. fracture. Four months later, each participant was contacted by telephone and, where necessary, undertook a clinical examination The study by Paul and colleagues, described above, sought to to determine whether he or she had developed complex regional determine factors that caused people with Parkinson’s disease to pain syndrome (CRPS). The analysis examined associations fall. The task for the investigators was to establish whether between the clinical variables and the development of CRPS. putative aetiologic factors (impaired physical capacity or impaired Both of these studies were cohort studies, but they had quite cognition) really did cause falls. The measure of the success of that different aims. The Paul study was designed to examine factors study is the extent to which it was able to distinguish between that cause a health outcome (in this case, falls). That is, the aim of putative aetiologic factors that truly do and do not cause falls. In the first study was to understand the mechanisms or aetiology of contrast, the study by Moseley and colleagues was explicitly not falls in people with Parkinson’s disease. In contrast, the Moseley concerned with determining causes of CRPS; it was designed, study was designed to identify factors that predict a health instead, to identify prognostic markers. Whether those markers are outcome (the development of CRPS after wrist fracture). Its aim causal or not was of no relevance, because the goal was to predict, was to generate prognoses for people who have had a wrist not to understand aetiology. The primary measure of the success fracture. of the Moseley study is the extent to which it was able to accurately In general, two types of cohort studies can be distinguished: predict who would go on to develop CRPS. aetiologic studies, which are concerned with understanding the mechanisms that cause health outcomes, and prognostic studies, Some readers might think that the distinction between cohort which are concerned with prediction of health outcomes. studies of aetiology and prognosis is simply that studies of Even though the same research design—a cohort study—can be aetiology monitor healthy participants in order to determine who used to answer questions about both aetiology and prognosis, develops the disease of interest, whereas studies of prognosis the way in which cohort studies of aetiology and prognosis are monitor people who already have a health condition in order to designed and analysed should be very different. Surprisingly, determine who develops particular disease-related outcomes.13 there have been few explicit discussions of this distinction in the However, as these examples show, that way of distinguishing epidemiology literature. Perhaps that is because, historically, studies of aetiology and prognosis is problematic: cohort studies of epidemiology has overwhelmingly been concerned with questions initially healthy participants could either investigate factors that about aetiology. The purpose of this Research Note is, therefore, to cause disease to develop, or factors that predict who will develop provide an introduction to differences in the design and analysis of disease; and cohort studies of people who already have a disease cohort studies of aetiology and prognosis. could either seek to identify exposures that cause sequelae of the For more extensive discussions of design and analysis of studies disease, or exposures that predict who will develop sequelae of the of aetiology and prognosis, the interested reader is referred to the disease. excellent introductory textbook by Grobbee and Hoes.3 In addition, Holland4 provides a rigorous grounding in the concept of causation, Many epidemiologists are reluctant to use the word ‘cause’ and Jewell5 provides a clear, entry-level presentation of design of when they write reports of their cohort studies. Instead, they claim epidemiological studies aimed at understanding causation. Hernan that they are seeking to identify ‘associations’. The justification for this practice appears to be that epidemiological studies use observational designs, which provide a less rigorous foundation for http://dx.doi.org/10.1016/j.jphys.2014.07.005 1836-9553/ß 2014 Australian Physiotherapy Association. Published by Elsevier B.V. All rights reserved.

242 Appraisal Research Note inferences about causation than experimental (ie, randomised) In studies of aetiology, the objective is to determine the causal studies, so observational studies should not be used to support effect, so it is essential to control for confounding. This can be claims about causation. However, demonstration of the existence achieved at the design stage by sampling study participants from of an association between an exposure and a health outcome is of within each stratum of the confounder, and at the analysis stage by little or no intrinsic interest. Ultimately, demonstration of an estimating associations within strata or using statistical modelling association is only useful if the association can be shown to be to ‘adjust for’ confounders.16 These control strategies are usually either causal or predictive.14 It may be difficult to establish imperfect, because the confounder may be measured imperfectly causation, but establishing causation is, nonetheless, often the (random measurement error produces what is known as ‘regres- ultimate objective. sion dilution bias’17 in statistical adjustments) and because statistical models used to adjust for confounding may be specified The reluctance of some epidemiologists to make claims about incorrectly (eg, non-linear relationships may be treated as linear causation means that it is often not clear whether particular cohort relationships). To the extent that there are important confounders, studies are designed to investigate aetiology or prognosis. The lack failure to properly control for confounding will produce biased of clarity is not helped by the use of vague terminology: reports of estimates of causal effect in aetiologic cohort studies. A researcher epidemiological studies refer to the exposures of interest as ‘risk who seeks to study aetiology must therefore identify all potentially factors’ or ‘predictors’ without ever making it clear if the interest is important confounders, measure them without error, and properly in causal factors (for the purposes of understanding disease control or adjust for the confounders. To the extent that the aetiology) or predictive factors (for the purposes of making researcher is unable to do that, he or she will obtain potentially prognosis). Epidemiologists should not be shy about the objectives biased estimates of causal effects. of their cohort studies; they should be explicit about whether the aim is to investigate aetiology or prognosis. In prognostic cohort studies, there is no need to control for confounders. To the extent that confounders are strongly associated Choice of exposure variables with outcomes, it may be useful to obtain data on exposure to confounders and incorporate confounders as predictor variables in a In a classic paper,4 Holland described how he and Donald Rubin, prognostic model. However, in prognostic studies it is not obligatory pioneers in the development of research methods for establishing to disentangle causes and confounders; this is because, if the causation, proposed the slogan ‘No causation without manipula- objective is simply to make an accurate prognosis, it does not matter tion’. They meant that it is not possible to talk meaningfully about if the prognosis is based on causal variables or confounders. Even if the causal effects of a variable unless it is clear how that variable the association between exposure and outcome is due purely to could be manipulated. For example, Holland contends that it is confounding, exposure to confounders can act as a marker of usually meaningless to talk about the causal effect of race because outcome. At least in theory, when the intention is to study prognosis, it is not possible to conceive of how a person’s race could be known confounders can be intentionally omitted from a statistical changed. The causal effect of an exposure variable must be model without compromising the validity of the predictions. In understood as the difference in an individual’s outcomes with and practice, it may be wise to include known strong confounders in without exposure to that variable,4 yet it is hard to conceive what a prognostic models, because doing so is likely to increase the person would be if he or she was other than his or her own race. In generalisability of the model to other populations. contrast, we can imagine that a person might or might not be exposed to different levels of physical activity, so we can validly Analysis ask questions about the causal effects of, say, training schedules on risk of developing lower limb overuse injuries in distance runners. The broad approach to analysing aetiologic studies and Rubin has a slightly more nuanced interpretation; in his view, it is prognostic studies is necessarily quite different. One of the most only possible to define the causal effects of a variable when it is important differences is that the analysis of aetiology should be clear what it means for a particular individual to be both exposed driven by theory, whereas the analysis of prognosis can, to a greater and not exposed to that variable.15 He would contend that extent, be driven by the data. Here, ‘theory’ is used to mean existing epidemiologists should not seek to determine the causal effect of knowledge about causes of, and associations with, the outcome. exposures that do not satisfy this criterion. The same constraint does not apply to studies of prognosis. Variables that cannot be Theory can drive the analysis of aetiologic studies in several manipulated, such as race and gender, are potentially strong ways. For example, theory will suggest plausible confounders, predictors in many contexts. which need to be controlled for in the analysis. It might also suggest whether the relationships between particular exposures Confounding and outcomes are likely to be non-linear. In addition, theory can indicate where there is the possibility of strong interactions Many associations are not causal. One of the reasons that between exposures that need to be modelled. Finally, theory might associations between exposures and outcomes occur, even when indicate whether the exposures could lie on the same causal the exposure and outcome are not causally related, is confounding. pathway. (Two exposures lie on the same causal pathway when (Another reason is the closely related concept of selection.6) one exposure causes a second exposure, which itself causes the Confounding occurs when an exposure and its outcome share a outcome of interest. In that case, the second exposure is called a common cause.6 Consider, for example, the weak association mediator.) If the aim is to identify causal effects, great care must be between cognitive impairment and the subsequent risk of falling. taken in the construction of statistical models that incorporate That association may be causal: cognitive impairment could cause mediators. For this reason, and because it is essential to identify falls. But it is also possible that the association between cognitive and control for all potentially important confounders, aetiologic impairment and risk of falling is a non-causal association caused by studies often use complex statistical models. These models may confounding. Confounding might arise if, for example, severe include many variables, continuous variables may have non-linear Parkinson’s disease causes both cognitive impairment and falls. relationships with outcomes, there may be interactions between In that case, disease severity would be a confounder of the variables, and the model may allow for mediators lying on the relationship between cognitive impairment and falls. In the same causal pathway. presence of confounding, the strength of the association between an exposure and an outcome does not provide a measure of the In prognostic studies, any statistical model that accurately strength of the causal effect of the exposure on the outcome. predicts outcomes can do the job. It matters relatively little if the statistical model is specified incorrectly. That is, it is of little consequence if particular causal variables are omitted from

Appraisal Research Note 243 the statistical model, or if a non-linear relationship is modelled as a relative risk or odds ratio. The uncertainty in (or imprecision of) linear relationship, or if an interaction or a mediator effect is estimates of the average causal effects are usually quantified ignored, and so on. (The caveat is that an incorrectly specified with confidence intervals. statistical model is not likely to predict outcomes as strongly as a correctly specified model, and may not perform well when applied In prognostic studies, there is no interest in causal effects, so to a population that differs from the study sample.) there is no need to contrast outcomes of exposed and unexposed people. Instead, the primary purpose is to estimate the expectation There are compelling reasons to ignore complexity when of the outcome. For continuous outcomes, the expectation is the developing prognostic models. Complex prognostic models are mean outcome, and for binary outcomes, the expectation is the difficult to use in clinical practice. They can only be used to predict proportion of people experiencing that outcome. Prognostic outcomes of individuals if many measurements have been studies may also generate predictions in specific subpopulations, obtained from those individuals. Also, quite a bit of calculation either by stratifying risk estimates or by using statistical models to is required to generate predictions from complex models – usually generate estimates for any combination of predictive factors.16 enough to necessitate the use of a computer, but busy clinicians Even then, the ultimate aim should still be to estimate the expected don’t always have the time or inclination to use a computer to outcome within each stratum, not to contrast the risk in different generate prognoses. The most useful predictive models require strata. data on only a few easily measured variables and are simple enough to be applied to individual patients without a It is necessary to quantify the performance of prognostic computer. The study by Moseley and colleagues2 is useful because models. And in prognostic models, performance is determined not it identified a very simple prediction rule: people who experience by the precision of estimates of effect (as in studies of causation) high levels of pain (scores of 5 or more on a 10-point scale) in the but by the ability to correctly discriminate individual people’s week after wrist fracture are at high risk of developing CRPS. outcomes.18 When outcomes are measured on a binary scale, discrimination is frequently summarised with the ‘area under the As it is not necessary to properly model the effects of curve’ statistic. When outcomes are measured on continuous confounders in prognostic studies, prognostic models can be scales, discrimination can be quantified with a correlation constructed largely without regard to theory. Consequently, the coefficient (such as r) or a coefficient of determination (r2). The process of developing a predictive model tends to be more data amount of uncertainty in an individual’s prognosis can be driven than theory driven – the researcher can rely heavily on quantified with prediction intervals. statistical algorithms to find the best weighting of an optimally selected set of predictors that strongly predicts outcomes. The One statistic that is used infrequently in studies of prognosis reliance on data-driven statistical procedures has some important with binary outcomes, but which would appear to be particularly consequences. Most importantly, data-driven processes for build- useful, is the predictive likelihood ratio (see the Moseley study2 for ing statistical models risk ‘over-fitting’. That is, data-driven model one example of a study that reported predictive likelihood ratios). building may generate models that generate very accurate Likelihood ratios can be used to predict the outcomes for an predictions in the study sample, but perform relatively poorly individual based both on prior knowledge about the individual’s when applied to other people from the same population.18 The prognosis and the particular values of the predictor variables for simplistic application of stepwise variable selection procedures is that individual. They are widely used in the interpretation of known to be particularly problematic.12 The problem is greatest diagnostic tests26 and arguably should be used much more in when many variables are considered as possible predictors and studies of prognosis. Likelihood ratios can be estimated directly (if the outcome is binary) when relatively few people experience with conventional logistic regression models.27 the outcome of interest (ie, when there are few ‘cases’), therefore, interpretation is most straightforward when the number of Interpretation candidate predictors is small and the number of cases is large. A widely used guideline is that there should be no more than one The primary motivation for studying the aetiology of a health candidate predictor for every 10 cases19 (also see Vittinghoff and condition is to suggest mechanisms through which health McCulloch 20). A range of statistical approaches, including bootstrap interventions might act and, in that way, to inform development variable regression and penalised regression, have been developed of new interventions. To the extent that an exposure causes a in an attempt to deal with the problem of over-fitting.18,21–24 health condition, interventions that reduce that exposure can prevent the health condition from developing. And to the extent That is not to say that over-fitting cannot also be a problem in that a health intervention has its effects solely by reducing an studies of aetiology. An undisciplined approach to the analysis of exposure, the maximal possible effect of the health intervention is aetiologic studies can also produce over-fitting. Ultimately, any equal to the causal effect of the exposure. statistical model that uses data-driven algorithms – even careful, theory-driven analyses of causation (using, for example, the Prognostic studies do not provide estimates of causal effects, so approach described by Hosmer and Lemeshow25) – must be they should not be used to guide decisions about selection of replicated in other studies before the findings can be considered health interventions. Instead, the value of prognostic studies is that credible. they can generate prognostic information that can be used to advise patients about the probability of developing particular Statistics of interest health outcomes. In addition, data from prognostic studies can be used to identify people with a poor prognosis. Identification of There is one more important and under-recognised difference people with a poor prognosis is potentially useful because many in the way that cohort studies of prognosis and aetiology should health interventions, particularly preventive interventions, are of be analysed. In studies of aetiology, the aim is to quantify the most benefit for people with the worst prognoses.28 causal effect – the extent to which an aetiologic factor modifies health outcomes. So studies of aetiology must contrast the Summary outcomes of people who are and are not exposed to the aetiologic factor. When the outcome is measured on a This Research Note has described differences between cohort continuous scale, the most common way to quantify the causal studies designed to investigate aetiology and prognosis. While it effect is in terms of the mean difference in outcomes of exposed will often be the case that questions about both aetiology and and unexposed people. When the outcome is binary, the causal prognosis are best answered with cohort studies, the design and effect is commonly quantified with a ratio statistic, like the analysis of such studies must be quite different. The differences are summarised in Table 1.

244 Appraisal Research Note Table 1 Differences in the design and analysis of cohort studies of aetiology and prognosis. Aetiology Prognosis Objective is to predict health outcomes. Primary interest is in outcomes. Objective is to determine the causes of particular health outcomes. Primary interest is in effects. Only exposures that can be manipulated are of interest because Any exposure can be a predictor. only they can have definable causal effects. Analysis should be theory driven. Analysis can be data driven, so care must be taken with ‘over-fitting’. Models must include all non-ignorable determinants of Confounding is irrelevant. Elaborate models are often unhelpful. Simple models with outcome, including all confounders. few predictors are preferred because they are more useful in clinical practice. Exposures (putative causes) must be measured with Exposures (predictors) should be easily measured. little or no error. Model must be correctly specified or estimates of causal effects Predictions can still be accurate, even if the model is incorrectly specified. may be biased. May have to model non-linear effects, interactions and mediator variables. Analysis involves contrasting outcomes with and without exposure. Analysis involves estimating expected outcomes. Can be used to develop new health interventions. Can be used to inform people of their prognoses and to identify those at high risk for whom intervention is most likely to be effective. Robert D Herberta,b 11. Royston P, et al. BMJ. 2009;338:1373–1377. aNeuroscience Research Australia 12. Steyerberg EW. Clinical Prediction Models: a Practical Approach to Development, bSchool of Medical Sciences, University of New South Wales, Validation, and Updating. New York: Springer; 2009:497. NSW, Australia 13. de Bie RA. Physiother Theory Pract. 2001;17:161–171. 14. Kaufman JS, et al. Soc Sci Med. 2003;57:2397–2409. References 15. Rubin DB. J Am Statist Assoc. 1986;81:961–962. 16. Rothman KJ, et al. 3rd ed Modern Epidemiology. Philadelphia, PA: Lippincott-Raven; 1. Paul SS, et al. Neurorehabil Neural Repair. 2014;28:282–290. 2. Moseley GL, et al. J Pain. 2014;15:16–23. 2008:738. 3. Grobbee DE, Hoes AW. Clinical Epidemiology: Principles, Methods, and Applications 17. MacMahon S, et al. Lancet. 1990;335:765–774. 18. Harrell Jr FE, et al. Stat Med. 1996;15:361–387. for Clinical Research. Sudbury, MA: Jones and Bartlett; 2009:413. 19. Peduzzi P, et al. J Clin Epidemiol. 1996;49:1373–1379. 4. Holland PW. J Am Statist Assoc. 1986;81:945–960. 20. Vittinghoff E, McCulloch CE. Am J Epidemiol. 2007;165:710–718. 5. Jewell NP. Statistics for Epidemiology. New York: Chapman & Hall/CRC Press; 2004. 21. Harrell FE. Regression Modeling Strategies: With Applications to Linear Models, 6. Hernan MA, Robins JM. Causal Inference. I: Causal Inference Without Models 2013, Logistic Regression, and Survival Analysis. New York: Springer; 2001:568. http://www.hsph.harvard.edu/miguel-hernan/causal-inference-book/. 22. Steyerberg EW, et al. Stat Med. 2000;19:1059–1079. 7. Hernan MA, Robins JM. Causal Inference. II: Causal Inference With Models 2013, 23. Steyerberg EW, et al. Med Decis Making. 2001;21:45–56. 24. Austin PC. Stat Med. 2008;27:3286–3300. http://www.hsph.harvard.edu/miguel-hernan/causal-inference-book/. 25. Hosmer DW, Lemeshow S. Applied Logistic Regression. 2nd edn New York: Wiley; 8. Altman DG, et al. BMJ. 2009;338:b605. 9. Moons KG, et al. BMJ. 2009;338:b606. 2000. 10. Moons KG, et al. BMJ. 2009;338:1317–1320. 26. Grimes DA, Schulz KF. Lancet. 2005;365:1500–1505. 27. Knottnerus JA. Med Decis Making. 1992;12:93–108. 28. Glasziou PP, Irwig LM. BMJ. 1995;311:1356–1359.

