Article Text
PDF
Abstract
Objective This study aims to evaluate the effect of a performance-focused swimming programme on motor function in previously untrained adolescents with cerebral palsy and high support needs (CPHSN) and to determine whether the motor decline typical of adolescents with CPHSN occurred in these swimmers.
Methods A Multiple-Baseline, Single-Case Experimental Design (MB-SCED) study comprising five phases and a 30-month follow-up was conducted. Participants were two males and one female, all aged 15 years, untrained and with CPHSN. The intervention was a 46-month swimming training programme, focused exclusively on improving performance. Outcomes were swim performance (velocity); training load (rating of perceived exertion min/week; swim distance/week) and Gross Motor Function Measure-66-Item Set (GMFM-66). MB-SCED data were analysed using interrupted time-series simulation analysis. Motor function over 46 months was modelled (generalised additive model) using GMFM-66 scores and compared with a model of predicted motor decline.
Results Improvements in GMFM-66 scores in response to training were significant (p<0.001), and two periods of training withdrawal each resulted in significant motor decline (p≤0.001). Participant motor function remained above baseline levels for the study duration, and, importantly, participants did not experience the motor decline typical of other adolescents with CPHSN. Weekly training volumes were also commensurate with WHO recommended physical activity levels.
Conclusions Results suggest that adolescents with CPHSN who meet physical activity guidelines through participation in competitive swimming may prevent motor decline. However, this population is clinically complex, and in order to permit safe, effective participation in competitive sport, priority should be placed on the development of programmes delivered by skilled multiprofessional teams.
Trial registration number ACTRN12616000326493.
- Para-Athletes
- Rehabilitation
Data availability statement
Data are available on reasonable request.
Statistics from Altmetric.com
WHAT IS ALREADY KNOWN ON THIS TOPIC
Compared with ambulant people with cerebral palsy, gross motor function declines in non-ambulant people with cerebral palsy and high support needs (CPHSN). These patients are also less physically active, and it is plausible that relative inactivity contributes to motor decline; however, this premise has not been investigated.
WHAT THIS STUDY ADDS
This study demonstrated that previously inactive adolescents with CPHSN who undertook performance-focused swimming training with multiprofessional guidance over 46 months improved sports performance and maintained gross motor function during a life stage when population-based modelling predicted gross motor decline.
HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY
This study introduces the novel concept of ‘Para Sport as Medicine’ and suggests that performance-focused sports training programmes delivered by multiprofessional teams may be an effective means of preventing motor decline among people with CPHSN, as well as conferring a range of psychosocial and well-being benefits.
Background
Cerebral palsy (CP) is the most common neuromotor disorder affecting children and non-progression of the underlying neuropathology is a defining feature of CP.1 In children with CP who are ambulant—Gross Motor Function Classification System (GMFCS) levels I and II—gross motor function improves from birth to approximately 7–9 years of age and then plateaus. However, among children with CP who are non-ambulant, have high support needs and are classified as GMFCS levels IV and V (CPHSN), early developmental gains are generally followed by a decline in motor function throughout adolescence and into early adulthood.2 The underlying causes of this decline are poorly understood although it has been suggested that reduced access to neurological care, the development of new neurological conditions3 or poor management of hypertonia during periods of growth4 may contribute.
We posit that insufficient habitual physical activity may contribute to motor function decline in adolescents with CPHSN. Specifically, the majority of children and adolescents with CP are insufficiently active for good health5 and, compared with those who are ambulant, those with CPHSN are more sedentary and less physically active.6 It is plausible that the relatively greater gross motor decline of adolescents with CPHSN is caused, at least in part, by their relatively low levels of habitual physical activity.
