Article Text

Download PDFPDF

Does telephone-delivered exercise advice and support by physiotherapists improve pain and/or function in people with knee osteoarthritis? Telecare randomised controlled trial
  1. Rana S Hinman1,
  2. Penny K Campbell1,
  3. Belinda J Lawford1,
  4. Andrew M Briggs2,
  5. Janette Gale3,
  6. Caroline Bills3,
  7. Jessica Kasza4,
  8. Anthony Harris5,
  9. Simon D French1,6,
  10. Stephen J Bunker7,
  11. Andrew Forbes4,
  12. Kim L Bennell1
  1. 1 Centre for Health, Exercise and Sports Medicine, The University of Melbourne, Melbourne, Victoria, Australia
  2. 2 School of Physiotherapy and Exercise Science, Curtin University, Perth, Western Australia, Australia
  3. 3 HealthChange Australia, Melbourne, Victoria, Australia
  4. 4 School of Public Health and Preventative Medicine, Monash University, Melbourne, Victoria, Australia
  5. 5 Centre for Health Economics, Monash University, Melbourne, Victoria, Australia
  6. 6 Department of Chiropractic, Macquarie University, Sydney, New South Wales, Australia
  7. 7 Medibank Private, Melbourne, Victoria, Australia
  1. Correspondence to Professor Rana S Hinman, Centre for Health, Exercise and Sports Medicine, The University of Melbourne, Melbourne, VIC 3010, Australia; ranash{at}


Objective Evaluate a physiotherapist-led telephone-delivered exercise advice and support intervention for people with knee osteoarthritis.

Methods Participant-blinded, assessor-blinded randomised controlled trial. 175 people were randomly allocated to (1) existing telephone service (≥1 nurse consultation for self-management advice) or (2) exercise advice and support (5–10 consultations with a physiotherapist trained in behaviour change for a personalised strengthening and physical activity programme) plus the existing service. Primary outcomes were overall knee pain (Numerical Rating Scale, range 0–10) and physical function (Western Ontario and McMaster Universities Osteoarthritis Index, range 0–68) at 6 months. Secondary outcomes, cost-effectiveness and 12-month follow-up were included.

Results 165 (94%) and 158 (90%) participants were retained at 6 and 12 months, respectively. At 6 months, exercise advice and support resulted in greater improvement in function (mean difference 4.7 (95% CI 1.0 to 8.4)), but not overall pain (0.7, 0.0 to 1.4). Eight of 14 secondary outcomes favoured exercise advice and support at 6 months, including pain on daily activities, walking pain, pain self-efficacy, global improvements across multiple domains (overall improvement, improved pain, improved function and improved physical activity) and satisfaction. By 12 months, most outcomes were similar between groups. Exercise advice and support cost $A514/participant and did not save other health service resources.

Conclusion Telephone-delivered physiotherapist-led exercise advice and support modestly improved physical function but not the co-primary outcome of knee pain at 6 months. Functional benefits were not sustained at 12 months. The clinical significance of this effect is uncertain.

Trial registration number Australian New Zealand Clinical Trials Registry (#12616000054415).

  • osteoarthritis
  • physiotherapy
  • knee

Statistics from


Hip and knee osteoarthritis (OA) is the 12th highest contributor to disability in adults aged 50–69 years.1 Knee OA affects 10%–25% of people aged 60 years and over,2 causing pain, physical dysfunction and reduced quality of life, and imposing enormous economic burden.3 4 Guidelines emphasise non-drug non-surgical interventions,5–8 focusing on active rehabilitation. Advice and information for self-management, exercise and weight control are core management, with drugs, injections and manual therapy considered adjunctive.7

Exercise is advocated for all people with knee OA.5–8 Beneficial effects include reduced pain, and improved function and quality of life.9 10 However, uptake of exercise by people consulting a doctor in primary care for OA is poor.11 Rates of drug prescription are higher than lifestyle management for knee OA.12 Although general practitioners believe self-management and lifestyle counselling are critically important,13 they feel unskilled in such roles.14 15 Compounding the problem, most people with knee OA do not meet physical activity guidelines for good health.16