Journal of Physiotherapy 60 (2014) 235 Journal of PHYSIOTHERAPY journal homepage: www.elsevier.com/locate/jphys Appraisal Critically Appraised Papers A caregiver-mediated home-based intervention improves physical functioning and social participation in people with chronic stroke Synopsis Summary of: Wang T-C, Tsai AC, Wang J-Y, Lin Y-T, Lin K-L, Impact Scale, Berg Balance Scale, 10-metre Walk Test, 6-minute Chen JJ, et al. Caregiver-mediated intervention can improve Walk Test, Barthel Index, and Caregiver Burden Scale. Results: All physical functional recovery of patients with chronic stroke: a 51 participants completed the study. At the end of the training, the randomized controlled trial. Neurorehabil Neural Repair 2014;doi: experimental group had significantly more improvement than 10.1177/1545968314532030. the control group in scores of the Stroke Impact Scale, including the composite physical (by 11.9 points, 95% CI 6.5 to 17.3), social Question: Does a caregiver-mediated home-based intervention participation (by 11.4 points, 95% CI 2.9 to 19.9), and general improve physical functioning and social participation in people recovery scores (by 17.2 points, 95% CI 10.0 to 24.4). The with chronic stroke? Design: Randomised controlled trial with experimental group also had significantly better outcomes than blinded outcome assessment. Setting: Home-based intervention in controls in the Berg Balance Scale (by 5.3 points, 95% CI 2.0 to 8.6), Taiwan. Participants: Key inclusion criteria were: diagnosis of free-walking velocity (by 8.9 cm/s, 95% CI 2.1 to 15.7), 6-minute hemispheric stroke; > 6 months post-onset; Brunnstrom stage III-V; Walk Test (95% CI 26.3 m, 95% CI 8.2 to 44.4), and Barthel Index (by home dwelling; and having family, friends or co-workers as 6.6 points, 95% CI 1.8 to 11.5). There were no significant differences caregivers. Key exclusion criteria were: recurrent stroke; dementia; between the groups on the Caregiver Burden Scale score. global or receptive aphasia; severe orthopaedic problems; and being Conclusion: The caregiver-mediated home-based intervention in medically unstable. Randomisation of 51 participants allocated 25 to this study improves physical functioning and social participation in the experimental group and 26 to the control group. Interventions: people with chronic stroke. The experimental group was given weekly, personalised caregiver- mediated home-based intervention training by a physiotherapist for Marco YC Pang 12 weeks. The caregiver was asked to encourage the patient to Department of Rehabilitation Sciences, perform the prescribed exercises and task-specific training at least The Hong Kong Polytechnic University, Hong Kong twice weekly. The control group received weekly visits or telephone calls from the therapist without exercise or task-specific training http://dx.doi.org/10.1016/j.jphys.2014.08.012 interventions during the same period. Outcome measures: Stroke Commentary the study is remarkable, and needs to be considered in the context of the flexible and cost-effective model of exercise delivery. A major component of rehabilitation after stroke is exercise therapy that serves to minimise the effects of brain cell damage In the current economic climate, it is imperative that healthcare and optimise motor re-learning.1 Physiotherapists have tradition- professionals identify similar interventions that can aid recovery ally been the mediators of post-stroke exercise therapy, but it has and optimise carer involvement following stroke, while being been suggested that the amount of therapy that is delivered is not mindful of available resources. This study responds to the clear optimal.2 While the evidence for novel exercise interventions to need for the provision of an evidence-based intervention that can promote motor recovery after stroke is growing, the involvement be delivered in the community setting and that is acceptable to of family and carers in a structured program of exercise delivery people with stroke and their carers. after stroke has remained largely unexamined.3 Rose Galvin The approach to exercise delivery by Wang and colleagues School of Physiotherapy, Royal College of Surgeons in Ireland, Ireland builds on the concept that people with stroke and their caregivers take responsibility for the bulk of exercise therapy.4 Wang and References colleague’s study has significant clinical and research implications, as it is the first randomised controlled trial to evaluate the delivery 1. Langhorne P, et al. Lancet Neurol. 2009;8:741–754. of exercise to people with chronic stroke by individuals who are 2. Galvin R, et al. Top Stroke Rehabil. 2008;15:365–377. not healthcare workers. Furthermore, the intervention demon- 3. Galvin R, et al. Stroke. 2011;42:681–686. strated significant gains to the person with stroke at the level of 4. De Weerdt W, et al. Lancet. 2002;359:182–183. impairment, activity limitation and participation restriction, with no additional burden on the carer. Finally, the lack of attrition from http://dx.doi.org/10.1016/j.jphys.2014.08.014 1836-9553/ß 2014 Australian Physiotherapy Association. Published by Elsevier B.V. All rights reserved.

Journal of Physiotherapy 60 (2014) 237 Journal of PHYSIOTHERAPY journal homepage: www.elsevier.com/locate/jphys Appraisal Critically Appraised Papers A program of neck exercises can prevent neck pain in office workers Synopsis Summary of: Sihawong R, Janwantanakul P, Jiamjararangsi diary. The usual care control group did not receive any prescribed W. Effects of an exercise programme on preventing neck pain exercise. Outcome measures: The primary outcome was pain among office workers: a 12-month cluster-randomized controlled greater than 30 mm on a 100 mm visual analogue scale during the trial. Occup Environ Med 2014;71:63–70. 12-month intervention period. Secondary outcome measures were neck flexor range and neck flexor endurance on a randomly Question: Does an exercise program that is focused on selected sub-group of participants (n = 40). Results: A total of stretching and endurance training reduce the 12-month incidence 534 participants (94%) completed the study. Over the 12 months, of neck pain in office workers? Design: Cluster randomised 32 of 267 (12%) participants in the intervention group, and 72 of controlled trial with concealed allocation of sites. Setting: Twelve 270 (27%) participants in the control group reported neck pain. offices in Thailand. Participants: People aged 18 to 55 years, There was a reduced risk of neck pain in the intervention group of working full time, with reduced neck flexion range and neck about half (Hazard rate ratio = 0.45, 95% CI 0.28 to 0.71). Exercise muscle endurance were included. Key exclusion criteria were: adherence (percentage of prescribed sessions completed) for the symptoms in the spine in the previous 6 months, pregnancy, a intervention group was 30 to 34% for stretching, and 57% for history of trauma to the spine in the previous 12 months, and endurance exercise. The sub-groups did not significantly differ for performing regular exercise. Randomisation of 567 participants change in neck range or neck flexor endurance, as interaction allocated 285 to the intervention group and 282 to the control effects were non-significant. Conclusion: A 12-month program of group. Interventions: The intervention group was prescribed an neck exercises prescribed for office workers with reduced range exercise program of stretching and endurance exercises. The and poor muscle endurance led to a reduced incidence of neck pain. exercise program consisted of stretching exercises for four neck muscles, each held for 30 seconds, twice each working day. Nicholas Taylor Endurance training consisted of 10 repetitions of sustained holds of Section Editor, Journal of Physiotherapy neck flexor muscles, twice each week for 12 months. Exercise participants received a reminder via mobile phone every working http://dx.doi.org/10.1016/j.jphys.2014.08.003 day for the first 3 months, and recorded exercise completion in a Commentary that neck pain experienced by participants did not have much impact on daily function. Alternatively, perhaps the NDI did not Finding ways to prevent neck pain is important because a capture significant upper limb-associated activity limitations that significant proportion of office workers who have neck pain report are often related to neck pain.3,4 Future research should measure associated activity limitations and often experience recurrent additional outcomes related to activity limitation, productivity, symptoms.1 Although neck pain most likely develops because of medical usage, etc, so that workers, clinicians, and other stake- exposure to multiple physical and psychosocial factors over time, it holders can make well-informed decisions about implementing has been suggested that new episodes of neck pain might be exercise-related strategies that may help to prevent neck pain. prevented by improving the physical capabilities of the neck and shoulder muscles.1 However, previous clinical trials on the effective- Robert J Nee ness of this prevention strategy have shown conflicting results.2 School of Physical Therapy, Pacific University, Hillsboro, USA Judging the potential value of a preventive intervention requires References weighing the resources required to implement the intervention against the relative severity of the event to be prevented. Therefore, 1. Cote P, et al. Spine. 2008;33(4S):S60–S74. the definition of neck pain (pain > 30 mm on a 100-mm visual 2. Sihawong R, et al. J Manipulative Physiol Ther. 2011;34:62–71. analogue scale that lasted > 24 hours) should be considered when 3. Mehta S, et al. Spine. 2010;35:2150–2156. interpreting the results from Sihawong et al because it does not 4. Olson W, et al. Man Ther. 2013;18:492–497. communicate the level of activity limitation associated with the pain experience. Regardless of group assignment, participants who http://dx.doi.org/10.1016/j.jphys.2014.08.015 reported neck pain had an average Neck Disability Index (NDI) score of only 7 to 8 points (0 to 100 scale). These low scores suggest 1836-9553/ß 2014 Australian Physiotherapy Association. Published by Elsevier B.V. All rights reserved.

Journal of Physiotherapy 60 (2014) 233 Journal of PHYSIOTHERAPY journal homepage: www.elsevier.com/locate/jphys Appraisal Trial Protocol A workplace exercise versus health promotion intervention to prevent and reduce the economic and personal burden of non-specific neck pain in office personnel: protocol of a cluster-randomised controlled trial V Johnston a, S O’Leary b,c, T Comans d,e, L Straker f, M Melloh g,h, A Khan a, G Sjøgaard i a School of Health and Rehabilitation Sciences, The University of Queensland, Brisbane; b Department of Physiotherapy, Royal Brisbane and Women’s Hospital, Queensland Health, Brisbane; c NHMRC Centre for Clinical Research Excellence in Spinal Pain, Injury and Health, The University of Queensland, Brisbane; d Centre for Applied Health Economics, School of Medicine, Griffith University, Brisbane; e Population and Social Health Research Program, Griffith Health Institute, Logan; f School of Physiotherapy and Exercise Science, Curtin University, Perth; g Centre for Medical Research, The University of Western Australia, Australia; h Centre for Health Sciences, Zurich University of Applied Sciences, Winterthur, Switzerland; i Institute of Sport Sciences and Clinical Biomechanics, Faculty of Health Sciences, University of Southern Denmark, Odense, Denmark Abstract Introduction: Non-specific neck pain is a major burden to office personnel will be randomly allocated to either an intervention industry, yet the impact of introducing a workplace ergonomics or control arm in work group clusters. Analysis: Analysis will be on and exercise intervention on work productivity and severity of an ‘intent-to-treat’ basis and per protocol. Multilevel, generalised neck pain in a population of office personnel is unknown. Research linear models will be used to examine the effect of the intervention question: Does a combined workplace-based best practice on reducing the productivity loss in dollar units (AUD), and severity ergonomic and neck exercise program reduce productivity losses of neck pain and disability. Discussion: The findings of this study and risk of developing neck pain in asymptomatic workers, or will have a direct impact on policies that underpin the prevention decrease severity of neck pain in symptomatic workers, compared and management of neck pain in office personnel. to a best practice ergonomic and general health promotion program? Design: Prospective cluster randomised controlled trial. Trial registration: Australian New Zealand Clinical Trials Regis- Participants and setting: Office personnel aged over 18 years, try (ANZCTR). Registration number: ACTRN12612001154897. Was and who work > 30 hours/week. Intervention: Individualised this trial prospectively registered: Yes. Funded by: National best practice ergonomic intervention plus 3 Â 20 minute weekly, Health and Medical Research Council Project Grant. Funder progressive neck/shoulder girdle exercise group sessions for approval number: APP1042508. Anticipated completion: 2016. 12 weeks. Control: Individualised best practice ergonomic Correspondence: Dr Venerina Johnston, School of Health and Reha- intervention plus 1-hour weekly health information sessions bilitation Sciences, The University of Queensland, St Lucia, Australia. for 12 weeks. Measurements: Primary (productivity loss) and Email: [email protected] secondary (neck pain and disability, muscle performance, and quality of life) outcome measures will be collected using Full protocol: Available on the eAddenda at doi:10.1016/j.jphys. validated scales at baseline, immediate post-intervention and 2014.08.007 12 months after commencement. Procedure: 640 volunteering http://dx.doi.org/10.1016/j.jphys.2014.08.007 Commentary Recent data highlights the rising problem of musculoskeletal exercise programs or health information sessions, this comparison is conditions with neck pain now the fourth most burdensome particularly relevant. condition globally in terms of years lived with disability.1 With the combination of more people in sedentary office work and A key strength of this protocol is the plan to both assess health intensification of work practices, it is critical to effectively address outcomes for individuals and to conduct a cost-benefit analysis for musculoskeletal conditions in the workplace. productivity. For health interventions to be meaningfully translated into policy and practice, they must be economically viable. Given the The protocol by Johnston and colleagues can be seen as a intensive nature of the combined ergonomic and exercise interven- significant development on previous research. They propose to tion to be studied, weighing the cost of intervention against the investigate a combination of best practice ergonomic interventions productivity of employees will be pivotal in deciding whether such with exercise for the upper quadrant, compared with a combina- interventions should be recommended in the workplace. tion of ergonomic intervention and health promotion sessions. Niamh Moloney The design of the study is impressive. The exercise programs are Discipline of Physiotherapy, University of Sydney, Australia individualised and progressive, following best practice guidelines in exercise prescription.2 Importantly, the comparison group in this References study will receive an intervention equal in time to the exercise- training group. The comparison group, receiving health promotion 1. Vos T, et al. Lancet. 2012;380:2163–2196. information sessions, may gain health and productivity benefits 2. Kraemer WJ, et al. Med Sci Sports Exerc. 2002;34:364–380. from these sessions (eg, on stress management and keeping active; 3. Hush JM, et al. Eur Spine J. 2009;18:1532–1540. both of which have been identified as influential in neck pain).3 In deciding whether to opt for expensive and intensive supervised http://dx.doi.org/10.1016/j.jphys.2014.09.002 1836-9553/ß 2014 Australian Physiotherapy Association. Published by Elsevier B.V. All rights reserved.

Journal of Physiotherapy 60 (2014) 234 Journal of PHYSIOTHERAPY journal homepage: www.elsevier.com/locate/jphys Appraisal Trial Protocol Effect of Schroth exercises on curve characteristics and clinical outcomes in adolescent idiopathic scoliosis: protocol for a multicentre randomised controlled trial Sanja Schreiber a, Eric C Parent a,b, Douglas M Hedden a,b, Marc Moreau a,b, Doug Hill a,b, Edmond Lou a,b a University of Alberta; b Alberta Health Services, Alberta, Canada Abstract reported perceived spinal appearance and quality of life) will be assessed at baseline, and every 3 months until 1-year follow-up. Introduction: The promising results of Schroth scoliosis-specific Analysis: Data will be analysed using intention-to-treat linear mixed exercises for adolescent idiopathic scoliosis found in low-quality models. Discussion: The results will demonstrate whether Schroth studies will be strengthened by confirmation in a randomised exercises combined with standard of care can improve outcomes in controlled trial. Research questions: 1. Are Schroth exercises adolescents with idiopathic scoliosis. This study has potential to combined with standard care for 6 months more effective than influence clinical practice worldwide, where exercises are not standard care alone in improving radiographic and clinical outcomes routinely prescribed for adolescents with idiopathic scoliosis. for adolescents with idiopathic scoliosis? 2. Will the outcomes of the control group (who will be offered Schroth therapy delayed by Trial registration: ClinicalTrials.gov. Registration number: 6 months) improve after 6 months of Schroth therapy? 3. Are the NCT01610908. Was this trial prospectively registered: Yes. Funded effects maintained 6 months after discontinuing the supervised by: Scoliosis Research Society, Glenrose Rehabilitation Hospital intervention? Design: This is an assessor-blinded and statistician- Foundation, SickKids Foundation and CIHR – Institute of Human blinded randomised controlled trial with transfer of the controls to Development, Child and Youth Health (IHDCYH) New Investigator the exercise group after 6 months. Participants and setting: Two Grant. Funder approval number: SickKids Foundation and CIHR – hundred and fifty-eight consecutive adolescents with idiopathic Institute of Human Development, Child and Youth Health (IHDCYH) scoliosis, aged 10 to 16 years, treated with or without a brace, with New Investigator Grant No. NI14-018R. Anticipated completion: curves between 10 and 45 deg Cobb and Risser sign  3 will be 1 September 2016. Correspondence: Eric Parent, Department of Phys- recruited from three scoliosis clinics. Intervention: Combined with ical Therapy, University of Alberta, Canada. Email: eparent@ualberta. standard care, the Schroth group will receive five individual training ca; [email protected] sessions, followed by weekly group classes and daily home exercises for 6 months. Control: Controls will only receive standard care Full protocol: Available on the eAddenda at doi:10.1016/j.jphys. consisting of observation or bracing, and will be offered Schroth 2014.08.005 therapy 6 months later. Measurements: Curve severity (Cobb angle) and vertebral rotation will be assessed from radiographs at baseline, http://dx.doi.org/10.1016/j.jphys.2014.08.005 6 and 12 months. Secondary clinical outcomes (back muscle endurance, surface topography measures of posture, and self- Commentary Some authorities advocate exercise as a means of slowing or whether a particular type of exercise – Schroth exercise – combined reversing the progression of adolescent idiopathic scoliosis. Other with standard care is more effective than standard care alone in authorities have argued that exercise is ineffective. In some slowing or reversing curve progression over a period of 6 months. countries, exercise is routinely prescribed for adolescents with This trial and three smaller trials that are currently underway (see moderate scoliosis and in other countries it is not. http://www.who.int/ictrp/en/) will soon make it possible, for the first time, to make evidence-based decisions about whether or not The existing evidence says little about the effectiveness of exercise should be offered as an intervention for adolescent exercise. There have been at least 10 systematic reviews of clinical idiopathic scoliosis. trials on the effects of exercise for adolescent idiopathic scoliosis, and there has even been a review of the reviews.1 The plethora of Rob Herbert reviews contrasts with the dearth of evidence. The relevant Neuroscience Research Australia, Sydney, Australia Cochrane review2 of trials published up to March 2011 concludes: ‘there is low quality evidence from one randomised controlled References study that exercises as an adjunctive to other conservative treatments increase the efficacy of these treatments’ (p2). In the 1. Weiss H-RR.. Pol Ann Med. 2012;19:72–83. debate about the effects of exercise for adolescent idiopathic 2. Romano M, et al. Cochrane Database Syst Rev. 2012;8:CD007837. scoliosis, opinions are much stronger than evidence. http://dx.doi.org/10.1016/j.jphys.2014.08.008 The well-designed, large, multicentre trial described by Schreiber and colleagues will provide a rigorous answer to the question of 1836-9553/ß 2014 Australian Physiotherapy Association. Published by Elsevier B.V. All rights reserved.