Unfortunately, people with CPHSN are grossly under-represented in exercise training studies. A recent review identified that only 3% of participants were either GMFCS level IV or V.7 While evidence indicates that physical activity can improve gross motor function in people with CP, the effect has only been demonstrated in children at GMFCS levels I and II,8 and as they do not experience the motor decline associated with GMFCS IV and V, the generalisability of the finding is not known. Additionally, interventions have been brief (8–12 weeks) with limited follow-up, so the extent to which improvements are maintained is not known.8
Competitive Para swimming is a type of physical activity open to people with CPHSN, and it provides them with critical avenues to engage in competitive swimming, one of Australia’s most popular and culturally significant sports, particularly among children and adolescents.9 Para swimmers devote the majority of their sports participation time to performance-focused training, defined as training that is planned and undertaken for the primary purpose of maximising sports performance.9 Because of its focus on performance enhancement—primarily maximising swimming velocity10—performance-focused swimming training is clearly distinct from other conventional aquatic therapies (including hydrotherapy) which explicitly focus on therapeutic outcomes and which have been shown to be effective.11 However, while performance-focused training does not have therapeutic goals, personal testimony from experienced Para swimmers with CP indicates they attribute large, meaningful improvements in physical function to such training.9 However, to date, the veracity of this testimony has not been evaluated in swimmers with CP, including swimmers with CPHSN.
Investigating whether performance-focused swimming training prevents motor decline in people with CPHSN presents considerable challenges. Severe functional limitations increase the time cost of participation for people with CPHSN by 8–13 times,12 significantly increasing research costs. Further, the heterogeneity that characterises CP is greatest in this population, who are often affected by a greater number of comorbidities that are more severe13 and many of these comorbidities (eg, seizure disorders, eating and drinking difficulties and pain) act as independent prognostic variables in exercise training trials. In group-based research designs such as randomised controlled trials (RCTs), the result of this heterogeneity is predictable, systematic between-participant differences in exercise training responses which act to amplify noise and threaten internal validity.
Single-case experimental research designs (SCEDs), where each participant acts as their own control, account for logistical and research design challenges in people with CPHSN and offer a methodologically robust alternative to RCTs.14 The SCED generates high-level evidence (equivalent to RCTs15) using small samples, permits tailoring of the intervention to each participant and produces individual outcomes—the SCED is one of the few designs in which it is possible to detect if, when and to what extent each participant responds to the intervention.16 These design features are particularly advantageous for studies in people with CPHSN: a relatively small, heterogeneous population who require tailored interventions and support which meets their personal needs, and who have been largely excluded from the literature to date.10
Therefore, this study employed a single-case experimental design to address two primary aims: to evaluate the effect of a performance-focused swimming programme on gross motor functioning in previously untrained, inactive adolescents with CPHSN and to determine whether the motor decline typical of adolescents with CPHSN occurred in swimmers who trained and competed regularly for one Paralympic cycle over a 46-month period.
Methods
The ParaSTART (Sports Training And Research Team) programme was established to facilitate research presented in this manuscript and other projects. ‘Para’ indicates a focus on people who are eligible to compete in Para sport.17 The programme specialises in physically demanding Para sports training for people with high support needs—those using wheeled mobility and requiring personal assistance for fundamental tasks of daily living. A brief vignette is available here—https://youtu.be/HxCRf7hHj7k.
Participants
Young, inactive people with CPHSN were recruited from a 30 km radius from the University of Queensland, Brisbane, Australia. Key inclusion and exclusion criteria are described fully in the published protocol.10 Participants were one female and two males with CPHSN, aged 15–16 years on enrolment, classified as GMFCS IV. None were achieving WHO physical activity guidelines, and they had not previously participated in performance-focused sports training. Two other participants were screened and did not meet inclusion criteria, and one other participant was screened and excluded due to contraindication to the intervention (see online supplemental appendix 1). Included participants provided assent, and participants’ parents/guardians provided informed consent on enrolment. Table 1 describes clinical characteristics, sport classes and stroke preference of each participant.
Supplemental material
Participant characteristics
Study design
To evaluate the effect of a performance-focused swimming programme on gross motor functioning, a Multiple-Baseline, Single-Case Experimental Design (MB-SCED) was used. It took place over 16 months, between March 2017 and July 2018, and comprised five phases A1 (baseline)-B1-A2-B2-A3: where ‘A’ phases represent periods of no training or training withdrawal, and ‘B’ phases represent training exposures, each being 16 weeks duration, all standard training block duration for competitive swimmers.10 18 Two features of this design make it particularly strong. First, there are repeated measures throughout all phases—a total of 102 data collection points, exceeding the 75 data points required for this design according to SCED guidelines.19 Second, the transitions between training and withdrawal phases were temporally staggered,10 and the 5-phase design presented a total of 12 opportunities to detect an experimental effect, 4 transitions for each participant (from an A-B or B-A phase). The MB-SCED methods are reported fully in the published protocol.10 The trial was registered (Australian and New Zealand Clinical Trial Registry number ACTRN12616000326493).