Difficulty accessing appropriately qualified health professionals, such as physiotherapists, is a barrier to exercise participation in people with OA.17 Physiotherapy is traditionally provided face to face, however, for many people, access is limited by geography.18 This is particularly problematic in regional and remote areas, even though these populations suffer from arthritis more than urban peers.19 Even in urban areas, travelling to appointments can be difficult for people with pain and mobility restrictions. Telephone services may increase access to physiotherapists, yet no trial has evaluated telephone-delivered physiotherapist-led exercise advice for people with knee OA without face-to-face care. The primary aim of this study was to determine the effectiveness of adding exercise advice and support by physiotherapists to an existing nurse-led musculoskeletal telephone service on pain and function in adults with knee OA. Secondary aims were to evaluate clinical effectiveness of the intervention on self-efficacy, physical activity, kinesiophobia, satisfaction, quality of life and global perceived change and cost-effectiveness.



Pragmatic superiority parallel-group randomised (1:1) controlled trial, prospectively registered, approved by the institutional ethics committee and reported according to Consolidated Standards of Reporting Trials recommendations.20 The protocol21 and qualitative evaluations22 23 are published.


Community volunteers were recruited across Australia (March 2016–October 2017) via advertisements (consumer organisations, social media, clubs, clinics, radio and newspapers) and our volunteer database. Participants were screened via electronic survey and over the phone. Inclusion criteria were OA clinical criteria7 (1) ≥45 years; (2) activity-related knee pain and (3) morning knee stiffness ≤30 min. Additional criteria were knee pain (previous week) of ≥4 on Numerical Rating Scale (NRS; 0=no pain; 10=worst pain possible) and pain for ≥3 months. Exclusion criteria were (1) waiting list for/planning knee/hip surgery (next 12 months); (2) arthroplasty; (3) knee arthroscopy (prior 12 months); (4) undertaking exercise prescribed by a clinician; (5) unable to speak/read English; (6) inflammatory arthritis; (7) lower limb neurological condition or (8) uncontrolled cardiovascular condition. Medical clearance was required for anyone (1) reporting a fall (past 12 months); (2) housebound due to immobility; or (3) failing exercise screening24 due to increased risks of adverse events from unsupervised exercise at home.

Randomisation, concealment and blinding

The randomisation schedule (random permuted blocks size 6–12, stratified by gender) was prepared by a biostatistician. Participants allocated exercise advice and support were randomly allocated a physiotherapist (random permuted blocks of size 16). Allocation was concealed by password-protected software. A researcher not involved in recruitment or outcome assessment accessed the schedule. Limited disclosure was used to blind participants, who were also the assessors (outcomes participant reported). Participants were informed that the trial was comparing two different forms of telephone-delivered self-management advice. We did not disclose intervention components, health professionals involved or the hypothesis. Nurses (who delivered care to both control and intervention groups) were blinded but physiotherapists (who delivered care to the intervention group only) were not.

Existing service

The control group received the existing service from the Musculoskeletal Help Line (Musculoskeletal Australia, Victoria, Australia), delivered by one of four nurses. It provides information about OA; treatments and self-management strategies; community resources; assistance navigating services; emotional support and care escalation when needed. Participants received one call from a nurse, with additional calls if required.

Exercise advice and support

In addition to the existing service, this group had 5–10 consultations with a physiotherapist over 6 months. The intervention is described elsewhere.21 Briefly, eight physiotherapists were trained (HealthChange Methodology (HealthChange Australia)) to deliver person-centred exercise-based care using theoretically informed behavioural change techniques. Details about training21 and its outcomes25 26 are published.

Physiotherapists helped devise an action plan including home-based strengthening exercise and physical activity. For strengthening, physiotherapists chose from a list of 14 exercises (online supplementary appendix table 1), aiming for 5–6 exercises performed three times per week. Physiotherapists aimed to prescribe a programme and dosage that was ‘hard’ to ‘very hard’ to perform to stimulate strength gains that would translate to improved function. Physiotherapists assisted participants to develop a physical activity plan aimed at increasing physical activity. To support the intervention, participants were provided an information folder, exercise bands, and access to a bespoke website for exercise videos.

Supplemental material

Fidelity was assessed via number and duration of consultations, and review of consultation notes. The exercise advice and support group recorded adherence to strengthening and physical activity using 11-point NRSs (0=strongly disagree; 10=strongly agree). Physiotherapists rated participant adherence using an 11-point NRS (0=not at all; 10=completely as instructed).