Journal of Physiotherapy 60 (2014) 236 Journal of PHYSIOTHERAPY journal homepage: www.elsevier.com/locate/jphys Appraisal Critically Appraised Papers Exercise classes supervised by a physiotherapist may be better at restoring function after frozen shoulder than individual physiotherapy Synopsis Summary of: Russell S, Jariwala A, Conlon R, Selfe J, Richards J, the Constant Score measured at 6 weeks, 6 months and 12 months, Walton M. A blinded, randomized controlled trial assessing with a score of 100 denoting the highest level of functioning. conservative management strategies for frozen shoulder. J Shoulder Secondary outcome measures included the Oxford Shoulder Score, Elbow Surg. 2014;23:500-507. the Hospital Anxiety and Disability Scale and shoulder range of motion. Results: A total of 61 participants (81%) completed the Question: Does one type of physiotherapy intervention study. Across the 12 months, the Constant Score increased improve shoulder function in people with frozen shoulder more significantly more in the exercise group, by 11 units (95% CI 5 to than other types of physiotherapy interventions? Design: Ran- 17 units), compared with individual physiotherapy, and by 20 units domised controlled trial with concealed allocation and blinded (95% CI 14 to 26 units), compared with home exercise. The Constant outcome assessment. Setting: Physiotherapy outpatient clinics in Score increased significantly more in individual physiotherapy, by the United Kingdom. Participants: Inclusion criteria were: people 10 units (95% CI 4 to 16 units), compared with home exercise. The with a diagnosis of frozen shoulder (insidious onset of pain and improvement in Oxford Shoulder Score was significantly more in stiffness with reduction of range of motion of at least 50% of external the exercise group than in the individual physiotherapy or home rotation, and without underlying radiologic abnormality) and exercise groups. The improvements in the Hospital Anxiety and symptoms present for at least 3 months. Exclusion criteria were: Disability Scale anxiety scores and range of motion were signifi- history of trauma to the shoulder, shoulder inflammatory joint cantly greater in both physiotherapy groups than in the home disease and cervical spine disease. Randomisation of 75 participants exercise group. Conclusion: An exercise class supervised by a allocated 25 to an exercise group, 24 to an individual physiotherapy, physiotherapist may be more effective at restoring function and 26 to a home exercise program. Interventions: All groups for patients with frozen shoulder than individual musculoskeletal received instruction on shoulder exercises and an information physiotherapy or a home exercise program alone. booklet. In addition, the exercise group participated in a twice- weekly physiotherapist-led exercise group class for 6 weeks; [95% CIs calculated by the CAP Editor.] participants performed a 50-minute exercise circuit of 12 stations of range of motion exercises for the shoulder and thoracic spine. The Nicholas Taylor individual physiotherapy group received two individual sessions Section Editor, Journal of Physiotherapy each week for 6 weeks from a musculoskeletal physiotherapist. Individual treatment could include manual techniques, massage, http://dx.doi.org/10.1016/j.jphys.2014.08.011 stretching and heat. Outcome measures: The primary outcome was Commentary Frozen shoulder causes pain, physical impairments and and colleagues refer to the ‘pain-predominant’ and ‘stiffness- potential anxiety.1 Primary care clinicians should be able to predominant’ classification.3 Loosely interpreting this, they exclud- recognise frozen shoulder, provide reassurance and initiate a ed patients with symptoms of less than 3 months in order to treatment pathway that is informed by efficacy and cost. Individual minimise those in the ‘early pain predominant phase’. This is physiotherapy, group physiotherapy and home exercises have encouraging, but it is uncertain as to how a stricter definition of different cost implications, but there have been no previous head- stiffness-predominant (patient-reported predominance of stiffness to-head comparisons of their efficacy for frozen shoulder. over pain3) would influence the size and direction of effect. Also uncertain are the implications for intra-articular corticosteroid Commendably, Russell and colleagues conducted the first injection, which appears to be efficacious – especially when randomised controlled trial of all three of these approaches. Based combined with physiotherapy.4 Future research should integrate on a patient-reported outcome measure with a known minimal the aspect of intra-articular injection with Russell and colleagues’ clinically important difference, group therapy was found to be important contribution. statistically and clinically superior to home exercises. The Hospital Anxiety and Disability Scale anxiety scores were significantly lower Nigel Hanchard in the physiotherapy groups; this possibly reflects contact with a Health and Social Care Institute, Teesside University, United Kingdom knowledgeable therapist and fellow patients in the group. This contrasts with the likely experience of many people with frozen References shoulder. Alarmingly, 83% of the patients referred for this study had been labelled with false-positive diagnoses. If false negatives are 1. Jones S, et al. BMJ Open. 2013;3:e003452. nearly as prevalent, then isolation, confusion and anxiety may be the 2. Hanchard NCA, et al. Physiother. 2011;97:115–125. norm. 3. Hanchard NCA, et al. Physiother. 2012;98:117–120. 4. Maund E, et al. Health Technol Assess. 2012;16:1–243. While there are various systems for classifying the phases of frozen shoulder, its phasic nature is accepted by clinicians. Clinicians http://dx.doi.org/10.1016/j.jphys.2014.08.017 modify treatment accordingly,2 yet researchers typically disregard this clinical wisdom with implications for applicability. Russell 1836-9553/ß 2014 Australian Physiotherapy Association. Published by Elsevier B.V. All rights reserved.

Journal of Physiotherapy 60 (2014) 238 Journal of PHYSIOTHERAPY journal homepage: www.elsevier.com/locate/jphys Appraisal Critically Appraised Papers Multimodal physiotherapy may be no better than sham treatment for people with hip osteoarthritis Synopsis Summary of: Bennell KL, Egerton T, Martin J, Abbott JH, Metcalf B, ultrasound and inert gel lightly applied to the hip region. McManus F, et al. Effect of physical therapy on pain and function in Participants in both groups attended 10 individual physiotherapy patients with hip osteoarthritis: a randomized clinical trial. JAMA. sessions over 12 weeks; twice in the first week, once a week for 2014;311(19):1987-1997. 6 weeks, then approximately once every 2 weeks. Outcome measures: Primary outcomes were pain on a 100-mm visual Question: Does a multimodal physiotherapy program lead to analogue scale and physical function, measured on the Western greater improvements in pain and physical function than sham Ontario and McMaster Universities Index (0 to 68), assessed by a physiotherapy among people with hip osteoarthritis? Design: blinded assessor at weeks 13 and 36. Results: A total of 96 patients A randomised, controlled trial with concealed allocation and (94%) completed the 13-week assessment; there were no 24-week follow-up. Setting: Nine private physiotherapy clinics in statistically significant differences between the two groups. The Melbourne, Australia. Participants: Men and women aged 50 years mean difference in improvement for pain was 6.9 mm (95% CI À3.9 or older with hip osteoarthritis, according to the American College to 17.7), and 1.4 units (95% CI À3.8 to 6.5) for function, both of Rheumatology classification criteria, with an average pain favouring the sham treatment. Significantly more participants intensity during the past week of at least 40 on a 100-mm visual reported adverse events in the active group than in the sham analogue scale, and at least moderate difficulty with performing treatment group (41 versus 14%, p = 0.003). No significant daily activities. Key exclusion criteria included: lower limb between-group differences in change were observed 24 weeks surgery; physiotherapy, chiropractic treatment or prescribed after the intervention. Conclusion: A multimodal physiotherapy exercises in past 6 months; more than 30 minutes daily walking; program did not result in greater improvement in pain and and regular exercise more than once a week. Randomisation function than sham treatment for people with symptomatic hip allocated 49 people to the physiotherapy program and 53 to the osteoarthritis. sham treatment. Interventions: The physiotherapy program was semi-standardised with core components typical of clinical T$F[RSNDAIME]Ka˚ re BirgerDFIRST]NE.[$AM ANH[RMU]S$DTE agena and D]IRTMSNFME[A$ argrethAMESTD.$[F]RNI TGRNU]SMD$[AE rotleb practice (manual therapy; spine mobilisation; deep tissue aNational Advisory Unit on Rehabilitation in Rheumatology, massage; muscle stretches; home exercises performed four times/week; education and advice; and provision of a walking Diakonhjemmet Hospital, Oslo stick, if appropriate), plus optional techniques and exercises bDivision of Surgery and Clinical Neuroscience, depending on assessment findings. Participants were instructed to perform unsupervised home exercises three times a week during Oslo University Hospital, Norway the 6-month follow-up. The sham intervention included inactive http://dx.doi.org/10.1016/j.jphys.2014.08.013 Commentary A clear conclusion can be drawn from this high-quality strength, aerobic capacity and/or range of motion.4 However, the randomised controlled trial: this multimodal physiotherapy pro- exercise part in the study of Bennell et al was mainly delivered as gram did not give additional clinical benefit over a placebo- home exercises, which limited the control of performance and controlled sham intervention, and was associated with relatively adherence to the program; this provides a possible explanation for frequent, but mild, adverse effects. the observed lack of clinical benefit. The need for individually tailored, supervised exercise programs of adequate dosage alongside Physiotherapy is typically delivered as a comprehensive package education is still the current recommendation for people with hip of care; therefore, Bennell et al aimed to test the hypothesis that a osteoarthritis.4,5 multimodal program could have beneficial effects on pain and function. However, focusing on several elements within the time [MDIFHS]TRAE$N anneTD[FIRS$.MANE] NS$T]RUDEAMD[ agfinrud, TDMNSRIEARF]$[ ikke HeleneFIRS[TD$NEAM.] []ANMED$RUMTS oe and ]SNE[RANTD$MFI ina]TIFERAM.[NSD$ ØMNRD$[S]TEUA stera˚ s limit of a treatment session may result in an ineffective dosage of Department of Rheumatology, Diakonhjemmet Hospital, Oslo, Norway each element. Mixed programs may even raise the risk of adverse interaction effects.1 Thus, one implication of the findings of this trial References is that physiotherapists should select modalities most appropriate for each individual patient rather than multimodal programs. 1. Abbott JH, et al. Osteoarthritis Cartilage. 2013;21(4):525–534. 2. Pisters MF, et al. Arthritis Care Res. 62(8):1087–1094. Supervision of exercise sessions increases the adherence to 3. Garber CE, et al. Med Sci Sports Exerc. 2011;43(7):1334–1359. exercise programs, and better adherence has been shown to 4. Fernandes L, et al. Ann Rheum Dis. 2013;72(7):1125–1135. improve long-term results in people with osteoarthritis.2 The 5. Fransen M, et al. Cochrane Database Syst Rev. 2014;4:CD007912. effects of exercise programs are dependent on dosage and progression,3 and recommendations underline the importance of http://dx.doi.org/10.1016/j.jphys.2014.08.016 meeting the minimal requirements to improve or maintain muscle 1836-9553/ß 2014 Australian Physiotherapy Association. Published by Elsevier B.V. All rights reserved.

Journal of Physiotherapy 60 (2014) 232 Journal of PHYSIOTHERAPY journal homepage: www.elsevier.com/locate/jphys Appraisal Trial Protocol PREvention STudy On preventing or reducing disability from musculoskeletal complaints in music school students (PRESTO): protocol of a randomised controlled trial Vera AE Baadjou a,b, Jeanine AMCF Verbunt a,b,c, Marjon DF van Eijsden-Besseling c, Ans LW Samama-Polak d, Rob A de Bie e, Rob JEM Smeets a,b,c a Department of Rehabilitation Medicine, CAPHRI, Maastricht University; b Adelante Centre of Expertise in Rehabilitation and Audiology, Hoensbroek; c Department of Rehabilitation Medicine, Maastricht University Medical Centre; d Postural exercise therapist method Mensendieck, Voorburg; e Department of Epidemiology, Musculoskeletal Disorders research group, CAPHRI, Maastricht University, Maastricht, The Netherlands Abstract Introduction: Up to 87% of professional musicians develop Shoulder and Hand questionnaire). The secondary outcome work-related complaints of the musculoskeletal system during measures are pain, quality of life and changes in health behaviour. their careers. Music school students are at specific risk for Multilevel mixed-effect logistic or linear regression analyses will developing musculoskeletal complaints and disabilities. This study be performed to analyse the effects of the program on the aims to evaluate the effectiveness of a biopsychosocial prevention aforementioned outcome measurements. Furthermore, cost-effec- program to prevent or reduce disabilities from playing-related tiveness, cost-utility and feasibility will be analysed. Discussion: It musculoskeletal disorders. Secondary objectives are evaluation of is believed that this is the first comprehensive randomised cost-effectiveness and feasibility. Methods: Healthy, first or controlled trial on the effect and rationale of a biopsychosocial second year students (n = 150) will be asked to participate in a prevention program for music students. multicentre, single-blinded, parallel-group randomised controlled trial. Students randomised to the intervention group (n = 75) will Trial registration: Nederlands Trial Register. Registration participate in a biopsychosocial prevention program that number: NTR3561. Was this trial prospectively registered: addresses playing-related health problems and provides postural Yes, date: 16-8-2012. Funded by: University Fund Limburg/SWOL, training according to the Mensendieck or Cesar methods of Ans Samama Fund. Funder approval number: not applicable. postural exercise therapy, while incorporating aspects from Anticipated completion: follow-up will continue until June 2016. behavioural change theories. A control group (n = 75) will Correspondence: Vera AE Baadjou, Maastricht University, participate in a program that stimulates a healthy physical activity FHML, Department of Rehabilitation Medicine, Maastricht, The level using a pedometer, which conforms to international Netherlands. Email: [email protected] recommendations. No long-term effects are expected from this control intervention. Total follow-up duration is two years. The Full protocol: Available on the eAddenda at doi:10.1016/ primary outcome measure is disability (Disabilities of Arm, j.jphys.2014.09.001 http://dx.doi.org/10.1016/j.jphys.2014.09.001 ß 2014 Australian Physiotherapy Association. Published by Elsevier B.V. All rights reserved. Commentary Performing artists experience high rates of musculoskeletal The investigators intend to prevent contamination of the injuries, but research in the area of musculoskeletal injury intervention between participants at the same institution by prevention for performing artists is sparse, particularly so for asking them not to divulge the nature of their intervention to other musicians.1 Musicians have highly repetitive workloads, varied students. However, in an age of social media the intervention musculoskeletal stresses related to particular instruments, individ- secrecy encouraged by the researchers may not be maintained to ual rehearsal and performance schedules, and psychological stresses the desired level. related to public performances. While these loads are similar to those experienced by athletes, the crossover of knowledge from Given the individual nature of each intervention, the aim of sports physiotherapy is limited. Professional musicians have examining the feasibility and cost of such an intervention is expressed a desire for healthcare and injury-prevention education commendable. The necessity of examining the delivery cost of a to be delivered earlier in their training than is currently happening.1 personnel-intensive program against the hoped-for benefit of reduced performance-related musculoskeletal disorders is impor- The protocol by Baadjou et al will fill a large gap in the evidence tant in this age of cost-driven healthcare. about injury prevention programs to reduce performance-related musculoskeletal disorders. The study will not only be one of the Conflict of interest declaration: None. first randomised controlled trials in this population, but also the first to use a biopsychosocial intervention. In musicians training at RI$CNTAMD[FE]S laireMANT.RIF$D[]ES Hiller a tertiary (pre-professional) level, an individual biopsychosocial Faculty of Health Sciences, University of Sydney, Australia prevention intervention will be compared to a general increased physical activity level intervention. Reference 1. Chan C, et al. Front Psychol. 2014;5:1–14. http://dx.doi.org/10.1016/j.jphys.2014.08.010 1836-9553/ß 2014 Published by Elsevier B.V. on behalf of Australian Physiotherapy Association.

Journal of Physiotherapy 60 (2014) 217–223 Journal of PHYSIOTHERAPY journal homepage: www.elsevier.com/locate/jphys Research Diabetes that impacts on routine activities predicts slower recovery after total knee arthroplasty: an observational study Nurudeen Amusat a, Lauren Beaupre b,c, Gian S Jhangri d, Sheri L Pohar e, Scot Simpson f, Sharon Warren a, C Allyson Jones b a Faculty of Rehabilitation Medicine, University of Alberta; b Department of Physical Therapy, Faculty of Rehabilitation Medicine, University of Alberta; c Orthopedic Research, Capital Health; d School of Public Health, University of Alberta; e Canadian Agency for Drugs and Technologies in Health, Ottawa; f Faculty of Pharmacy and Pharmaceutical Sciences, University of Alberta, Edmonton, Canada KEY WORDS ABSTRACT Total knee arthroplasty Question: In the 6 months after total knee arthroplasty (TKA), what is the pattern of pain resolution and Diabetes functional recovery in people without diabetes, with diabetes that does not impact on routine activities, Pain and with diabetes that does impact on routine activities? Is diabetes that impacts on routine activities an Function independent predictor of slower resolution of pain and functional recovery after TKA? Design: Recovery Community-based prospective observational study. Participants: A consecutive cohort of 405 people undergoing primary TKA, of whom 60 (15%) had diabetes. Participants with diabetes were also asked preoperatively whether diabetes impacted on their routine activities. Participants were categorised into three groups: no diabetes (n = 345), diabetes with no impact on activities (n = 41), and diabetes that impacted activities (n = 19). Outcome measures: Pain and function were measured using the Western Ontario and McMaster Universities (WOMAC) Osteoarthritis Index within the month before surgery and 1, 3 and 6 months after surgery. Demographic, medical and surgical factors were also measured, along with depression, social support and health-related quality of life. Results: No baseline differences in pain and function were seen among the three groups (p > 0.05). Adjusting for age, gender and contralateral joint involvement across the 6 postoperative months, participants with diabetes that impacted on routine activities had pain scores that were 8.3 points higher (indicating greater pain) and function scores that were 5.4 points higher (indicating lower function) than participants without diabetes. Participants with diabetes that doesn’t impact on routine activities had similar recovery to those without diabetes. Conclusion: People undergoing TKA who report preoperatively that diabetes impacts on their routine activities have less recovery over 6 months than those without diabetes or those with diabetes that does not impact on routine activities. Physiotherapists could institute closer monitoring within the hospital and community settings for people undergoing TKA who perceive that diabetes impacts on their routine activities. [Amusat N, Beaupre L, Jhangri GS, Pohar SL, Simpson S, Warren S, Jones CA (2014) Diabetes that impacts on routine activities predicts slower recovery after total knee arthroplasty: an observational study. Journal of Physiotherapy 60: 217–223] ß 2014 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/3.0/). Introduction prolonged hospitalisation) and mortality, compared to those without diabetes.4,5,10,11 Glycaemic control is a significant factor Prevalence of arthritis among adults with diabetes is high, with estimates of 48% and 52%.1,2 This is not unexpected, because both in the postoperative recovery phase of TKA. People whose diabetes arthritis and diabetes are more prevalent in older adults and have common risk factors such as obesity and cardiovascular disease. is not well controlled have higher odds of perioperative When conservative management is exhausted for arthritis, total knee arthroplasty (TKA) is a successful elective surgery to alleviate complications and mortality than those with well-controlled pain and improve function.3 Estimates of diabetes prevalence in diabetes.5 Clinical outcomes such as the Knee Society score12 people undergoing TKA range from 8 to 12%,4,5 although more recent estimates are as high as 22%.6 The increased prevalence of diabetes appear to be comparable over the long term, regardless of diabetes among people undergoing primary TKA is believed to be related to status.13,14 Although pain relief and functional recovery are increasing life expectancy, obesity and overall diabetes rates.6 primary clinical goals after TKA, few studies have examined the Similar to other surgical procedures,7–9 people with diabetes who undergo TKA are at a higher risk of surgical complications, impact of diabetes on pain and functional recovery after joint systemic complications, non-routine hospital discharges (eg, arthroplasty.13,15 Measures of function in older adults are predictive of health utilisation and mortality.16 Observational studies suggest that the greatest amount of pain relief and functional improvement occurs within the first 6 months,17–19 yet it is unclear whether the recovery pattern over this time period is different for people who have diabetes. The http://dx.doi.org/10.1016/j.jphys.2014.09.006 1836-9553/ß 2014 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/3.0/).