Following the MB-SCED, a 30-month follow-up period commenced during which participants continued a schedule of regular training and monitoring. Data from the full 46 months (16-month MB-SCED and 30-month follow-up period) provided a basis for comparing participant motor function with predicted motor decline.2 During the 30-month follow-up period, the training phases were extended from 16 weeks to longer training blocks aligning with the competitive swimming season, and the withdrawal phases were incorporated between seasons to facilitate recovery. Gross motor function, swimming performance and training load were longitudinally monitored.
Intervention
The intervention comprised performance-focused swimming training over the course of four consecutive competitive swimming seasons (one Paralympic cycle). The term ‘performance-focused’ refers to the fact that the sole aim of all strategies employed was to improve competitive swimming performance over 50 m. A comprehensive description of the training programme is available in the published protocol.10 Training aimed to achieve three main goals:
Improve water safety skills.
Minimise hydrodynamic drag forces.
Maximise propulsive forces.
Training was delivered by a multiprofessional team comprising qualified physiotherapists, exercise physiologists and swim coaches, supported by a multiprofessional medical team. Training session frequency increased from once per week to five times per week as the training phases progressed. Training session intensity and duration varied but aimed to gradually increase over time. The participants were paired with a typically developing volunteer training buddy who provided training assistance.
Outcomes
In the MB-SCED, repeated measures of swimming performance and gross motor function were conducted throughout five phases: A1 (Baseline)-B1-A2-B2-A3 with staggered exposure/withdrawal sequences.10 In accordance with SCED guidelines,19 a minimum of five data points occurred for each participant in each phase.10 During the baseline phase, participant 1 completed 5 data points, participant 2 completed 8 data points and participant 3 completed 11 data points. All participants then completed: phase B1 (8 data points), phase A2 (5 data points), phase B2 (8 data points) and phase A3 (5 data points).10
Swimming velocity
A full description of the test protocol, including rationale, is reported in the published prototol.10 To summarise, each participant completed a maximum-effort swimming trial. The duration of each participant’s test was based on the 2017 World Para Swimming Championships 50 m freestyle qualifying time for the participant’s class, and they swam their preferred stroke as fast and as far as possible in this allotted time. The distance covered was recorded and average swimming velocity was calculated.
Gross motor function
The Gross Motor Function Measure-66-Item Set (GMFM-66-IS) has excellent levels of overall agreement with the full version of the GMFM-66 when measuring change over time (Intraclass Correlation Coefficient or ICC≥0.9).20 Given the time-intensive nature of the full test, the short item set was appropriate for use in this study as a repeated measure of gross motor function. Scores for the tasks within the item set were entered into the Gross Motor Ability Estimator programme to obtain the final GMFM-66-item score.
Training load
Training load comprised the frequency (training sessions per week), duration (minutes spent training) and intensity, which in this study was quantified using the session-RPE (rating of perceived exertion) method.21 Each participant rated each training session intensity on the OMNI RPE scale22 which ranges from 0 (extremely easy) to 10 (extremely hard), and this rating was multiplied by the session duration to produce a given number of session RPE minutes. Weekly totals for RPE minutes were calculated.
Randomisation/blinding
Assessments were conducted by a physiotherapist with expertise in the assessment of gross motor function in people with CP. The assessor was blinded to the intervention and whether each participant was in a period of training or withdrawal at the time of assessment. Participants were randomised to either a 10-week (5 data point), 16-week (8 data point) or 22-week (11 data point) baseline period.