Participants completed outcomes at baseline, 6 and 12 months (electronically or paper based). The primary time point was 6 months. To minimise missing data, participants were provided a $A50 gift voucher for completing all questionnaires. Primary outcomes were valid and reliable measures recommended for OA trials.27 Overall average knee pain (previous week) was measured via NRS (0=no pain; 10=worst pain possible). Difficulty with physical functioning (previous 48 hours) was measured by the Western Ontario and McMaster Universities Osteoarthritis Index28 function subscale (WOMAC, range 0–68 (maximum dysfunction)).

Secondary outcomes included: (1) knee pain on daily activities (WOMAC pain subscale; range 0–20 (higher pain)); (2) knee pain on walking (NRS; range 0–10 (higher pain)); self-efficacy for (3) pain and (4) function (Arthritis Self-Efficacy Scale29; range 1–10 (higher self-efficacy)); (5) fear of movement (Brief Fear of Movement Scale for OA30; range 0–24 (lower fear)); (6) physical activity (Physical Activity Scale for the Elderly31; range (0 to >400 (more physical activity)); behavioural determinants of exercise via (7) Barriers to Physical Activity Scale32 (range 0–92 (greater perceived barriers)) and (8) Benefits of Physical Activity Scale32 (range 14–70 (greater perceived benefits)); (9) health-related quality of life (Assessment of Quality of Life (AQoL) instrument33 (AQoL-8D); range −0.04 to 1.00 (better quality of life)); (10) global changes (overall; pain; function) via 7-point scales (‘much worse’ to ‘much better’), including change in physical activity (‘much less’ to ‘much more’) and (11) satisfaction (‘extremely unsatisfied’ to ‘extremely satisfied’).

Cost-effectiveness was assessed as additional cost per quality-adjusted life years gained at 6 and 12 months (online supplementary appendix 1). Adverse events and healthcare usage were recorded via custom surveys.

Sample size

We aimed to detect an effect size of 0.5 on primary outcomes at 6 months, consistent with evidence from a Cochrane Review.9 To account for potential clustering, we assumed physiotherapists would treat the same number of participants and an intracluster correlation of 0.05.34 We assumed a between-participant SD of 2.2 for pain and 11.6 for function, and a baseline to 6-month correlation in scores of 0.29 and 0.51, respectively.35 With analysis of covariance adjusted for baseline scores and clustering, we required 70 people per group for pain and 54 per group for function with 80% power. Assuming 20% attrition, we recruited 175 people. This sample size also allowed detection of minimum clinically important between-group differences of 1.8 pain units36 and six function units37 with 90% power.

Statistical methods

Analyses were performed using Stata (V.15) on an intention-to-treat basis. To account for physiotherapist clustering, participants assigned the existing service were assumed to be in clusters of size 1. For continuous outcomes, mean (95% CI) difference in change (baseline minus follow-up) was estimated using mixed-effects linear regression models adjusting for baseline, gender, group and time, and an interaction between group and time. A random intercept for participant accounted for multiple measurements. Clustering by physiotherapist was accounted for by including a random intercept for participants allocated exercise advice and support only, with separate residual variances for the two groups. For missing outcomes, multiple imputation with chained equations and predictive mean matching drawing from the three nearest neighbours for continuous outcomes, and using chained logistic regression models for binary outcomes, was applied. Imputations were generated separately for each group. Estimates from 20 imputed data sets were combined using Rubin’s rules.38 For the global changes, participants ‘moderately better’ or ‘much better’ (‘moderately more’ or ‘much more’) were classified as improved. Participants ‘moderately satisfied’ or ‘extremely satisfied’ were classified as satisfied. Binary outcomes were compared with logistic regression models fitted using generalised estimating equations. Models included terms for gender, time and group, with an interaction between time and group. The approach for binary outcomes differed from the protocol21 to allow estimation of more useful population-averaged effects.

We assessed intervention effects under scenarios of hypothetical full adherence (‘complier average causal effects’) using a two-stage least squares instrumental variables approach, first defining ‘full adherence’ as ≥5 consultations (minimum number dictated by our intervention protocol), then treating number of consultations as continuous to estimate a dose–response relationship.39 40 This involved fitting two models jointly: a linear regression model for difference in change in outcome adjusted for baseline, stratifying variables and number of consultations (first dichotomised then included as a continuous variable), and a linear regression model for number of consultations adjusted for group, baseline and stratifying variables. Existing service participants were assumed to have zero physiotherapist consultations (monotonicity assumption). When consultations were treated continuously, difference in outcome between people with zero consultations and those with five consultations was calculated. In further analyses, we adjusted for physiotherapist visits (outside the trial) in regression models. All hypotheses were two sided.