218 Amusat et al: Diabetes and recovery after total knee arthroplasty prognostic characteristic of diabetes on recovery after joint Outcome measures arthroplasty has traditionally been evaluated in terms of the presence or absence of diabetes, not in terms of functional difficulty The primary outcome measure was the Western Ontario and that is associated with diabetes. Evidence in high-functioning, older McMaster Universities Osteoarthritis Index (WOMAC), a self- women suggests that self-reported difficulty in performing activities administered health questionnaire that is designed to measure is a strong indicator of preclinical disability.20 Specifically, asking disability of the osteoarthritic knee.21 Participants were asked to people about their preclinical difficulty with functional activities respond specifically about the knee that was being replaced. The appears to be informative of forthcoming disability. WOMAC index yields aggregate scores for joint-specific pain (five items), stiffness (two items) and physical function (17 items). Each The primary aim of the present study was to determine whether item uses a 5-point Likert scale. The range of subscale scores people with diabetes have different patterns of recovery for both ranged from 0 to 100 points, with a score of 0 indicating no pain or pain and function over 6 months after TKA than those without dysfunction. Because improvements of 23 points for joint pain and diabetes. Better defining the pre-surgical effect of diabetes on the 19 points for joint function on the WOMAC index are typically recovery of TKA will have direct clinical importance when rated by people as somewhat better as opposed to equal,22 the screening for surgical candidates and planning postoperative differences between groups were considered against this thresh- management. From a rehabilitation perspective, diabetes was old. The WOMAC index has been found to be valid, reliable, and defined in terms of the impact that it has on function, because it responsive in people with arthritis and after arthroplasty.21,23,24 may provide a far richer depiction of the severity of the condition on pain and functional outcomes for TKA. The a priori hypothesis Diabetes status was determined by self-report and/or medical specified that participants with diabetes who identified prior to chart. Because one of the primary outcomes was functional status, surgery that diabetes affected their routine activities would have a participants were asked to rate how much impact diabetes had on slower recovery after TKA than those without diabetes or with performing their routine activities by using a 4-point Likert scale diabetes that did not affect routine activities. Therefore, the (none, mild, moderate or severe). Participants were asked this at specific research questions for the present study were: baseline and at the three follow-up interviews. They were not reminded of their ratings in prior interviews. Using their baseline 1. In the 6 months after TKA, what is the pattern of pain relief and rating, participants were then classified into one of three groups: functional recovery in people without diabetes, with diabetes no diabetes, diabetes with no functional impact, and diabetes with that does not impact on routine activities, and with diabetes that functional impact. does impact on routine activities? Each participant’s overall health status was evaluated using the 2. Is diabetes that impacts on routine activities an independent Health Utilities Index Mark 3 (HUI3) – a generic, multi-attribute predictor of slower resolution of pain and functional recovery in utility measure of health-related quality of life. Because people people undergoing TKA? with diabetes have a substantial illness burden directly related the disease itself, its treatment, complications and the comorbid Method medical conditions that are prevalent in diabetes, a generic health measure was used to capture overall health. The HUI3 includes Design eight attributes of health-related quality of life, including: vision, hearing, speech, ambulation, dexterity, emotion, cognition, and This community-based, prospective, observational study pain.25,26 The overall score for the HUI3 was calculated using a recruited a consecutive cohort of participants who were undergo- multi-attribute utility function, with scores ranging from –0.36 to ing TKA within a Canadian health region. 1.0. Negative scores are assigned to health states that are considered to be worse than dead, a score of zero reflects the Participants health state dead and 1.0 reflects perfect health (full function on all eight attributes of the HUI3). A difference of at least 0.03 was The study included people who were scheduled for a primary considered to be a meaningful change for the HUI3. Construct elective TKA at one of the three tertiary hospitals within the validity of the HUI3 in type-2 diabetes and in people with health region (Edmonton, Alberta, Canada) if they were aged osteoarthritis has been reported previously.27–29 The HUI3 is also 40 years or older, residing within the health region, and able to valid in people who need a total hip arthroplasty due to speak English. People were excluded if they had hemiarthroplas- osteoarthritis.29 ties uni-compartmental revisions, or emergency arthroplasties. No bilateral joint arthroplasties were performed in this cohort. The Centre for Epidemiologic Studies Depression Scale (CES-D) was used to screen for depressive symptoms. The scale has All patients were managed using the health region’s clinical 20 items and each item is scored on a 4-point ordinal level, which pathway for TKA to ensure standardised medical, pharmacological generates a total score with a range from 0 to 60.30 The CES-D has and rehabilitative care during their hospital stay. All 29 orthopae- good internal consistency with an alpha of 0.85 in the general dic surgeons who were practising at one of the three hospitals population and has satisfactory test-retest reliability.31 Partici- within the health region gave permission for their patients to be pants were categorised into two groups: 0 to 15 indicated absent contacted for participation in the study. depressive symptoms, and 16 or higher indicated depressive symptoms.30 Using this threshold had high sensitivity (100%) and After consent was obtained, participants were interviewed specificity (88%) for depression in the previous month in a during their preadmission clinic visit within the month prior to community-based sample of older adults between the ages of surgery. Follow-up interviews were completed at 1, 3 and 6 months 55 and 85 years.32 after surgery. In-person interviews were completed at the pread- mission clinic visit and the follow-up interviews were conducted by To evaluate social support, participants completed the 19-item telephone. Home interviews were conducted for participants who Medical Outcomes Study Social Support Survey (MOS),33 which were unable to complete telephone interviews. A trained research includes items related to tangible support, affection, positive social assistant, who was an allied health professional not directly involved interaction, and emotional or informational support. The total in the care of the participants, conducted the interviews. Chart score is a weighted average of all items, rescaled to range from 0 to reviews using a standardised data-collection form were performed 100, with higher scores representing greater available social after hospital discharge to obtain surgical and perioperative support. information, including: type and number of in-hospital postopera- tive complications; discharge status; length of stay; and medical Comorbid conditions were identified from a list of predefined information including diabetes, height and weight. comorbid conditions obtained from the Charlson Comorbidity Index34 and the Canadian National Population Health Survey.35 No gold standard exists regarding the measurement of comorbidity.

Research 219 Although hospital-based comorbidity measures perform better at WOMAC pain and function scores, depression, kidney disease, MOS predicting variables of health service utilisation, self-reported social support score, HUI3 score, other weight-bearing joint comorbid conditions is as useful when measuring the impact of the involvement, age and gender were treated as fixed effects where comorbidity on health-related quality of life.36 The specific the fixed effects describe the mean change in the population. comorbidities were derived from self-report and/or admission conditions listed in the hospital chart. A p-value was considered to be statistically significant if less than 0.05 for main level factors and if less than 0.10 for interaction Data analysis terms. Age, gender, and other weight-bearing joint involvement variables were kept in all the multivariable models regardless of Descriptive statistics were used to characterise the cohort and their statistical significance. All statistical testing was performed univariate analyses were performed. Although participants were with two-tailed tests. asked to rate the impact of diabetes on routine activities, the mild, moderate and severe categories were collapsed into one category Results because very few participants reported moderate or severe impact. Participants who did not report having diabetes but had a Of the 500 people who were scheduled for TKA, 405 (81%) diagnosis of diabetes in the chart were categorised as having participated in the study. The characteristics of participants are diabetes without impact on their routine activities. presented in Table 1. The mean age of the cohort was 68 years (SD 10) and 249 (62%) were female. In total, 380 (94%) participants Linear mixed modelling was used to examine the pattern of had two or more comorbid conditions, among which 60 (15%) recovery for WOMAC pain and function scores over the four time had diabetes. Hypertension was the most prevalent comorbidity points because non-linear equations, as opposed to a linear (n = 216, 53%) followed by low back pain (n = 155, 38%). equation, provided the best fit for predicting pain and function Contralateral joint involvement affected 117 (18%) at the hip scores over the 6 months. Linear mixed modeling also allowed and 298 (25%) at the knee. Postoperative in-hospital complications available data to be used at each time period, unlike repeated occurred in 18% of participants with diabetes and 13% of measures analysis, which requires complete datasets over all time participants without diabetes. The most common types of periods.19 The linear mixed models included parameters that complications were postoperative delirium (n = 17, 4%), joint or estimated either pain or function for TKA before surgery, and the wound infection (n = 15, 4%) and urinary tract infection (n = 14, rate of change during the recovery. The square of time was also 3%). The mean length of stay in acute care was 6 days (SD 3). included as an estimate of change in the recovery rate because of the quadratic relationship over time for WOMAC pain and function The diagnosis of diabetes had 97% exact agreement between scores. The model had two levels, which consisted of one level for chart review and participant reports. Of the 60 participants with the within-individual change over time and the other for between- diabetes, 19 (32%) participants reported that diabetes impacted their individual differences in change over time. ability to perform daily routine activities. The number of partici- pants with self-reported diabetes remained relatively constant over In the multivariate linear mixed models, variables were the 6 months. Eighty percent of participants with diabetes had selected using both forward selection and backward elimination hospital admission glucose levels above 6.0 mmol/L and 65% were procedures. Forward selection started with a simple linear mixed taking either oral hypoglycaemics or insulin for their diabetes. model, then considered all of the reasonable one-step-more- complicated models and chose the one with the smallest p-value No significant differences were seen between the diabetic and for the new parameter. This continued until no additional variables non-diabetic participants for age (p = 0.42), gender (p = 0.26), or had a significant p-value. Backward elimination started with a chronic comorbidities such as heart disease, kidney disease and complicated model, including all those variables with a p- visual impairment, as presented in Table 1. Participants with value < 0.2 in the univariate linear mixed model, and removed diabetes that impacted on routine activities had a mean body mass the variable with the largest p-value at each step, as long as that p- index (BMI) of 35.8 kg/m2 (SD 7.1), which was higher than value was larger than 0.05. In the final multivariable linear mixed participants with diabetes that did not impact on routine activities models, all variables with a p-value of less than 0.05 or clinically (mean 33.7 kg/m2, SD 6.6) and participants without diabetes important variables with a p-value close to 0.05 were kept in the (mean 31.7 kg/m2, SD 6.3). models. Within this model, time squared, diabetes status, baseline Pre-operative WOMAC pain and function scores were similar among the three groups (Figure 1). At 1, 3 and 6 months after Table 1 Baseline characteristics of participants by diabetes status. Characteristics Total No diabetes Diabetes without Diabetes with (n = 405) (n = 345) impact on routine impact on routine activities (n = 41) activities (n = 19) Age (yr), mean (SD) 68 (10) 68 (10) 68 (9) 65 (12) Gender (female), n (%) 249 (62) 216 (63) 19 (46) 14 (74) Marrried or de facto, n (%) 260 (64) 220 (64) 29 (71) 11 (58) MOS score (0 to 100), mean (SD) 79 (22) 78 (23) 83 (22) 78 (18) Completed high school, n (%) 316 (78) 71 (79) 30 (73) 15 (79) Body mass index (kg/m2), mean (SD) 32.1 (6.5) 31.7 (6.3) 33.7 (6.6) 35.8 (7.1) Comorbidities (n), mean (SD) 5.7 (2.2) 5.5 (2.2) 6.7 (2.0) 7.3 (2.2) Cardiac disease, n (%) 90 (22) 76 (22) 10 (24) 3 (16) Depression (CES-D ! 16), n (%) 85 (21) 72 (21) 4 (10) 3 (16) Kidney disease, n (%) 2 (11) Visual impairment, n (%) 16 (4) 12 (4) 2 (5) 9 (47) Other weight-bearing joint involvement a, n (%) 139 (34) 120 (35) 10 (24) 15 (79) In-hospital complications, n (%) 346 (86) 295 (86) 36 (88) 3 (16) 55 (14) 44 (13) 8 (20) 2 (11) joint-related medical-related 11 (3) 7 (2) 2 (5) 1 (5) joint and medical-related 40 (10) 33 (10) 6 (15) 0 (0) Acute care length of stay (d), mean (SD) 0 (0) 6.0 (2.3) Discharged directly home from acute care, n (%) 4 (1) 4 (1) 7.3 (5.6) 7 (44) 6.3 (3.0) 6.2 (2.6) 21 (55) 231 (60) 203 (61) MOS = Medical Outcomes Study Social Support Survey, CES-D = Centre for Epidemiologic Studies Depression Scale. a Contralateral (hip, knee and/or ankle) and/or ipsilateral weight-bearing (hip and/or ankle) joint(s).

[(Figure_1)TD$]GI220 Amusat et al: Diabetes and recovery after total knee arthroplasty group than the other two groups at 3 (p < 0.01) and 6 months (p < 0.05). At baseline, the overall HUI3 scores for the three groups differed by more than 0.03, which was the threshold that was adopted as being clinically meaningful. Participants who reported diabetes that impacted on routine activities had a mean score of 0.38 (SD 0.32), while those with diabetes without impact scored 0.47 (SD 0.21) and those with no diabetes scored 0.50 (SD 0.25), as shown in Figure 1. These health-related quality of life scores improved over the 6 months after surgery in all three groups. While participants with diabetes that impacted on routine activities reported lower overall health at all four time points, differences of 0.03 or greater were not seen between the other two groups over the three follow- up time points. The numerical data used to generate Figure 1 are available in Table 2 in the eAddenda. The unadjusted parameter estimates in Table 3 show that participants with diabetes that impacted on routine activities reported less reduction in pain over the 6 months after surgery than the other two groups. Poorer health status, less perceived social support, living alone, kidney disease, and depression at baseline predicted less reduction in pain over the 6 months after surgery. Several baseline factors (health status, perceived social support, living alone, kidney disease and depression) were also significantly predictive of functional recovery over the 6 months. When adjusting for other factors such as age, gender and other weight-bearing joint involvement in the multivariable model (Table 4), variables associated with less reduction in pain included diabetes with an impact on routine activities, depression and less social support, and kidney disease. Similarly, variables associated with less functional improvement included diabetes with an impact on routine activities, poorer health status, kidney disease and less social support. Over the course of recovery, pain scores were an average of 8.3 units higher, which indicated greater pain in the group with diabetes that impacted on activities compared to the group without diabetes. Function scores were an average of 5.4 units higher, indicating lower function in the diabetes with impact group than the group without diabetes. Figure 1. Unadjusted mean Westarn Ontario and McMaster Osteoarthritis Index Discussion (WOMAC) scores and overall Health Utility Index 3 (HUI3) scores for total knee arthroplasty within the preoperative month (pre-op) and at the three postoperative The results of this longitudinal study suggest that recovery over measurement times. Error bars indicate standard deviations. (a) WOMAC pain 6 months after TKA was slower in participants who reported scores with lower scores representing less pain. (b) WOMAC function scores with diabetes that impacts on routine activities than either those lower scores indicating greater function. (c) Overall HUI3 scores. Negative scores without diabetes or those with diabetes that does not impact on assigned to health states are considered to be worse than dead, a score of zero routine activities. Although there were no differences in pain or reflects the health state ‘dead’ and 1.0 indicates ‘perfect health’. function before surgery among the three groups, different patterns of recovery were seen, depending upon the perception of impact of surgery, participants with diabetes that impacted on routine diabetes on functional activities. Participants with diabetes that activities had greater pain scores than the other two groups. These impacted on their activities had less resolution of pain and less differences were of a magnitude that people typically consider to functional improvement than the other two groups. be somewhat different.22 A similar pattern was also seen with the WOMAC function scores. Participants with diabetes that impacted Preoperative joint pain and function were similar for the three on routine activities had poorer function than the other two groups groups, yet clinical differences for overall health (HUI3 scores) (Figure 1). Although no statistically significant differences were were seen among the three groups over the four time points. The seen among the groups at 1 month, function scores were lower health status in this study cohort could be due, in part, to significantly poorer for participants in the diabetes with impact greater pain and ambulatory deficits, which are attributes that are heavily weighted in the overall HUI3 score. Over the course of the present study, the three groups had considerably lower health status, as seen with lower HUI3 scores when compared to the general community-dwelling population with diabetes without comorbidities (0.88), those with one comorbidity (0.77 to 0.79), and those with two comorbidities (0.64 to 0.66).37 To our knowledge, this is the first study to show that the severity of diabetes, as indicated by its perceived impact on function, was predictive of recovery after TKA. While most studies have defined diabetes as a dichotomous variable or in terms of glycemic control, asking participants to report the impact of a condition on routine activities provides insight into the functional impact of the condition. This has direct implications for physiotherapists in their assessment of people undergoing TKA.