Statistical methods
Interrupted time-series simulation (ITSSIM) analysis23 was used to calculate a standardised mean difference effect size, d, and an unstandardised mean difference, D, for each participant’s transition from A-B or B-A in the outcomes of swimming performance and GMFM-66. Standardised effect sizes were interpreted as follows: small, 0.20–0.49; moderate, 0.50–0.80 and large, greater than 0.80.24 The 5-phase design comprised a total of 12 transitions—4 transitions (from an A-B or B-A phase) for each of three participants. The criterion for inferring causality was statistically significant effects for at least three transitions25).
The longitudinal non-linear fluctuations in GMFM, as a function of participant age, were evaluated using a generalised additive model with a penalised cubic regression spline basis function and visualised using the ‘ggplot’ function from the ‘ggplot2’ package (R Studio V.1.3.1056, PBC, Boston, Massachusetts, USA). A Gaussian distribution with an identity link function was used to produce the general additive model. Five knots were included in the model, positioned at quartiles of the observed data points. The fitted smoothed coefficients resulting from the analysis were plotted along with the 95% CIs. The GMFM fluctuations that could be expected to occur in people who are of the same age as those in the current study were plotted according to the original models developed by Hanna et al.2
Equity, diversity and inclusion and patient involvement statement
Equity and patient voice were fundamental to our justification for this study and at the forefront of the discussion of results, implications for future research and clinical practice. This work is driven by the voices of people with disabilities who have high support needs and who acted as consumer advisors for the ParaSTART programme of research. Our research team comprises both males and females from three countries and includes senior, mid-career and early-career academics.
Results
MB-SCED to evaluate the effects of a performance-focused swimming programme
Figure 1 presents training load, GMFM-66 and swimming performance data for each participant over the five phases of the 16-month SCED study. Training load is presented graphically in the three panels on the left side of figure 1. It shows that, in accordance with a multiple baseline design, the baseline (phase A1) is 11, 17 and 23 weeks for participants 1, 2 and 3, respectively. Training load during baseline and the two withdrawal periods (A2 and A3) was zero. Table 2 presents an overview of the training load completed in each of the two training phases phase B1 and B2. Total RPE minutes accrued during B1 were 10888, 7572 and 11 028 for participants 1, 2 and 3, respectively. Total RPE minutes accrued during B2 were 14673, 10 008 and 11 164, respectively.
Training load, swimming performance and GMFM-66 data for each participant throughout the five-phase A1-B1-A2-B2-A3 SCED study. GMFM-66, Gross Motor Function Measure-66-Item Set; SCED, Single-Case Experimental Design.
Training load data for each participant presented by training phase
The three middle panels of figure 1 present swimming velocity, with each participant achieving greater swimming velocity in each training phase. Table 3 presents the results of the ITSSIM analysis for swimming velocity. Data are presented for each participant and each phase transition. All participants achieved increases in swimming velocity each time the intervention was introduced (transitions A1-B1 and A2-B2), and effect sizes were moderate-large (0.61–3.75). In one instance—participant 3, transition A2-B2—the increase in swimming velocity was not statistically significant (p=0.11). Responses to withdrawal of the intervention (transitions B1-A2 and B2-A3) were more variable. Swimming performance in participants 2 and 3 decreased in the B1-A2 transition and effect sizes were moderate-large (−0.69 to −1.82); swimming velocity increased in participant 1 in this transition, though the effect size was small (0.39). Swimming velocity in participants 1 and 3 decreased in the B2-A3 transition and effect sizes were large (−1.17 to −1.64); swimming velocity increased in participant 2 in this transition, though the effect size was small (0.29).
Swimming performance and GMFM-66-IS ITSSIM results
GMFM-66 scores are presented in the three right panels of figure 1 and the results of the ITSSIM analysis are presented in table 3. All participants achieved increases in GMFM-66 score each time the intervention was introduced (transitions A1-B1 and A2-B2). Effect sizes were large in transition A1-B1 (1.15–2.26), but small-moderate in transition A2-B2 (0.11–0.74). GMFM-66 score decreased in all participants each time the intervention was withdrawn (transitions B1-A2 and B2-A3). Effect sizes were moderate-large in transition B1-A2 (−0.50 to −2.01) and small-large in transition B2-A3 (−0.47 to −2.28).