Patient and public involvement

No patients were involved in study design. Representatives of a consumer advocacy organisation (Arthritis and Osteoporosis Victoria, now known as Musculoskeletal Australia), a private health insurer (Medibank), the national physiotherapy professional body (Australian Physiotherapy Association) and industry (HealthChange Australia) helped design the trial and obtain funding. The existing service was delivered by nurses on the Musculoskeletal Help Line run by Musculoskeletal Australia.


We enrolled 175 participants, with 165 (94%) and 158 (90%) completing both primary outcomes at 6 and 12 months, respectively (figure 1). Groups were similar at baseline, except more of the existing service group tended to report other joint problems (table 1). Participants who withdrew were generally comparable to those who did not (online supplementary appendix table 2).

Figure 1

Flow of participants through the trial.

Table 1

Baseline characteristics of participants, by intervention group, reported as mean (SD) unless otherwise stated

Three (3%) participants from each group did not consult with a nurse. Topics discussed with nurses were similar across groups (online supplementary appendix table 3). Three (3%) participants allocated exercise advice and support did not consult with a physiotherapist. Each physiotherapist consulted with a mean (SD) of 11 (2) participants (online supplementary appendix table 4 describes initial call), and participants had 6 (2) physiotherapist consultations. Seventy-six (87%) participants had ≥5 consultations. Mean (SD) call duration was 47 (6) and 20 (1) min for initial and follow-ups, respectively. Adherence (online supplementary appendix table 5) exceeded 7 out of 10 units on all NRS’s at 6 months. Adverse events (online supplementary appendix table 6) were uncommon but more frequent with exercise advice and support. Cointerventions were similar across groups (online supplementary appendix tables 6 and 7).

Primary outcomes

Both groups demonstrated clinically important improvements in pain and function at 6 and 12 months (tables 2 and 3). At 6 months, evidence of a between-group difference in change in function favoured exercise advice and support (4.7 units (95% CI 1.0 to 8.4)) but there was no evidence of a difference in pain (0.7 units (0.0 to 1.4)). There were no between-group differences at 12 months.

Table 2

Mean (SD) scores on continuous outcome measures across time, by treatment group

Table 3

Change within groups, and difference in change between groups (adjusted for baseline value of outcome, gender, and physiotherapist clustering), for continuous outcomes, using multiply imputed data

Sensitivity analyses of complete case data (online supplementary appendix table 8) yielded similar results, except there was some evidence of a small difference in pain at 6 months favouring exercise advice and support. Varying scenarios of hypothetical adherence gave similar results (online supplementary appendix table 9). When adjusting for physiotherapist visits outside the trial (online supplementary appendix table 10), between-group differences in adherent participants were evident and favoured exercise advice and support.

Secondary outcomes

At 6 months (tables 2 and 3), exercise advice and support demonstrated a greater reduction in pain on daily activities (1.2 units (95% CI 0.2 to 2.1)) and pain on walking (1.0, 0.1 to 1.8), and a greater increase in pain self-efficacy (−1.2, −1.8 to –0.6). Of these, only pain self-efficacy was different at 12 months, favouring exercise advice and support (−1.0, −1.7 to –0.4). At 12 months, exercise advice and support participants also reported a greater increase in physical activity (−31, −60 to –1)). Sensitivity analyses of complete case data (online supplementary appendix table 8) yielded similar results.

At 6 months, more exercise advice and support participants reported improvement overall (risk difference 39.1% (95% CI 25.5% to 52.7%) table 4), in pain (41.3%, 95% CI 28.5% to 54.2%)), in function (37.2%, 95% CI 23.2% to 51.2%) and in physical activity levels (32.1%, 95% CI 20.3% to 43.8%). More of this group were satisfied (52.3%, 95% CI 40.3% to 64.2%). At 12 months, more of the exercise advice and support group were improved in function (22.4%, 95% CI 9.3% to 35.6%) and physical activity (17.9%, 95% CI 4.3% to 31.4%), and were satisfied (37.6%, 95% CI 23.3% to 51.9%). Sensitivity analyses (online supplementary appendix table 10) yielded similar results. The direct cost of exercise advice and support was $A514 per participant (online supplementary appendix 2), with no evidence that it saved other health service resources.