Research 221 Table 3 Unadjusted parameter estimates for the Western Ontario and McMaster Universities Osteoarthritis Index (WOMAC) pain and function scores.a Category Referent Pain Function Factors Coeff (95% CI) p-value Coeff (95% CI) p-value Sociodemographic male À0.03 (À0.12 to 0.06) 0.51 0.04 (À0.04 to 0.13) 0.32 Age (yr) not Gender with others À0.15 (À2.11 to 1.80) 0.88 À1.07 (À2.91 to 0.77) 0.25 female High school education < 30 kg/m2 0.98 (À1.30 to 3.26) 0.40 2.01 (À0.13 to 4.14) 0.065 completed no diabetes À0.06 (À0.10 to À0.02) 0.01 À0.04 (À0.08 to À0.002) 0.04 MOS score (0 to 100) Living status < 16 2.82 (0.64 to 5.01) 0.01 3.50 (1.45 to 5.54) 0.001 alone not À4.97 (À9.23 to À0.72) 0.02 À6.37 (À10.50 to À2.25) 0.003 Health HUI3 (–0.36 to 1.0) À0.68 (À2.60 to 1.23) 0.48 0.44 (À1.36 to 2.24) 0.63 BMI ! 30 kg/m2 À0.08 (À3.12 to 2.97) 0.96 À0.10 (À2.99 to 2.79) 0.95 8.48 (4.22 to 12.75) <0.001 5.77 (1.71 to 9.84) 0.006 Medical À1.46 (À3.39 to 0.47) 0.60 Diabetes status 1.34 (À2.09 to 4.77) 0.14 À0.49 (À2.30 to 1.33) 0.41 without impact 1.31 (À1.32 to 3.94) 0.44 1.05 (À1.44 to 3.54) 0.83 with impact 4.69 (0.01 to 9.39) 0.33 À0.35 (À3.57 to 2.88) 0.02 High blood pressure 0.05 5.34 (0.95 to 9.73) Cardiac disease 3.39 (0.79 to 6.00) 0.08 Circulatory problems 2.24 (0.87 to 5.35) 0.01 2.23 (0.26 to 4.72) 0.98 Kidney disease 0.16 À0.04 (À2.58 to 2.51) CES-D ! 16 Other joint involvement b CES-D = Centre for Epidemiologic Studies Depression Scale, HUI3 = Health Utilities Index Mark 3, MOS = Medical Outcomes Study Social Support Survey. a Intercept, time, time-square, baseline pain or function, baseline x time were also included in the models. Referent reported for dichotomous variables. b Contralateral (hip, knee and/or ankle) and/or ipsilateral weightbearing (hip and/or ankle) joint(s). Although the severity of diabetes has been evaluated in terms of While no single condition is completely responsible for the glycemic control in people with total joint arthroplasty,5 it was outcome after total joint arthroplasty, other conditions associated found that admission fasting blood glucose levels were not with diabetes also had significant deleterious effects on recovery, significant in explaining the 6-month trajectories for pain and such as depression and kidney disease. Depression is not surprising function. Glycemic control was predictive of complications, because evidence has recognised that psychosocial symptoms such mortality, increased length of stay, and higher hospital charges as depression are associated with osteoarthritis38,39 and less pain after total joint arthroplasty in a large patient sample.5 Others have relief and functional gains after TKA.40,41 Chronic kidney disease is not evaluated the severity of the diabetes, but rather evaluated a serious complication of diabetes,42,43 yet kidney disease had an chronic conditions as a simple count to capture the burden of independent effect on recovery after TKA. The interaction between illness or treated diabetes as a dichotomous factor. Many of these diabetes and kidney disease was not significant. This is most likely approaches do not take into account the severity or functional because this cohort had a small proportion of kidney disease. The impact of the disease when evaluating outcomes after joint effect of kidney disease on recovery after TKA has not been arthroplasty. explicitly examined in the literature and warrants further Table 4 Model parameter estimates for the Western Ontario and McMaster Universities Osteoarthritis Index (WOMAC) pain and function scores. Factors a Pain Function Coeff (95% CI) p-value Coeff (95% CI) p-value Intercept 25.9 (16.1 to 35.7) < 0.001 19.9 (10.6 to 29.2) < 0.001 Time À8.56 (À10.10 to À7.02) < 0.001 À6.90 (À8.20 to À5.60) < 0.001 Time-square < 0.001 < 0.001 Diabetes 1.41 (1.23 to 1.59) 1.12 (0.96 to 1.28) 0.94 0.59 without impact 0.11 (À2.91 to 3.12) < 0.001 À0.79 (À3.70 to 2.11) < 0.01 with impact 8.28 (4.05 to 12.51) 5.42 (1.39 to 9.46) CES-D ! 16 3.24 (0.57 to 5.90) 0.02 0.048 Kidney disease 4.63 (À0.02 to 9.24) 0.049 4.39 (0.03 to 8.75) 0.049 MOS À0.05 (À0.09 to À0.01) 0.04 À0.04 (À0.08 to À0.001) 0.01 HUI3 À5.41 (À9.55 to À1.26) 0.14 Gender (female) À0.63 (À2.55 to 1.28) 0.58 0.39 Age (yrs) À0.03 (À0.12 to 0.07) 0.58 À1.38 (À3.20 to 0.45) 0.65 Other joint involvement b 0.96 (À1.73 to 3.64) 0.48 0.04 (À0.05 to 0.12) Baseline pain c < 0.001 0.60 (À1.97 to 3.18) < 0.001 Baseline function c 0.68 (0.62 to 0.75) < 0.001 Baseline  Time < 0.001 0.69 (0.62 to 0.75) Variance Estimates À0.10 (À0.12 to À0.07) À0.08 (À0.10 to À0.07) < 0.001 Within-patient residual < 0.001 < 0.001 Intercept 165 (151 to 180) 0.002 130 (119 to 142) < 0.001 Time 22.4 (11.9 to 42.3) 26.8 (16.9 to 42.7) < 0.001 4.1 (2.8 to 6.2) 3.1 (2.1 to 4.7) CES-D = Centre for Epidemiologic Studies Depression Scale, HUI3 = Health Utilities Index Mark 3, MOS = Medical Outcomes Study Social Support Survey. a Random effect was the intercept that indicates the average score at baseline and time. All other factors were treated as fixed effects. b Contralateral (hip = knee and/or ankle) and/or ipsilateral weight-bearing (hip and/or ankle) joint(s). c Baseline WOMAC scores.

222 Amusat et al: Diabetes and recovery after total knee arthroplasty examination, given the profile of people who are at high risk for Acknowledgements: Nil. chronic kidney disease, such as diabetes or hypertension, also receiving TKA. Correspondence: Dr. Allyson Jones, Department of Physical A strength of our study was the method used to define the Therapy, Faculty of Rehabilitation Medicine, University of Alberta, functional impact of diabetes. Diabetes was examined in the context of functional difficulty in performing routine activities, Edmonton, Canada. Email: [email protected] which was congruent with the measured outcomes, joint-specific pain and function. Rather than evaluating a single point in time References after surgery, the trajectory of recovery was examined, when the greatest gains are reported to occur when recovering from total 1. Cheng YJ, Imperatore G, Caspersen CJ, Gregg EW, Albright AL, Helmick CG. Preva- joint arthroplasty. Lastly, the external validity of the findings were lence of diagnosed arthritis and arthritis-attributable activity limitation among based on a community-based cohort within a universal healthcare adults with and without diagnosed diabetes: United States, 2008–2010. Diabetes system rather than recruitment from a single centre. Care. 2012;35:1686–1691. Some limitations also warrant recognition, in particular, 2. Bolen J, Hootman J, Helmick CG, Murphy L, Langmaid G, Caspersen CJ. Arthritis as a defining diabetes status in this cohort. Diabetes was determined potential barrier to physical activity among adults with diabetes. United States, by self-report, chart review or both. In particular, 12 (20%) 2005 and 2007. Morbidity and Mortality Weekly Report. 2008;57:486–489. participants with diabetes documented in the chart did not report having diabetes. The preoperative assessment was performed 3. Jones CA, Voaklander DC, Johnston DW, Suarez-Almazor ME. Health related quality during the month prior to surgery and it is possible that some of of life outcomes after total hip and knee arthroplasties in a community based these participants were newly diagnosed. Nevertheless, a small population. J Rheumatol. 2000;27:1745–1752. degree of misclassification of diabetes is a limitation that needs to be recognised. There was a relatively small subgroup of 4. Bolognesi MP, Marchant Jr MH, Viens NA, Cook C, Pietrobon R, Vail TP. The impact participants who reported that diabetes impacted on their routine of diabetes on perioperative patient outcomes after total hip and TKA in the United activities, yet they had a large and statistically significant effect in States. J Arthroplasty. 2008;23:92–98. the univariate and multivariable models for WOMAC pain and function scores. Although this was a community-based study that 5. Marchant Jr MH, Viens NA, Cook C, Vail TP, Bolognesi MP. The impact of glycemic included three hospitals and 29 surgeons, the small number of control and diabetes mellitus on perioperative outcomes after total joint arthro- participants with diabetes may be due, in part, to only those who plasty. J Bone Joint Surg Am. 2009;91:1621–1629. were medically fit being recommended for this elective surgery. 6. Cram P, Lu X, Kates SL, Singh JA, Li Y, Wolf BR. TKA volume, utilization, The findings from this study indicate that diabetes, along with and outcomes among Medicare beneficiaries, 1991–2010. JAMA. 2012;308: other associated comorbid conditions, is complex and burden- 1227–1236. some. Knowing which conditions account for the amount of impairment during recovery will provide direction to institute 7. Browne JA, Cook C, Pietrobon R, Bethel MA, Richardson WJ. Diabetes and early treatment priorities, both within the hospital and community postoperative outcomes following lumbar fusion. Spine. 2007;32:2214–2219. settings. Physiotherapy after total joint arthroplasty is effective during the post-discharge recovery period44,45 and providing 8. Cook C, Tackett S, Shah A, Pietrobon R, Browne J, Viens N, et al. Diabetes and targeted treatment for a subset of people who are at risk of slower perioperative outcomes following cervical fusion in patients with myelopathy. recovery may maximise their rehabilitation potential. To identify Spine. 2008;33:E254–E260. that subset, physiotherapists can simply ask during preoperative screening whether diabetes impacts on routine activities. People 9. Ganesh SP, Pietrobon R, Cecilio WA, Pan D, Lightdale N, Nunley JA. The impact who are identified in this way can be monitored more closely over of diabetes on patient outcomes after ankle fracture. J Bone Jt Surg Am. 2005; the 6 months following surgery. 87:1712–1718. What is already known on this topic: People undergoing a 10. England SP, Stern SH, Insall JN, Windsor RE. TKA in diabetes mellitus. Clin Orthop total knee arthroplasty who also have diabetes are at increased Relat R. 1990;130–134. risk of surgical complications, systemic complications, pro- longed hospitalisation and mortality. 11. Yang K, Yeo SJ, Lee BP, Lo NN. TKA in diabetic patients: a study of 109 consecutive What this study adds: Diabetes is also associated with slower cases. J Arthroplasty. 2001;16:102–106. resolution of pain and recovery of function after total knee arthroplasty, but only if the diabetes is severe enough that the 12. Noble PC, Scuderi GR, Brekke AC, Sikorskii A, Benjamin JB, Lonner JH, et al. person perceives preoperatively that it impacts on the comple- Development of a new Knee Society scoring system. Clin Orthop Relat R. 2012; tion of routine daily activities. Physiotherapists can therefore 470:20–32. prospectively identify people who are at risk of slower recovery after total knee arthroplasty simply by asking those with 13. Robertson F, Geddes J, Ridley D, McLeod G, Cheng K. Patients with Type 2 diabetes diabetes if their diabetes impacts on their daily activities. mellitus have a worse functional outcome post knee arthroplasty: a matched cohort study. Knee. 2012;19:286–289. eAddenda: Table 2 can be found online at doi:10.1016/j.jphys. 2014.09.006 14. Moon HK, Han CD, Yang IH, Cha BS. Factors affecting outcome after TKA in patients with diabetes mellitus. Yonsei Med J. 2008;49:129–137. Ethics approval: The Health Research Ethics Board at the University of Alberta approved this study. All participants provided 15. Clement ND, Macdonald D, Burnett R, Breusch SJ. Diabetes does not influence the informed consent before data collection began. early outcome of total knee replacement: A prospective study assessing the Oxford knee score, short form 12, and patient satisfaction. Knee. 2013;20:437–441. Competing interests: None declared. Source(s) of support: This study was funded, in part, by grants 16. Reuben DB, Rubenstein LV, Hirsch SH, Hays RD. Value of functional status as a from the Alberta Heritage Foundation for Medical Research, Royal predictor of mortality: results of a prospective study. Am J Med. 1992;93:663–669. Alexandra Foundation, University of Alberta Hospital Foundation, and the Edmonton Orthopaedic Research Trust. Drs. Allyson Jones 17. Ethgen O, Bruyere O, Richy F, Dardennes C, Reginster JY. Health-related quality of and Lauren Beaupre received salary support from the Alberta life in total hip and TKA. A qualitative and systematic review of the literature. J Bone Heritage Foundation for Medical Research and the Canadian Joint Surg Am. 2004;86–A:963–974. Institutes of Health Research. 18. Jones CA, Pohar S. Health-related quality of life after total joint arthroplasty: a scoping review. Clin Geriatr Med. 2012;28:395–429. 19. Kennedy DM, Stratford PW, Hanna SE, Wessel J, Gollish JD. Modeling early recovery of physical function following hip and knee arthroplasty. BMC Musculoskel Disord. 2006;7:100. 20. Fried LP, Bandeen-Roche K, Chaves PH, Johnson BA. Preclinical mobility disability predicts incident mobility disability in older women. J Gerontol A Biol Sci Med Sci. 2007;55:M43–M52. 21. Bellamy N, Buchanan WW, Goldsmith CH, Campbell J, Stitt LW. Validation study of WOMAC: a health status instrument for measuring clinically important patient relevant outcomes to antirheumatic drug therapy in patients with osteoarthritis of the hip or knee. J Rheumatol. 1988;15:1833–1840. 22. Escobar A, Quintana JM, Bilbao A, Arostegui I, Lafuente I, Vidaurreta I. Responsive- ness and clinically important differences for the WOMAC and SF-36 after total knee replacement. Osteoarthr Cartilage. 2007;15:273–280. 23. Bellamy N. WOMAC Osteoarthritis Index: A User’s Guide. Ontario: London Hospital and Research Centre; 1995. 24. Nilsdotter AK, Roos EM, Westerlund JP, Roos HP, Lohmander LS. Comparative responsiveness of measures of pain and function after total hip replacement. Arthritis Rheum. 2001;45:258–262. 25. Feeny D, Furlong W, Torrance GW, Goldsmith CH, Zhu Z, Depauw S, et al. Multi- attribute and single-attribute utility functions for the Health Utilities Index Mark 3 system. Medical Care. 2002;40:113–128. 26. Furlong W, Feeny D, Torrance GW, Goldsmith C, Depauw S, Boyle M, et al. Multiplicative Multi-attribute Utility Function for the Health Utilities Index Mark 3 (HUI3) System: A Technical Report. 1998; McMaster University Centre for Health Economics and Policy Analysis Working Paper, No. 98–11. 27. Grootendorst P, Feeny D, Furlong W. Health Utilities Mark 3. Evidence of construct validity for stroke and arthritis in a population health survey. Med Care. 2000; 38:290–299. 28. Maddigan SL, Feeny DH, Johnson JA, Dove I. A comparison of the health utilities indices Mark 2 and Mark 3 in type 2 diabetes. Med Decis Making. 2003;23:489–501. 29. Blanchard C, Feeny D, Mahon JL, Bourne R, Rorabeck C, Stitt L, et al. Is the Health Utilities Index valid in total hip arthroplasty patients? Qual Life Res. 2004;13: 339–348.

Research 223 30. Radloff LS. The CES-D Scale: a self-report depression scale for research in the 39. Knoop J, van der Leeden M, Thorstensson CA, Roorda LD, Lems WF, Knol DL, et al. general population. Appl Psych Meas. 1977;1:385–401. Identification of phenotypes with different clinical outcomes in knee osteoar- thritis: data from the Osteoarthritis Initiative. Arthrit Care Res. 2011;63:1535– 31. Hann D, Winter K, Jacobsen P. Measurement of depressive symptoms in cancer 1542. patients: evaluation of the Center for Epidemiological Studies Depression Scale (CES-D). J Psychosom Res. 1999;46:437–443. 40. Brander VA, Stulberg SD, Adams AD, Harden RN, Bruehl S, Stanos SP, et al. Predicting total knee replacement pain: a prospective, observational study. Clin 32. Beekman AT, Deeg DJ, Van Limbeek J, Braam AW, De Vries MZ, Van Tilburg W. Orthop Relat R. 2003;416:27–36. Criterion validity of the Center for Epidemiologic Studies Depression scale (CES-D): results from a community-based sample of older subjects in The Netherlands. 41. Lopez-Olivo MA, Landon GC, Siff SJ, Edelstein D, Pak C, Kallen MA, et al. Psychoso- Psych Med. 1997;27:231–235. cial determinants of outcomes in knee replacement. Ann Rheum Dis. 2011;70: 1775–1781. 33. Sherbourne CD, Stewart AL. The MOS social support survey. Soc Sci Med. 1991; 32:705–714. 42. Fox CS, Matsushita K, Woodward M, Bilo HJ, Chalmers J, Heerspink HJ, et al. Associations of kidney disease measures with mortality and end-stage renal 34. Charlson ME, Pompei P, Ales KL, MacKenzie CR. A new method of classifying disease in individuals with and without diabetes: a meta-analysis. Lancet. prognostic comorbidity in longitudinal studies: development and validation. J 2012;380:1662–1673. Chron Dis. 1987;40:373–383. 43. US Renal Data System. USRDS 2012 Annual Data Report: Atlas of Chronic Kidney 35. Statistics Canada. National Population Health Survey (2013) http://www.stat- Disease and End-Stage Renal Disease in the United States. National Institutes of can.gc.ca/concepts/nphs-ensp/nphs-ensp1-eng.htm. [Accessed 1 May 2013]. Health, National Institute of Diabetes and Digestive and Kidney Diseases. 2012:Bethesda, MD http://www.usrds.org/atlas.aspx. [Accessed 1 May 2013]. 36. Voaklander DC, Kelly KD, Jones CA, Suarez-Almazor ME. Self-report co-morbidity and health related quality of life - A comparison with record based co-morbidity 44. Coulter CL, Scarvell JM, Neeman TM, Smith PN. Physiotherapist-directed rehabili- measures. Soc Indic Res. 2004;66:213–228. tation exercises in the outpatient or home setting improve strength, gait speed and cadence after elective total hip replacement: a systematic review. J Physiother. 37. Maddigan SL, Feeny DH, Johnson JA. Health-related quality of life deficits associ- 2013;59:219–226. ated with diabetes and comorbidities in a Canadian National Population Health Survey. Qual Life Res. 2005;14:1311–1320. 45. Minns Lowe CJ, Barker KL, Dewey M, Sackley CM. Effectiveness of physiotherapy exercise after knee arthroplasty for osteoarthritis: systematic review and meta- 38. Hawker GA, Gignac MA, Badley E, Davis AM, French MR, Li Y, et al. A longitudinal analysis of randomised controlled trials. BMJ. 2007;335:812. study to explain the pain-depression link in older adults with osteoarthritis. Arthrit Care Res. 2011;63:1382–1390.

Journal of Physiotherapy 60 (2014) 209–216 Journal of PHYSIOTHERAPY journal homepage: www.elsevier.com/locate/jphys Research Educators and students prefer traditional clinical education to a peer-assisted learning model, despite similar student performance outcomes: a randomised trial Samantha Sevenhuysen a, Elizabeth H Skinner b, Melanie K Farlie a, Lyn Raitman a, Wendy Nickson c, Jennifer L Keating c, Stephen Maloney c, Elizabeth Molloy c, Terry P Haines a a Monash Health; b Western Health; c Monash University Melbourne, Australia KEY WORDS ABSTRACT Education Question: What is the efficacy and acceptability of a peer-assisted learning model compared with a Professional traditional model for paired students in physiotherapy clinical education? Design: Prospective, assessor- Students blinded, randomised crossover trial. Participants: Twenty-four physiotherapy students in the third year Learning of a 4-year undergraduate degree. Intervention: Participants each completed 5 weeks of clinical placement, utilising a peer-assisted learning model (a standardised series of learning activities undertaken by student pairs and educators to facilitate peer interaction using guided strategies) and a traditional model (usual clinical supervision and learning activities led by clinical educators supervising pairs of students). Outcome measures: The primary outcome measure was student performance, rated on the Assessment of Physiotherapy Practice by a blinded assessor, the supervising clinical educator and by the student in self-assessment. Secondary outcome measures were satisfaction with the teaching and learning experience measured via survey, and statistics on services delivered. Results: There were no significant between-group differences in Assessment of Physiotherapy Practice scores as rated by the blinded assessor (p = 0.43), the supervising clinical educator (p = 0.94) or the students (p = 0.99). In peer- assisted learning, clinical educators had an extra 6 minutes/day available for non-student-related quality activities (95% CI 1 to 10) and students received an additional 0.33 entries/day of written feedback from their educator (95% CI 0.06 to 0.61). Clinical educator satisfaction and student satisfaction were higher with the traditional model. Conclusion: The peer-assisted learning model trialled in the present study produced similar student performance outcomes when compared with a traditional approach. Peer- assisted learning provided some benefits to educator workload and student feedback, but both educators and students were more satisfied with the traditional model. Trial registration: ACTRN12610000859088. [Sevenhuysen S, Skinner EH, Farlie MK, Raitman L, Nickson W, Keating JL, Maloney S, Molloy E, Haines TP (2014) Educators and students prefer traditional clinical education to a peer-assisted learning model, despite similar student performance outcomes: a randomised trial. Journal of Physiotherapy 60: 209–216] ß 2014 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/3.0/). Introduction increase in demand, and the popularity of the 2:1 or ‘paired’ model — where two students are supervised by one clinical educator — is Health workforce shortages have been identified as a major issue growing. In theory, the paired model offers an immediate increase worldwide.1 In Australia, the increasing demand for healthcare in capacity, compared to the 1:1 model traditionally used in workers is challenging training and service delivery systems.2 Health physiotherapy placements. However, a search of four databases Workforce Australia identified ‘creating a more efficient training (Medline, CINAHL, SCOPUS and ERIC) up to June 2011, using key system’ as an important objective for 2012–2013.3 There has been a search terms synonymous with peer-assisted learning and substantial increase in the number of entry-level physiotherapy physiotherapy, yielded no randomised trials and little evidence programs in Australia in the past decade,4 but national shortages of of the actual effects of paired student models on student, educator physiotherapists persist.5 Clinical education is a prerequisite for or patient outcomes.7–11 Physiotherapy clinical educators consider program accreditation;6 however, the rising student numbers is peer-assisted learning models to be feasible8,9,12 and some prefer challenging the capacity of health service organisations to deliver this to the 1:1 model.12 Those authors recommend implementa- this fundamental component of physiotherapy education.4 tion of the paired student model in physiotherapy and reference the need for clinical educators to be prepared to facilitate peer Assigning multiple students to one educator in physiotherapy engagement. Despite the recommendation for the paired model, no clinical placements is one strategy being adopted to cope with this http://dx.doi.org/10.1016/j.jphys.2014.09.004 1836-9553/ß 2014 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/3.0/).

210 Sevenhuysen et al: Peer-assisted learning in physiotherapy clinical education studies have provided a reproducible framework, set of activities or degree, were eligible for inclusion if they were allocated to clinical specific tools to assist educators and learners in applying the placements at the health service. There were no exclusion criteria. model. Procedure Topping and Ehly13 defined peer-assisted learning as ‘the acquisition of knowledge and skill through active helping and Students were randomly paired and allocated to either supporting among status equals or matched companions’. traditional or peer-assisted learning groups for the duration of Implementation of paired student placements might vary for their 5-week cardiorespiratory and neurology clinical placements. several reasons, such as student and clinical educator preparation, Student pairs remained the same for both placements. Before placement environment and the cohesion of the student-peer random allocation occurred, a university staff member who was relationship.8,9,12,14–16 Peer interactions may take place in a not involved in the project allocated students to placements at the number of ways – from purely social support to formalised participating health service, based on student preferences. Prior to peer-assisted learning tasks. There is little knowledge of how the commencement of the study, participating clinical educators particular aspects of the peer interaction contribute to learning and were engaged in four 2-hour workshops that focused on how to maximise the impact on learning outcomes. development and facilitation of a peer-assisted learning model.21 Students attended a 2-hour tutorial on the first day of their peer- Qualitative investigations into physiotherapy education models assisted learning placement, at which they were introduced to the have reported that the company of another student on placement tools and expectations of the peer-assisted learning model. Blinded reduces student anxiety and aids learning.12,15–17 No study assessors with experience in using the Assessment of Physiother- provided a description or evaluation of the amount or type of apy Practice were seconded from the university and other health peer interaction occurring within the paired placements. A model services, and remunerated for their time. of paired student clinical education that specifically aims to facilitate peer-assisted learning may present immediate benefits Intervention within the placement and help to develop more sustainable and productive learner behaviours.18 The ability to collaborate with Peer-assisted learning model peers is highly valued by workplaces19 and is particularly In the absence of any published operational peer-assisted important in the provision of effective healthcare.20 learning model, the literature was mined for tools and frameworks Therefore, the research questions for the present study were: that could be used to facilitate peer-assisted learning between student pairs. Clinical educators participating in the trial worked 1. What are the effects of a paired student placement model that collaboratively to develop the model, utilising an iterative process incorporates specifically facilitated peer-assisted learning activi- that included four workshops, culminating in consensus (process ties, compared to a traditional teaching approach, on student and outcomes reported in more detail elsewhere).21 The final performance outcomes measured by external assessors blinded to model included a standardised series of tools that were utilised by group allocation, clinical educators and student self-assessment? students and educators during the peer-assisted learning clinical placements (Table 1), in addition to typical learning activities such 2. What are the effects of these models on the frequency of student as involvement in patient care, team meetings, tutorials and and educator participation in different learning/teaching administration. The peer-assisted learning tools could be used as activities, and the effects on their satisfaction with the clinical required, but a minimum number of applications was mandated placement? (Table 1). The minimum frequency was nominated by participating clinical educators in the workshops, based on the literature Method reviewed, and educator experience and opinions on model feasibility. While the peer-assisted learning framework encour- Design aged students to work with and learn from each other, the responsible clinical educator had supervisory responsibilities of This trial was a prospective, randomised, crossover trial minimising risk to patients and students, providing formative and comparing two models of physiotherapy clinical undergraduate summative feedback and assessment, and providing appropriate education: a traditional paired model and a peer-assisted learning education/guidance. paired model (Figure 1). Usual supervision (traditional model) Participants and setting The traditional model involved delivery of supervision accord- The trial was conducted in a tertiary metropolitan health ing to the usual practice of the clinical educators when supervising service from June to October 2011. Participating sites included pairs of students. This was not standardised but was characterised three acute hospitals, one sub-acute inpatient centre and one by supervisor feedback to learners and individualised learning outpatient rehabilitation centre. Physiotherapy students from activities including supervised practice, reflective learning and $DT)I_erugiF([G]1Monash University, in the third year of a four-year undergraduate Block 1 Block 2 Cardio Pair 1 Pair 2 Cardio Pair 4 Pair 3 Neuro Pair 3 Pair 4 Neuro Pair 2 Pair 1 PAL Trad PAL Trad Figure 1. Trial design. Cardio = cardiothoracic clinical placement, Neuro = neurological clinical placement, PAL = peer-assisted learning model, Trad = traditional model.

Research 211 Table 1 Risk identification The peer-assisted learning model.21 Complexity- Risk Matrix33 Domain Feedback Clinical reasoning Tool SNAPPS32 Structured Peer Educator Peer Verbal 2/pair/placement Structure feedback feedback observation feedback Structured Minimum frequency 3/pair/wk book book form triad Unstructured Unstructured Structured Unstructured 2/student/wk 2/student/wk 2/student/wk 1/pair/wk assessment. Peer-assisted learning activities were not scheduled or phase.21 Days where educators were absent were excluded from facilitated. the results. Outcome measures Students recorded a range of learning activity statistics, including number of times treating patients, observing, providing Outcome measures were defined a priori and completed by peer feedback, and engaging in facilitated peer learning activities. blinded assessors of clinical performance outcomes (who were not Learning activity statistics were recorded on a daily basis, using a part of the investigative team), clinical educators and students (ie, form created by educator participants during the model develop- self assessment). It was not possible to blind students or clinical ment.21 Days where students were absent were excluded from the educators to group allocation due to clear differences in the results. structure of the two education models. The Likert scale responses in the surveys were defined as: Primary outcome 1 = strongly disagree, 2 = disagree, 3 = neutral, 4 = agree, and The primary outcome measure was the Assessment of 5 = strongly agree. Physiotherapy Practice, scored by blinded outcome assessors, Data analysis supervising clinical educators, and students at the end of each 5- week placement. The Assessment of Physiotherapy Practice The Assessment of Physiotherapy Practice score was compared instrument is designed to monitor longitudinal evaluation of between groups using linear regression analysis. As this was a physiotherapy student performance in the clinical environment crossover trial, data were clustered by participants, and robust and has been shown to be reliable, with an ICC (2,1) of 0.92 (95% CI variance estimates were calculated to account for this data 0.84 to 0.96).22 It has been validated against a range of other dependency. The overall between-group result was not adjusted indicators (eg, stability in hierarchy of item difficulty, global rating for student characteristics, as student participants contributed scores) when applied by clinical educators who assessed students equally to both groups. When analysing the Assessment of during at least 4 weeks of clinical placement.23 The Assessment of Physiotherapy Practice scores by clinical area (cardiothoracic Physiotherapy Practice comprises 20 items in seven key areas that and neurological), the results were adjusted for pre-clinical map to the core competencies specified in the Australian Standards objective structured clinical examination (OSCE) score. In these for Physiotherapy.24 Each item is rated on a 5-level scale from 0 clinical area-specific analyses, results were not clustered by (infrequently/rarely demonstrates performance indicators) to 4 participant, as each participant only contributed to one education (demonstrates most performance indicators to an excellent approach within each clinical area. standard). The total Assessment of Physiotherapy Practice score ranges from 0 to 80, with a higher score representing better Educator workload statistics were added across the 5-week block performance. The standard error of measurement for the Assess- and divided by the number of days worked to yield an average ment of Physiotherapy Practice was low and the minimal number of minutes per day for each category. The between-group detectable change at 90% confidence was 7.9.23 difference was analysed using a linear mixed model. In this model, a random-effect term for educator was nested within one for site, Whilst the Assessment of Physiotherapy Practice ratings by the while education approach was a fixed effect. The educator survey supervising clinical educator and the students were longitudinal, results were analysed using the Wilcoxon signed-rank test as the blinded outcome assessors completed the Assessment of matched data. The number of student learning activities were added Physiotherapy Practice following a half-day observation of each across the 5-week block and divided by the number days present to student within the final 3 days of their placement. Although no yield an average number of occurrences per day for each category. data are currently available on the validity and reliability of the The between-group difference was analysed using a linear mixed Assessment of Physiotherapy Practice when used over a half-day model regression. The student survey results were also analysed period, the instrument provided the best option because it had using the Wilcoxon signed-rank test. construct validity for assessment of the target outcome, was used by students and educators in formative feedback on performance Results during the placement, was practical and feasible, and assessors were experienced in its application. The half-day assessment was Compliance with the trial method chosen as it afforded the introduction of blinded assessment, in comparison to the longitudinal assessments undertaken by clinical There were no dropouts in this study, but four student educators who could not be blinded to the education model being participants did not consent to being observed by the blinded delivered. outcome assessor. Therefore, the participant number for this outcome measure was 20, not 24. One educator did not complete Secondary outcomes the survey. Eight students did not complete the end-of-unit Satisfaction with the teaching and learning experience on satisfaction survey. completion of each model was measured via survey for both the Characteristics of the participants supervising clinical educator and the student. The six blinded assessors had more than 5 years of experience in Clinical educators recorded a range of workplace statistics, clinical practice and clinical education. They had current or recent including number of patients seen, time spent on administrative experience with physiotherapy students, either teaching on- tasks, direct teaching, student supervision, and quality assurance campus and/or as a clinical educator. The 14 clinical educators activities. Educator workload statistics were recorded at the end of were mostly aged between 20 and 30 years with a Bachelor-level each day on a form generated during the model development

212 Sevenhuysen et al: Peer-assisted learning in physiotherapy clinical education Table 2 Primary and secondary outcomes Characteristics of blinded assessors, clinical educators and students. Student performance Characteristic Blinded Clinical Students There were no significant differences in the Assessment of assessors educators (n = 24) Physiotherapy Practice scores between the peer-assisted learning (n = 6) (n = 14) and traditional models, whether awarded by the blinded assessor, the supervising clinical educator or the students. Similarly, there Age (yr), n (%) 0 (0) 0 (0) 8 (33) were no significant differences in the Assessment of Physiotherapy 18 to 20 0 (0) 4 (29) 13 (54) Practice scores between the peer-assisted learning and traditional 20 to 25 0 (0) 8 (57) 3 (13) models when analysed by clinical area (Table 3). 25 to 30 2 (33) 2 (14) 30 to 35 2 (33) 0 (0) 0 (0) Educator workload 35 to 40 2 (33) 0 (0) 0 (0) Analysis of educator workload statistics revealed no significant > 40 0 (0) 3 (50) 14 (58) between-group differences in any of the measured outcomes Gender, n female (%) 2 (33) (Table 4), with the exception of time spent on direct teaching and Qualification level, n (%) 1 (17) 11 (79) non-student-related quality assurance tasks (eg, projects designed 1 (7) to improve the quality of patient care). Despite minimal significant Bachelor 0 (0) 2 (14) differences in their daily workload data, educators reported that Graduate diploma 0 (0) they were more satisfied with the balance of their workload in the Master 2 (33) 16 (67) traditional model (Table 4). Tertiary education (yr), n (%) 4 (67) 2 (8) 2 6 (25) Educator satisfaction 4 0 (0) On completion of both models, clinical educators reported that 5+ 0 (0) 5 (36) Clinical practice experience (yr), n (%) 3 (50) 5 (36) they were less satisfied with the peer-assisted learning model 1 to 3 3 (50) 4 (28) overall, and in the areas of student anxiety, personal stress, time 3 to 5 0 (0) available for client service and their ability to observe and gauge 5 to 10 4 (67) students’ clinical ability (Table 5). > 10 2 (33) 3 (22) Clinical education experience (yr), n (%) 7 (50) When asked to rate on a Likert scale (1 = strongly disagree to <1 1 (17) 4 (28) 5 = strongly agree), clinical educators had a neutral response about 1 to 4 2 (33) 0 (0) their confidence in facilitating the peer-assisted learning strategies 5 to 10 3 (50) during the designated peer-assisted learning block (median 3, IQR > 10 3 to 4). Clinical educators also had a neutral response when asked if Time since engagement with students, n (%) 6 (42) their educational style and behaviours varied substantially for both current 4 (29) the peer-assisted learning and traditional clinical blocks (median 1 to 3 yr 3 (22) 3, IQR 3 to 4). When asked which model they would prefer to use in Involvement with students, n (%) 1 (7) the future, five educators stated they would use a ‘flexible peer- on-campus teaching assisted learning’ model, four indicated they would return to a clinical educator 5 (36) traditional model (but still in pairs), and four did not answer. both 4 (28) Confidence in clinical education, n (%) 5 (36) Student learning activities neutral 0 (0) There was no difference in the learning activities that students somewhat confident confident were exposed to in the areas of clinician observation, working very confident without observation, receiving individual feedback, participating Students educated (n/yr), n (%) in team meetings, time observed by the educator, administration 1 to 3 and statistics. In the peer-assisted learning model there was more 4 to 6 time spent by students observing their peers perform a full 8 to 12 assessment and treatment, and engaging in specific, facilitated Prior experience with peer-assisted peer interactions. Students received more verbal and written learning, n (%) feedback in the peer-assisted learning model. There was also more time spent in family meetings in the peer-assisted learning model; qualification. Their time in clinical practice and in clinical however, this was reported by a relatively small number education ranged from < 1 to 10 years. The average number of of participants. Five of the six pre-determined elements of the students they had educated per year before the study ranged from one to 12, indicating variable experience levels. Only one clinical educator felt ‘very confident’ in their clinical education skills and none had prior experience with peer-assisted learning. Students (n = 24) were mostly aged between 18 and 25 years and two-thirds had completed two years of tertiary education prior to clinical placements (Table 2). Table 3 Student performance outcomes as measured by the Assessment of Physiotherapy Practice (regression adjusted for pre-clinical OSCE result). APP scores Blinded assessor (n = 20) Educator (n = 24) Student (n = 24) PAL mean Trad mean Regression PAL mean Trad mean Regression PAL mean Trad mean Regression (SD) (SD) coefficient (SD) (SD) coefficient (SD) (SD) coefficient (95% CI) p (95% CI) p (95% CI) p Total score 40 43 –3 46 46 0 44 44 0 (0 to 80) (11) (10) (–7 to 7) (9) (11) (–10 to 4) (12) (10) (–5 to 5) 0.943 0.430 0.999 Cardiorespiratory unit total score 42 43 0 47 43 0 47 42 0 (0 to 80) (9) (13) (–1 to 1) (8) (3) (–1 to 0) (6) (–1 to 0) (7) 0.978 0.250 0.083 Neurological unit total score 39 43 0 45 49 0 41 47 0 (0 to 80) (13) (13) (–1 to 1) (9) (13) (–1 to 1) (9) (–1 to 1) (16) 0.909 0.982 0.590 APP = Assessment of Physiotherapy Practice, OSCE = objective structured clinical examination, PAL = peer-assisted learning model, Trad = traditional model.