Table 4

Number (percentage) of participants reporting global improvement, and satisfaction, with exercise advice and support relative to the existing service (adjusted for physiotherapist clustering and gender), using multiply imputed data


Telephone-delivered physiotherapist-led exercise advice and support resulted in modest improvements in our coprimary functional outcome at 6 months, but not pain. The clinical relevance of these findings is uncertain. Although the between-group improvement in function of 4.7 units (95% CI 1.0 to 8.4) favouring intervention is less than the 6 units advocated as clinically meaningful,37 the CIs included this difference. For both primary outcomes, larger intervention effects (and narrower CIs) emerged with complete case data and sensitivity analyses, including some evidence of a difference in our coprimary pain outcome favouring exercise advice and support. However, this difference is probably clinically irrelevant.36 In contrast, a secondary pain outcome (pain on walking) demonstrated a between-group effect (1.0 (0.1 to 1.8)) in the primary analysis, and the minimum clinically important difference of 1.8 units was within CIs. Thus, although we cannot conclude that the intervention effects on pain and function are clinically significant, we believe further research is warranted, particularly given consistent findings across 8 of 14 secondary outcomes favouring intervention.

How our findings compare with those of other studies

Telephone interventions for knee OA have been studied previously.41 Only two studies have involved physiotherapist consultations, evaluating telephone calls to follow-up in-person physiotherapist consultation(s).42 43 Our trial is unique because our physiotherapists prescribed strengthening and physical activity entirely over the telephone. We added our intervention to an existing musculoskeletal-specific telephone service provided by a consumer advocacy organisation. In contrast, another recent Australian randomised controlled trial evaluated the efficacy of a more generic government-funded telephone service.44 The intervention was coaching to support lifestyle improvements involving diet, physical activity, healthy weight maintenance and smoking cessation. Up to 10 coaching calls from allied health professionals were provided over 6 months. Findings showed no intervention effect on knee pain or weight compared with usual care in overweight/obese people with knee OA. In contrast to our study where 87% of people participated in at least five physiotherapist consultations, participant adherence in this trial was poor with only 51% continuing beyond the second call. Similarly, a trial in a USA veteran’s affairs medical centre evaluated the efficacy of telephone-based OA self-management in people with hip or knee OA.45 A health educator called participants monthly for 12 months to discuss educational topics, goals and action plans. There was no difference in pain compared with usual care, but a small pain reduction was observed compared with a generic (non-OA-specific) education control. No differences were observed in function. Two subsequent trials in the same veteran’s affairs setting by the same team combined a similar telephone-based patient intervention with patient-specific self-management recommendations delivered at the point of care by the primary care provider.46 47 Meta-analysis of these three trials showed no effect of the intervention on pain compared with usual care.41

The burden of knee OA is increasing.2 There are fewer physiotherapy services in regional and remote places than in major cities,48 and older people everywhere experience transport difficulties to attend health services.49 Telephone services provide reach and may address underutilisation of behavioural approaches to manage OA.11 12 16 Qualitative research showed our participants valued the convenience and accessibility of telephone-delivered care,23 and participants and physiotherapists believed a strong therapeutic alliance was developed over the phone.50 Generic telephone services promoting a healthy lifestyle are not effective for OA symptoms,44 nor is broad-based OA self-management education.41 Our trial suggests that a service focused on individualised progressive exercise and physical activity, provided by physiotherapists trained in behavioural change, offers the most promise.

Strengths and limitations

Study strengths include clinical inclusion criteria that maximise external validity. We enrolled people from all over Australia, spanning major cities to very remote areas, and aged from 45 to 85 years. Consultations focused on best-evidence physiotherapy that was personally tailored and supported by behavioural change principles. Although this resulted in a complex intervention requiring specialised physiotherapist training,25 26 this was worthwhile given the high satisfaction and excellent adherence observed. The trial was robust, with participant and assessor blinding, high retention and accompanied by cost-effectiveness analyses. Limitations include unblinded physiotherapists, exclusion of non-English-speaking participants and use of a proxy adherence measure (number of consultations) in sensitivity analyses. We also do not know whether participants performed their exercises at the desired intensity, nor did we prevent participants seeking care outside the trial. When we accounted for physiotherapist consultations outside the trial, between-group treatment effects among those who would have adhered to their assigned group became slightly more pronounced.