Research 213 Table 4 Educator workload statistics and satisfaction with the teaching experience (n = 14). Workload statistic PAL Trad Linear mixed model coefficient (95% CI) p-value Time spent on tasks (min/day), mean (SD) 75 (37) 79 (48) –3 (–15 to 9) 0.640 direct student supervision 19 (13) 15 (19) 2 (–2 to 7) 0.314 student-related administration tasks 11 (12) 12 (15) –4 (–7 to 0) 0.040 direct teaching 14 (19) 13 (17) 0 (–5 to 5) 0.997 student assessment 21 (13) 19 (15) 3 (–1 to 7) 0.112 student feedback 79 (59) 75 (55) 6 (–6 to 17) 0.306 non-student-related administration tasks 11 (18) 5 (11) 5 (1 to 10) 0.020 non-student-related quality assurance tasks 215 (77) 213 (104) 0.661 patient attributable activity 9 (10) 8 (10) –5 (–28 to 18) 0.077 overtime 3 (0 to 5) 0.240 8 (3) 9 (3) 0 (–1 to 0) 0.000 Combined caseload of educator and students (patients/day), mean (SD) 2 3 –0.5 Satisfaction with workload a (1 = strongly disagree to 5 = strongly agree), median IQR (2 to 4) (3 to 4) (–0.8 to –0.2) PAL = peer-assisted learning model, Trad = traditional model. a Agreement with the statement ‘I was satisfied with the balance of my workload this week’. Table 5 PAL Median (IQR) Trad p-value Educator survey results for each model at the end of intervention (n = 13). 2 (2 to 2) 3 (2 to 3) 0.002 Survey responses (1 = strongly disagree to 5 = strongly agree) 2 (2 to 2) 3 (3 to 3) 0.009 2 (2 to 3) 1 (1 to 2) 0.005 I was satisfied with the model of clinical education 2 (2 to 2) 3 (2 to 3) 0.003 I was effectively able to observe and gauge students’ clinical ability 2 (2 to 3) 1 (1 to 2) 0.008 I found the clinical education model personally stressful There was sufficient time available for client service The students displayed a high degree of anxiety PAL = peer-assisted learning model, Trad = traditional model. peer-assisted learning model were performed significantly more learning and to the value of their peers’ feedback on their own often in the peer-assisted learning placement, indicating adher- learning. Students had a neutral-to-negative response about the ence to the trial protocol (Table 6). value of the contribution the elements of the peer-assisted learning model made to their learning, with the exception of the clinical Student Satisfaction educator feedback book (Table 8). On completion of both models, students reported increased When asked which model they would prefer to use in the stress and reduced satisfaction with the peer-assisted learning future, 81% students indicated that they preferred the traditional model (Table 7). When asked to rate on a Likert scale (1 = strongly model to the peer-assisted learning model. disagree to 5 = strongly agree), students reported no difficulty providing or receiving feedback from a peer. They had a neutral Only one student reported an instance where they received response regarding the value of their contributions to their peers’ conflicting knowledge, feedback or advice from the supervisor and peer, which did not adversely alter the outcome of the placement. Table 6 Student placement profile (n = 24). Aspect of student placement PAL Trad Linear mixed model coefficient (95% CI) p-value Learning activities (n/day), mean (SD) 0.69 (0.90) 0.83 (1.07) 0.16 (–0.47 to 0.79) 0.622 observed clinician patient management 0.28 (0.41) 0.32 (0.51) 0.04 (–0.28 to 0.35) 0.809 observed another AHP delivering patient management 0.49 (0.43) 0.34 (0.47) –0.16 (–0.38 to 0.07) 0.176 observed peer performing an assessment 0.46 (0.46) 0.26 (0.39) –0.20 (–0.40 to 0.00) 0.056 observed peer performing a treatment 0.27 (0.34) 0.11 (0.23) –0.15 (–0.29 to –0.02) 0.028 observed peer performing a full assessment and treatment 0.99 (1.41) 0.39 (0.82) –0.58 (–1.36 to 0.19) 0.140 worked with peer without direct clinician observation 1.40 (1.52) 2.01 (1.51) 0.63 (–0.25 to 1.50) 0.161 worked individually without direct clinician observation 1.82 (1.64) 1.19 (1.59) –0.64 (–1.59 to 0.32) 0.191 worked without peer observation 0.61 (0.76) 1.05 (0.96) 0.45 (–0.04 to 0.93) 0.073 received verbal feedback without peer present 0.10 (0.12) 0.10 (0.12) –0.01 (–0.05 to 0.04) 0.807 received feedback against the APP22,23 without peer present 0.06 (0.14) 0.01 (0.04) –0.05 (–0.09 to –0.01) 0.014 participated in family meeting 0.55 (0.64) 0.64 (0.55) 0.12 (–0.23 to 0.47) 0.504 participated in team meeting 0.77 (0.72) 1.27 (1.23) 0.51 (–0.00 to 1.03) 0.051 observed by educator performing an assessment 0.93 (0.85) 1.40 (1.46) 0.47 (–0.13 to 1.07) 0.122 observed by educator performing a treatment 0.41 (0.47) 0.63 (0.74) 0.23 (–0.10 to 0.56) 0.170 observed by educator performing a full assessment and treatment 0.09 (0.21) 0.20 (0.34) 0.11 (–0.04 to 0.26) 0.146 observed by educator co-treating with a peer 0.36 (1.31) 0.23 (1.13) –0.12 (–0.94 to 0.70) 0.777 patient-related administration 0.07 (0.26) 0.00 (0.12) –0.07 (–0.21 to 0.06) 0.299 Statistics 0.56 (0.30) 0.01 (0.08) –0.54 (–0.65 to –0.44) 0.000 Elements of the peer-assisted learning model (n/day), mean (SD) 0.51 (0.45) 0.20 (0.47) –0.33 (–0.61 to –0.06) 0.018 0.37 (0.29) 0.00 (0.00) –0.36 –0.48 to –0.25) 0.000 discussed a completed SNAPPS32 form 0.32 (0.22) 0.07 (0.28) –0.26 (–0.42 to –0.09) 0.003 received written feedback in educator feedback book 0.01 (0.05) 0.00 (0.03) –0.01 (–0.03 to 0.01) 0.297 received written feedback in peer feedback book 0.68 (0.53) 0.31 (0.41) –0.37 (–0.63 to –0.10) 0.006 completed Peer Observation Form completed Complexity-Risk Matrix33 received verbal feedback with peer present AHP = allied health professional, APP = Assessment of Physiotherapy Practice assessment tool, PAL = peer-assisted learning model, Trad = traditional model.

214 Sevenhuysen et al: Peer-assisted learning in physiotherapy clinical education Table 7 PAL Trad p-value Student satisfaction results for each model at the end of intervention (n = 16). 2 (1 to 2) Median (IQR) 0.001 Survey responses (1 = strongly disagree to 5 = strongly agree) 2.5 (1 to 3) 0.052 1 (1 to 1) 3 (3 to 3) 0.275 I was satisfied with the model 1 (1 to 2) 3 (3 to 4) 0.867 In the model I received adequate education from my supervisor 4 (3 to 4) 1 (1 to 1) 0.103 I found it difficult to receive feedback from my supervisor 3.5 (3 to 4) 1 (0.75 to 2.25) 0.471 I found it difficult to discuss feedback with my supervisor 2 (1 to 2) 3.5 (3 to 4) 0.018 I found educational value from watching my supervisor working with a patient 3 (3 to 4) I found educational value in my supervisor’s feedback on my performance 3 (2 to 3.25) I found the model personally stressful PAL = peer-assisted learning model, Trad = traditional model. One student sought assistance from the university unit co- non-clinical tasks such as administration and quality assurance ordinator over the duration of the study. The student was activities.28 Peer-assisted learning works on the assumption that undertaking the traditional model at the time of the request for learners are intrinsically motivated, can act in a collaborative assistance. manner and do not require the clinical educator to direct all of their learning.19 This notion of reduced reliance on the clinical educator Discussion was demonstrated in the results where, in the peer-assisted learning model, clinical educators spent significantly less time on This study is the first randomised trial to investigate a peer- direct teaching and more time on non-student-related quality assisted learning model in the allied health sciences in a clinical assurance activities. education setting, and one of few randomised controlled trials to examine clinical education outcomes. The peer-assisted learning Interestingly, the reduction in the burden of direct teaching did model produced similar student performance outcomes compared not lead to greater satisfaction with the peer-assisted learning with a traditional approach. A recent randomised controlled trial model. This may be because the introduction of the peer-assisted investigating the use of simulation in clinical education also found learning model represented a change in ideology and practice, and comparable student outcomes across different models of clinical may have challenged clinical educators’ traditional and more education.25 This may indicate that ‘traditional’ clinical education familiar practices. A previous study reported that peer learning can be altered without measurable change in student performance processes challenge expectations of the educator’s roles and outcomes. Unlike simulation, the peer-assisted learning model responsibilities, and require a different understanding of ways to does not require additional equipment and therefore may be more approach teaching and learning.19 This may also explain why, economically viable for health services and education providers. despite there being no difference in the average number of patients seen or the student performance outcomes, clinical educators The results demonstrate that students were not concerned by reported less satisfaction with the time available for client service delivering feedback to a peer or receiving it from a peer, but placed and their ability to observe and gauge students’ clinical abilities in higher value on the feedback delivered by the clinical educator. the peer-assisted learning model. The implementation of the peer- This finding of learners attributing more value to feedback assisted learning model as part of a research trial also involved provided by experts compared with feedback from peers is additional data collection and administration, which may have consistent with feedback studies in higher education.26 If peer- added to the burden for both educators and students and assisted learning tasks could be made more valuable for students, contributed to dissatisfaction. The data collection was required this might play an important role in shifting the traditional view of for the outcomes of the trial, but would not be part of usual practice supervision and feedback from one being led solely by the clinical when implementing a peer-assisted learning model. educator, to one that is also shared among learners. In the peer-assisted learning model, students spent more time Physiotherapy clinical educators have previously reported that in formalised peer learning tasks without sacrificing other time spent directly teaching students is burdensome,27 and elements of the clinical education placement. This may demon- that having students in the workplace takes time away from strate that peer-assisted learning activities can be utilised in paired student placements without reducing access to other learning Table 8 activities. It may have indicated that students in peer-assisted Student perceptions of peer-assisted learning model at the end of both units learning were able to use their ‘downtime’ (ie, time when, in the (n = 16). traditional approach, they may have been waiting for their clinical educator to direct their learning) to complete the designated peer- Survey responses (1 = strongly Median (IQR) assisted learning tasks. disagree to 5 = strongly agree) 3 (3 to 4) The rigid structure of the formal peer-assisted learning activities I had valuable contributions to make to 2 (2 to 2) may have contributed to the dissatisfaction with the model, a notion my peer colleagues’ learning 2 (2 to 3) that is supported by the clinical educators citing a preference for a 3 (2.75 to 4) ‘flexible peer-assisted learning’ model in the future. To ensure I found it difficult to receive feedback 3 (2 to 3.25) consistency in the research protocol, the formal elements of the from a peer 2.5 (2 to 4) peer-assisted learning model were prescribed and did not vary 2 (1 to 3.25) throughout the placement. Principles of learning dictate that an I found it difficult to deliver feedback 2 (2 to 2.75) effective teaching strategy involves a progression of increasingly to a peer 4 (4 to 4) complex tasks as knowledge and skill increase.29 Although it was 3.5 (2 to 4) theoretically possible to increase complexity of the task within the I found educational value in my peer’s prescribed activities, this may have been difficult for clinical feedback on my performance educators and students to execute, given that it was their first experience with the tools. If paired student placement models are The SNAPPS32 form aided my learning utilised in clinical education, it may be important to consider The complexity-risk matrix aided incorporating flexibility in the type and number of peer-assisted learning activities facilitated each week, although the results of the my learning trial may have been different if this approach had been tested. The peer observation record aided my learning The peer feedback book facilitated my clinical education experience The educator feedback book facilitated my clinical education experience I found educational value in observing my peer receive feedback from the supervisor

Research 215 The time allocated to familiarise students with the tools and whether these competencies influence their capacities to operate expectations of the peer-assisted learning model in this study may in the workforce. have been insufficient, which may have contributed to students’ relative dissatisfaction with the formal tools and the model itself. While peer-assisted learning activities were integrated into Students’ willingness to engage in a different learning culture to the clinical education of paired students without sacrificing traditional, teacher-led practices can affect their engagement with student performance outcomes, both educators and students peer-assisted learning19 and has been recognised as being were more satisfied with the traditional approach. The peer- important to clinical educators.30 To help address this, it may be assisted learning model provided some benefits to educator of benefit to introduce the various tools in the pre-clinical period, workload, with clinical educators reducing time spent on direct and to invest time in orientating learners about the evidence of teaching and increasing time available for quality assurance both the short-term and long-term benefits of working with and activities. Students received more written feedback in the learning with peers.9–14,16,17,19,31 It is also possible that some peer-assisted learning model, but preferred educator feedback over elements of the peer-assisted learning model may have greater peer feedback. Students and educators cited the rigidity of the model acceptability to students than others, and this will be the focus of as a source of dissatisfaction. It is therefore recommended that ongoing investigations. clinical educators using a paired student model incorporate flexibility in the type and number of learning activities facilitated The project was conducted in one health service with one group in the placement. of clinical educators, which limits generalisability. Clinical educa- tor participants were volunteers and therefore a self-selecting What is already known on this topic: Peer-assisted learning group. Issues may have been missed that related specifically to incorporates learning activities undertaken by student pairs clinical educators who did not volunteer. For example, clinical and educators to facilitate peer interaction using guided strat- educators who volunteered may have been particularly enthusi- egies. The peer-assisted learning model has potential advan- astic or motivated about their clinical education role. There was tages in the clinical education of physiotherapy students. potential for response bias in the survey, as participants may have What this study adds: The peer-assisted learning model and built a relationship with the lead investigator through the research a traditional paired model of clinical education produced process. similar student performance outcomes. The peer-assisted learning model produced some modest benefits: educators In trials of educational approaches, keeping the intervention had more time for other work activities and students received consistent with a protocol can be seen as a limitation because it is more written feedback. Despite this, educators and students counter to best practice educational principles, such as tailoring preferred the traditional model. activities to the individual and increasing complexity as the student’s mastery improves. However, the minimum number of Ethics approval: The Monash Health and Monash University tasks in the peer-assisted learning approach was necessary to Human Research Ethics Committees approved this study. All permit measurement of adherence. participants gave written informed consent before data collection began. The reliability and validity of the Assessment of Physiotherapy Practice tool over a half-day observation, as was conducted by the Competing interests: None declared. blinded assessors, has not been investigated. However, the Source(s) of support: Monash Health Allied Health Research Assessment of Physiotherapy Practice has construct validity for Unit. such an application and a superior method for assessment of Acknowledgements: Monash Health physiotherapy clinical clinical performance in physiotherapy clinical education was not educators and students. available. In addition, the results did not differ when longitudinal Correspondence: Samantha Sevenhuysen, Allied Health, assessments by educators were considered and the Assessment of Monash Health, Victoria, Australia. Email: sam.sevenhuysen@ Physiotherapy Practice has been demonstrated to be both reliable monashhealth.org and valid under these conditions. References Clinical educators developed and then immediately tested the peer-assisted learning model, with no opportunity to refine the 1. World Health Organization. Working Together for Health, The World Health Report model based on their practical experiences. Educators and (2006) http://www.who.int/whr/2006/whr06_en.pdf [accessed 20/06/2013]. students were learning and testing the model simultaneously, which may have affected the results. 2. National Health Workforce Taskforce. Health Workforce in Australia and Factors for Current Shortages (2009) http://www.ahwo.gov.au/documents/NHWT/The%20 Despite resulting in equivalent student performance outcomes, health%20workforce%20in%20Australia%20and%20factors%20influencing%20 there was resistance to using the peer-assisted learning model current%20shortages.pdf [accessed 15/12/2012]. from both learners and educators. For learners, expert observation of performance and expert delivered feedback is preferred over 3. Health Workforce Australia 2012-13 Work Plan. https://www.hwa.gov.au/sites/ peer observation because ‘it means more’ (more understanding of uploads/hwa-work-plan-2012-13-approved-SCoH-20120810.pdf [accessed 15/ performance standards, more experience in observation, more 12/2012]. strategies for improvement tested). For educators, a strict peer- assisted learning model may represent threats to patient/student 4. Health Workforce Australia. National Simulated Learning Project Report for Physio- safety, to quality feedback and to well-worn, familiar routines in therapy (2012) http://www.hwa.gov.au/sites/uploads/simulated-learning- clinical supervision. The resistance needs to be acknowledged, and physiotherapy-report-201108.pdf [accessed 15/12/2012]. more studies are required to determine whether the challenge is in the change of routine for both parties (expanding the envelope of 5. Human Capital Alliance. Recruitment and retention of allied health professionals in comfort) or simply because the peer-assisted learning activities are Victoria - A literature review (2005) http://www.humancapitalalliance.com.au/ not as potent as teacher-led activities. downloads/DH35%20Overall%20Main%20Report%20Final.pdf [accessed 15/12/ 2012]. Further research could evaluate whether incorporating peer- assisted learning activities into a paired student placement in a 6. Australian Physiotherapy Council. Accreditation of Entry Level Physiotherapy flexible way optimises clinical educator and student satisfaction. Programs: a manual for universities (2009) http://www.physiocouncil.com.au/ There may be improvement in clinical educator and student Accreditation [accessed 10/07/2010]. satisfaction if certain peer-assisted learning activities become more familiar and are incorporated into ‘usual practice’ or there 7. Blakely C, Rigg J, Joynson K, Oldfield S. Supervision models in a 2:1 acute care may remain a strong preference for traditional, supervisor-led placement. Brit J Occup Ther. 2009;72:515–517. learning activities. Longitudinal studies could investigate how students evolve in their peer learning practices over time, and 8. Lekkas P, Larson T, Kumar S, Grimmer K, Nyland L, Chipchase L, et al. No model of clinical education for physiotherapy students is superior to another: a systematic review. Aust J Physiother. 2007;53:19–28. 9. Moore A, Morris J, Crouch V, Martin M. Evaluation of physiotherapy clinical educational models: comparing 1:1, 2:1 and 3:1 placements. Physiotherapy. 2003;89:489–501. 10. Roberts NJ, Brockington S, Doyle E, Pearce LM, Bowie AJ, Simmance N, et al. Pilot study of an innovative model for clinical education in dietetics. Nutr Diet. 2009;66:39–46. 11. Strohschein J, Hagler P, May L. Assessing the need for change in clinical education practices. Phys Ther. 2002;82:160–172.

216 Sevenhuysen et al: Peer-assisted learning in physiotherapy clinical education 12. Baldry-Currens JA, Bithell CP. The 2:1 clinical placement model: perceptions of 24. Australian Physiotherapy Council. Australian Standards for Physiotherapy (2006). clinical educators and students. Physiotherapy. 2003;89:204–218. http://www.physiocouncil.com.au/files/the-australian-standards-for-physiother- apy [accessed 10/12/2013]. 13. Topping K, Ehly S. Peer-assisted learning. London: Lawrence Erlbaum Associates; 1998. 14. Boud D. Situating academic development in professional work: using peer learn- 25. Watson K, Wright A, Morris N, McMeeken J, Rivett D, Blackstock F, et al. Can simulation replace part of clinical time? Two parallel randomised controlled trials ing. Int J Acad Dev. 1999;4:3–10. Med Educ. 2012;46:657–667. 15. DeClute J, Ladyshewsky RK. Enhancing clinical competence using a collaborative 26. Boud D, Molloy E, eds. In: Effective Feedback in Higher and Professional Education. clinical educational model. Phys Ther. 1993;73:683–689. London: Routledge; 2013. 16. Skøien AK, Vagstol U, Raaheim A. Learning physiotherapy in clinical practice: Student 27. Bearman M, Molloy E, Ajjawi R, Keating J. ‘Is there a Plan B?’: clinical educators interaction in a professional context. Physiother Theory Prac. 2009;25:268–278. supporting underperforming students in practice settings Teach High Educ. 2012. 17. Ladyshewsky RK, Barrie SC, Drake VM. A comparison of productivity and learning http://dx.doi.org/10.1080/13562517.2012.752732. outcome in individual and cooperative physical therapy clinical education models. 28. Sevenhuysen S, Haines TP. The slave of duty: why clinical educators across the Phys Ther. 1998;78:1288–1298. continuum of care provide clinical education in physiotherapy. Hong Kong Physi- 18. Leach DC, Fletcher SW. Perspectives on continuing education in the health profes- other J. 2011;29:64–70. sions: improving health care through lifelong learning. CHEST. 2008;134:1299–1303. 19. Sampson J, Boud D, Cohen R, Gaynor F. Designing Peer Learning. In: Paper presented 29. Merrill MD. First Principles of Instruction. In: Reigeluth CM, Carr A, eds. In: at the HERDSA Annual International Conference. 1999. Instructional Design Theories and Models: Building a Common Knowledge Base 20. World Health Organization. Framework for Action on Interprofessional Education & (Vol. III). New York:Routledge; 2009. Collaborative Practice (2010). http://whqlibdoc.who.int/hq/2010/WHO_HRH_HP- N_10.3_eng.pdf [accessed 15/03/2013]. 30. Chipchase LS, Buttrum PJ, Dunwoodie R, Hill AE, Mandrusiak A, Moran M. Char- 21. Sevenhuysen S, Farlie M, Nickson W, Keating JK, Raitman L, Skinner E, et al. The acteristics of student preparedness for clinical learning: clinical educator perspec- development of a peer-assisted learning model for the clinical education of tives using the Delphi approach. BMC Med Educ. 2012;12:112. physiotherapy students. J Peer Learn. 2013;6:30–45. 22. Dalton M, Davidson M, Keating JL. The assessment of physiotherapy practice (APP) 31. Ladyshewsky RK. Enhancing service productivity in acute care inpatient is a reliable measure of professional competence of physiotherapy students: a settings using a collaborative clinical education model. Phys Ther. 1995;75: reliability study. J Physiother. 2012;58:49–56. 503–510. 23. Dalton M, Davidson M, Keating J. The Assessment of Physiotherapy Practice (APP) is a valid measure of professional competence of physiotherapy students: a cross- 32. Wolpaw TM, Wolpaw DR, Papp KK. SNAPPS: A learner-centered model for outpa- sectional study with Rasch analysis. J Physiother. 2011;57:239–246. tient education. Acad Med. 2003;78:893–898. 33. Kneebone RL, Nestel D, Vincent C, Darzi A. Complexity, risk and simulation in learning procedural skills. Med Educ. 2007;41:808–814.

Journal of Physiotherapy 60 (2014) 234 Journal of PHYSIOTHERAPY journal homepage: www.elsevier.com/locate/jphys Appraisal Trial Protocol Effect of Schroth exercises on curve characteristics and clinical outcomes in adolescent idiopathic scoliosis: protocol for a multicentre randomised controlled trial Sanja Schreiber a, Eric C Parent a,b, Douglas M Hedden a,b, Marc Moreau a,b, Doug Hill a,b, Edmond Lou a,b a University of Alberta; b Alberta Health Services, Alberta, Canada Abstract reported perceived spinal appearance and quality of life) will be assessed at baseline, and every 3 months until 1-year follow-up. Introduction: The promising results of Schroth scoliosis-specific Analysis: Data will be analysed using intention-to-treat linear mixed exercises for adolescent idiopathic scoliosis found in low-quality models. Discussion: The results will demonstrate whether Schroth studies will be strengthened by confirmation in a randomised exercises combined with standard of care can improve outcomes in controlled trial. Research questions: 1. Are Schroth exercises adolescents with idiopathic scoliosis. This study has potential to combined with standard care for 6 months more effective than influence clinical practice worldwide, where exercises are not standard care alone in improving radiographic and clinical outcomes routinely prescribed for adolescents with idiopathic scoliosis. for adolescents with idiopathic scoliosis? 2. Will the outcomes of the control group (who will be offered Schroth therapy delayed by Trial registration: ClinicalTrials.gov. Registration number: 6 months) improve after 6 months of Schroth therapy? 3. Are the NCT01610908. Was this trial prospectively registered: Yes. Funded effects maintained 6 months after discontinuing the supervised by: Scoliosis Research Society, Glenrose Rehabilitation Hospital intervention? Design: This is an assessor-blinded and statistician- Foundation, SickKids Foundation and CIHR – Institute of Human blinded randomised controlled trial with transfer of the controls to Development, Child and Youth Health (IHDCYH) New Investigator the exercise group after 6 months. Participants and setting: Two Grant. Funder approval number: SickKids Foundation and CIHR – hundred and fifty-eight consecutive adolescents with idiopathic Institute of Human Development, Child and Youth Health (IHDCYH) scoliosis, aged 10 to 16 years, treated with or without a brace, with New Investigator Grant No. NI14-018R. Anticipated completion: curves between 10 and 45 deg Cobb and Risser sign  3 will be 1 September 2016. Correspondence: Eric Parent, Department of Phys- recruited from three scoliosis clinics. Intervention: Combined with ical Therapy, University of Alberta, Canada. Email: eparent@ualberta. standard care, the Schroth group will receive five individual training ca; [email protected] sessions, followed by weekly group classes and daily home exercises for 6 months. Control: Controls will only receive standard care Full protocol: Available on the eAddenda at doi:10.1016/j.jphys. consisting of observation or bracing, and will be offered Schroth 2014.08.005 therapy 6 months later. Measurements: Curve severity (Cobb angle) and vertebral rotation will be assessed from radiographs at baseline, http://dx.doi.org/10.1016/j.jphys.2014.08.005 6 and 12 months. Secondary clinical outcomes (back muscle endurance, surface topography measures of posture, and self- Commentary Some authorities advocate exercise as a means of slowing or whether a particular type of exercise – Schroth exercise – combined reversing the progression of adolescent idiopathic scoliosis. Other with standard care is more effective than standard care alone in authorities have argued that exercise is ineffective. In some slowing or reversing curve progression over a period of 6 months. countries, exercise is routinely prescribed for adolescents with This trial and three smaller trials that are currently underway (see moderate scoliosis and in other countries it is not. http://www.who.int/ictrp/en/) will soon make it possible, for the first time, to make evidence-based decisions about whether or not The existing evidence says little about the effectiveness of exercise should be offered as an intervention for adolescent exercise. There have been at least 10 systematic reviews of clinical idiopathic scoliosis. trials on the effects of exercise for adolescent idiopathic scoliosis, and there has even been a review of the reviews.1 The plethora of Rob Herbert reviews contrasts with the dearth of evidence. The relevant Neuroscience Research Australia, Sydney, Australia Cochrane review2 of trials published up to March 2011 concludes: ‘there is low quality evidence from one randomised controlled References study that exercises as an adjunctive to other conservative treatments increase the efficacy of these treatments’ (p2). In the 1. Weiss H-RR.. Pol Ann Med. 2012;19:72–83. debate about the effects of exercise for adolescent idiopathic 2. Romano M, et al. Cochrane Database Syst Rev. 2012;8:CD007837. scoliosis, opinions are much stronger than evidence. http://dx.doi.org/10.1016/j.jphys.2014.08.008 The well-designed, large, multicentre trial described by Schreiber and colleagues will provide a rigorous answer to the question of 1836-9553/ß 2014 Australian Physiotherapy Association. Published by Elsevier B.V. All rights reserved.

Journal of Physiotherapy 60 (2014) 236 Journal of PHYSIOTHERAPY journal homepage: www.elsevier.com/locate/jphys Appraisal Critically Appraised Papers Exercise classes supervised by a physiotherapist may be better at restoring function after frozen shoulder than individual physiotherapy Synopsis Summary of: Russell S, Jariwala A, Conlon R, Selfe J, Richards J, the Constant Score measured at 6 weeks, 6 months and 12 months, Walton M. A blinded, randomized controlled trial assessing with a score of 100 denoting the highest level of functioning. conservative management strategies for frozen shoulder. J Shoulder Secondary outcome measures included the Oxford Shoulder Score, Elbow Surg. 2014;23:500-507. the Hospital Anxiety and Disability Scale and shoulder range of motion. Results: A total of 61 participants (81%) completed the Question: Does one type of physiotherapy intervention study. Across the 12 months, the Constant Score increased improve shoulder function in people with frozen shoulder more significantly more in the exercise group, by 11 units (95% CI 5 to than other types of physiotherapy interventions? Design: Ran- 17 units), compared with individual physiotherapy, and by 20 units domised controlled trial with concealed allocation and blinded (95% CI 14 to 26 units), compared with home exercise. The Constant outcome assessment. Setting: Physiotherapy outpatient clinics in Score increased significantly more in individual physiotherapy, by the United Kingdom. Participants: Inclusion criteria were: people 10 units (95% CI 4 to 16 units), compared with home exercise. The with a diagnosis of frozen shoulder (insidious onset of pain and improvement in Oxford Shoulder Score was significantly more in stiffness with reduction of range of motion of at least 50% of external the exercise group than in the individual physiotherapy or home rotation, and without underlying radiologic abnormality) and exercise groups. The improvements in the Hospital Anxiety and symptoms present for at least 3 months. Exclusion criteria were: Disability Scale anxiety scores and range of motion were signifi- history of trauma to the shoulder, shoulder inflammatory joint cantly greater in both physiotherapy groups than in the home disease and cervical spine disease. Randomisation of 75 participants exercise group. Conclusion: An exercise class supervised by a allocated 25 to an exercise group, 24 to an individual physiotherapy, physiotherapist may be more effective at restoring function and 26 to a home exercise program. Interventions: All groups for patients with frozen shoulder than individual musculoskeletal received instruction on shoulder exercises and an information physiotherapy or a home exercise program alone. booklet. In addition, the exercise group participated in a twice- weekly physiotherapist-led exercise group class for 6 weeks; [95% CIs calculated by the CAP Editor.] participants performed a 50-minute exercise circuit of 12 stations of range of motion exercises for the shoulder and thoracic spine. The Nicholas Taylor individual physiotherapy group received two individual sessions Section Editor, Journal of Physiotherapy each week for 6 weeks from a musculoskeletal physiotherapist. Individual treatment could include manual techniques, massage, http://dx.doi.org/10.1016/j.jphys.2014.08.011 stretching and heat. Outcome measures: The primary outcome was Commentary Frozen shoulder causes pain, physical impairments and and colleagues refer to the ‘pain-predominant’ and ‘stiffness- potential anxiety.1 Primary care clinicians should be able to predominant’ classification.3 Loosely interpreting this, they exclud- recognise frozen shoulder, provide reassurance and initiate a ed patients with symptoms of less than 3 months in order to treatment pathway that is informed by efficacy and cost. Individual minimise those in the ‘early pain predominant phase’. This is physiotherapy, group physiotherapy and home exercises have encouraging, but it is uncertain as to how a stricter definition of different cost implications, but there have been no previous head- stiffness-predominant (patient-reported predominance of stiffness to-head comparisons of their efficacy for frozen shoulder. over pain3) would influence the size and direction of effect. Also uncertain are the implications for intra-articular corticosteroid Commendably, Russell and colleagues conducted the first injection, which appears to be efficacious – especially when randomised controlled trial of all three of these approaches. Based combined with physiotherapy.4 Future research should integrate on a patient-reported outcome measure with a known minimal the aspect of intra-articular injection with Russell and colleagues’ clinically important difference, group therapy was found to be important contribution. statistically and clinically superior to home exercises. The Hospital Anxiety and Disability Scale anxiety scores were significantly lower Nigel Hanchard in the physiotherapy groups; this possibly reflects contact with a Health and Social Care Institute, Teesside University, United Kingdom knowledgeable therapist and fellow patients in the group. This contrasts with the likely experience of many people with frozen References shoulder. Alarmingly, 83% of the patients referred for this study had been labelled with false-positive diagnoses. If false negatives are 1. Jones S, et al. BMJ Open. 2013;3:e003452. nearly as prevalent, then isolation, confusion and anxiety may be the 2. Hanchard NCA, et al. Physiother. 2011;97:115–125. norm. 3. Hanchard NCA, et al. Physiother. 2012;98:117–120. 4. Maund E, et al. Health Technol Assess. 2012;16:1–243. While there are various systems for classifying the phases of frozen shoulder, its phasic nature is accepted by clinicians. Clinicians http://dx.doi.org/10.1016/j.jphys.2014.08.017 modify treatment accordingly,2 yet researchers typically disregard this clinical wisdom with implications for applicability. Russell 1836-9553/ß 2014 Australian Physiotherapy Association. Published by Elsevier B.V. All rights reserved.

Journal of Physiotherapy 60 (2014) 238 Journal of PHYSIOTHERAPY journal homepage: www.elsevier.com/locate/jphys Appraisal Critically Appraised Papers Multimodal physiotherapy may be no better than sham treatment for people with hip osteoarthritis Synopsis Summary of: Bennell KL, Egerton T, Martin J, Abbott JH, Metcalf B, ultrasound and inert gel lightly applied to the hip region. McManus F, et al. Effect of physical therapy on pain and function in Participants in both groups attended 10 individual physiotherapy patients with hip osteoarthritis: a randomized clinical trial. JAMA. sessions over 12 weeks; twice in the first week, once a week for 2014;311(19):1987-1997. 6 weeks, then approximately once every 2 weeks. Outcome measures: Primary outcomes were pain on a 100-mm visual Question: Does a multimodal physiotherapy program lead to analogue scale and physical function, measured on the Western greater improvements in pain and physical function than sham Ontario and McMaster Universities Index (0 to 68), assessed by a physiotherapy among people with hip osteoarthritis? Design: blinded assessor at weeks 13 and 36. Results: A total of 96 patients A randomised, controlled trial with concealed allocation and (94%) completed the 13-week assessment; there were no 24-week follow-up. Setting: Nine private physiotherapy clinics in statistically significant differences between the two groups. The Melbourne, Australia. Participants: Men and women aged 50 years mean difference in improvement for pain was 6.9 mm (95% CI À3.9 or older with hip osteoarthritis, according to the American College to 17.7), and 1.4 units (95% CI À3.8 to 6.5) for function, both of Rheumatology classification criteria, with an average pain favouring the sham treatment. Significantly more participants intensity during the past week of at least 40 on a 100-mm visual reported adverse events in the active group than in the sham analogue scale, and at least moderate difficulty with performing treatment group (41 versus 14%, p = 0.003). No significant daily activities. Key exclusion criteria included: lower limb between-group differences in change were observed 24 weeks surgery; physiotherapy, chiropractic treatment or prescribed after the intervention. Conclusion: A multimodal physiotherapy exercises in past 6 months; more than 30 minutes daily walking; program did not result in greater improvement in pain and and regular exercise more than once a week. Randomisation function than sham treatment for people with symptomatic hip allocated 49 people to the physiotherapy program and 53 to the osteoarthritis. sham treatment. Interventions: The physiotherapy program was semi-standardised with core components typical of clinical T$F[RSNDAIME]Ka˚ re BirgerDFIRST]NE.[$AM ANH[RMU]S$DTE agena and D]IRTMSNFME[A$ argrethAMESTD.$[F]RNI TGRNU]SMD$[AE rotleb practice (manual therapy; spine mobilisation; deep tissue aNational Advisory Unit on Rehabilitation in Rheumatology, massage; muscle stretches; home exercises performed four times/week; education and advice; and provision of a walking Diakonhjemmet Hospital, Oslo stick, if appropriate), plus optional techniques and exercises bDivision of Surgery and Clinical Neuroscience, depending on assessment findings. Participants were instructed to perform unsupervised home exercises three times a week during Oslo University Hospital, Norway the 6-month follow-up. The sham intervention included inactive http://dx.doi.org/10.1016/j.jphys.2014.08.013 Commentary A clear conclusion can be drawn from this high-quality strength, aerobic capacity and/or range of motion.4 However, the randomised controlled trial: this multimodal physiotherapy pro- exercise part in the study of Bennell et al was mainly delivered as gram did not give additional clinical benefit over a placebo- home exercises, which limited the control of performance and controlled sham intervention, and was associated with relatively adherence to the program; this provides a possible explanation for frequent, but mild, adverse effects. the observed lack of clinical benefit. The need for individually tailored, supervised exercise programs of adequate dosage alongside Physiotherapy is typically delivered as a comprehensive package education is still the current recommendation for people with hip of care; therefore, Bennell et al aimed to test the hypothesis that a osteoarthritis.4,5 multimodal program could have beneficial effects on pain and function. However, focusing on several elements within the time [MDIFHS]TRAE$N anneTD[FIRS$.MANE] NS$T]RUDEAMD[ agfinrud, TDMNSRIEARF]$[ ikke HeleneFIRS[TD$NEAM.] []ANMED$RUMTS oe and ]SNE[RANTD$MFI ina]TIFERAM.[NSD$ ØMNRD$[S]TEUA stera˚ s limit of a treatment session may result in an ineffective dosage of Department of Rheumatology, Diakonhjemmet Hospital, Oslo, Norway each element. Mixed programs may even raise the risk of adverse interaction effects.1 Thus, one implication of the findings of this trial References is that physiotherapists should select modalities most appropriate for each individual patient rather than multimodal programs. 1. Abbott JH, et al. Osteoarthritis Cartilage. 2013;21(4):525–534. 2. Pisters MF, et al. Arthritis Care Res. 62(8):1087–1094. Supervision of exercise sessions increases the adherence to 3. Garber CE, et al. Med Sci Sports Exerc. 2011;43(7):1334–1359. exercise programs, and better adherence has been shown to 4. Fernandes L, et al. Ann Rheum Dis. 2013;72(7):1125–1135. improve long-term results in people with osteoarthritis.2 The 5. Fransen M, et al. Cochrane Database Syst Rev. 2014;4:CD007912. effects of exercise programs are dependent on dosage and progression,3 and recommendations underline the importance of http://dx.doi.org/10.1016/j.jphys.2014.08.016 meeting the minimal requirements to improve or maintain muscle 1836-9553/ß 2014 Australian Physiotherapy Association. Published by Elsevier B.V. All rights reserved.