Our findings have implications for both clinical practice and further research. Clinicians can consider implementing telephone-delivered models of exercise management for people with chronic knee pain who are unable (or unwilling) to attend face-to-face consultations with a physiotherapist and where physical dysfunction is a driver for seeking care. Although not evaluated in this study, a ‘blended’ approach combining face-to-face consultations with telephone-delivered care may also appeal to some patients and serve to increase utilisation of exercise-based OA care. In regional and remote geographical areas, where health services are limited and people are geographically dispersed, establishment of centralised telephone-delivered service models (from a regional physiotherapy or medical clinic ‘hub’) may help deploy physiotherapy services on broad scale. Further trials are needed to evaluate the effectiveness of telephone-delivered and blended models of physiotherapy care relative to usual care for people living in regional and remote areas, including research to better understand the barriers to implementing such services.


Telephone-delivered physiotherapist-led exercise advice and support modestly improved physical function but not the coprimary outcome of knee pain at 6 months. Functional benefits were not sustained at 12 months. Although clinical significance is uncertain, findings support further trials of telephone-delivered service models.

What are the findings?

  • Incorporating 5–10 sessions of physiotherapist-led exercise advice and behavioural change support into an existing national nurse-led musculoskeletal telephone service provided some modest benefits at 6 months, including improved physical function, improvements in some measures of pain and physical activity and improved self-efficacy for pain.

  • High satisfaction rates with the intervention show that telephone-delivered exercise and advice from physiotherapists is an acceptable mode of service delivery to people with knee osteoarthritis.

How might it impact on clinical practice in the future?

  • Burden of knee osteoarthritis (OA) is increasing due to population ageing and rising obesity rates, and there are inequities in availability of physiotherapy services in regional and remote locations compared with major cities.

  • Telephone services permit widespread reach and may help address underutilisation of behavioural approaches to manage knee OA. Clinicians should consider implementing telephone-delivered models of exercise management for people with chronic knee pain who are unable to attend face-to-face consultations with a physiotherapist and where physical dysfunction is a driver for seeking care.


We thank the participants who volunteered for this trial, and the nurses (Anne Lloyde, Clare Patterson, Margaret Ricardo and Leanne Mill) and physiotherapists (Christopher Snell, Justin Edwards, Leigh Iacovangelo, Mary McMahon, Nathan Wilson, Rebecca Manson, Sophie Heywood, Viktoria Molloy) involved in delivering the interventions.



  • Contributors RSH, KLB, SDF, AF, AH, AMB, SJB and JG conceived the study and obtained study funding and in-kind support; RSH, KLB, CB and JG designed the physiotherapy group intervention; JG designed the behavioural change methodology that the physiotherapists were trained in; CB and RSH trained the physiotherapists; PKC recruited participants and coordinated the trial; BJL led the qualitative studies that aided interpretation of trial findings; JK was responsible for statistical analyses; AH was responsible for cost-effective analyses; RSH drafted the manuscript and all authors read and approved the final version for submission.

  • Funding This trial was funded by the National Health and Medical Research Council (Partnership Project #1112133 and Centre of Research Excellence (#1079078)) and the Medibank Better Health Foundation, with in-kind support from Musculoskeletal Australia (formerly Arthritis and Osteoporosis Victoria), HealthChange Australia and the Australian Physiotherapy Association. RSH is supported by National Health and Medical Research Council Fellowship (#1154217). KLB is supported by a National Health and Medical Research Council Fellowship (#1058440).

  • Competing interests JG owns HealthChange Australia, which trained the physiotherapists in behavioural change methodology and employs CB as a training facilitator. AMB was a salaried employee of Arthritis and Osteoporosis Victoria (now known as Musculoskeletal Australia) at the time this trial was designed and grant funding awarded. SJB is employed as Medibank’s Clinical Research Advisor.

  • Patient consent for publication Not required.

  • Ethics approval Obtained from the human research ethics committee of the University of Melbourne (HREC #1544432).

  • Provenance and peer review Not commissioned; externally peer reviewed.

  • Data availability statement All data relevant to the study are included in the article or uploaded as online supplementary information.

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.