Comparison of measured and predicted motor function over 46 months
The raw weekly training load and modelled GMFM-66 data for each participant over 46 months are presented in figure 2. Training load remains relatively consistent over the entire period, although participant 1 has some large peaks in the third training period (aged 17 years). The red line indicates 750 RPE min/week, the volume of activity recommended for people with disabilities by the WHO.26
Longitudinal training load and GMFM-66 data. The left panel displays training load data for each participant between the age of 15/16 years and 19/20 years (displayed x-axis; note that the baseline period is not temporally represented). The red horizontal line denotes the RPE-minute value commensurate with national physical activity guidelines (750 RPE min/week). The right panel displays modelled GMFM data for each participant, with 95% CIs, and the red line denotes the projected trajectory of motor decline,2 from the median of baseline GMFM-66 scores. GMFM-66, Gross Motor Function Measure-66-Item Set; LCI, lower confidence interval; RPE, rating of perceived exertion; UCI, upper confidence interval.
The three right-hand panels of figure 2 present modelled GMFM-66 data for each participant. Scores increase in the first year of training (age 15–16 years), and then plateau in the subsequent 3 years into late adolescence. The red line in each GMFM panel is the predicted trajectory for GMFM-66 scores.2 For each participant, the red line originates from the median GMFM-66 score at baseline for each participant. The upward trend of modelled GMFM-66 measures for each participant contrasts with the predicted downward trend in GMFM-66 indicated by the red line.
Discussion
There were two main findings from this study. First, a performance-focused swimming training programme comprising training volumes commensurate with WHO physical activity recommendations and delivered by a skilled multiprofessional team conferred improved motor function in previously untrained, physically inactive people with CPHSN. The five-phase SCED demonstrated that motor function improved following training phases and declined following withdrawal phases in all participants, thereby indicating the relationship was causal—performance-focused swimming caused gross motor function to improve.
The second main finding was that, over a 46-month period, participant gross motor function initially improved and then plateaued around the new, improved level. These improvements occurred during a life stage when population-based modelling2 indicates that motor function typically declines. Specifically, the participants were aged 15/16 years at baseline and their GMFM-66 scores improved by between 2 and 7 points from their median baseline score and then plateaued until age 19/20 years. During the same life stage, population-based modelling predicts mean GMFM-66 scores typically fall by 4.2 points for people with GMFCS level IV CP.2 Thus, the difference between predicted and measured motor function for participants in this project was between 6.2 and 11.2 points on the GMFM-66 scale, a clinically meaningful difference. The plateau in motor function indicated a ceiling effect—participants may have, at least to some extent, maximised their gross motor capacity as measured using the GMFM-66.
Together, this study’s two main findings indicate that people with CP at GMFCS level IV who achieve physical activity guidelines during adolescence may not only prevent motor decline but improve it. The obverse of this finding is that the high prevalence of physical inactivity in this group during adolescence may account for declines which are currently accepted as clinically inevitable. This may have implications for clinical practice—and highlights the importance of including physical activity interventions as part of routine care of adolescents with CP.
Improvement in swimming velocity for all three participants validated our characterisation of the training programme as ‘performance focused’. Results support the veracity of previously reported athlete testimonies which claim that meaningful improvements in physical function are conferred by performance-focused sports training.9
We suggest three key features of the programme contributed to the results observed:
The competitive sport context: For young people with CPHSN, competitive sport has a number of advantages, and the views of ParaSTART participants have been published.27 In addition, competitive sport is age appropriate and culturally significant for many young people with CP; is routinely supported by multiprofessional teams; focuses on achievement of excellence, rather than identifying and remediating motor-sensory impairments and fosters personal interaction and teamwork28—critical features for youth who often experience social isolation and challenging periods of transition between adolescence and adulthood.29
Qualified multiprofessional staff: The delivery team included physiotherapists, exercise physiologists and coaches. They were supported by a medical doctor, dietician, occupational therapist, speech pathologist and sport psychologist. Heterogeneous, complex comorbdities and medical events/issues were managed during the programme: the number of comorbidities/medical issues for participants 1, 2 and 3 were N=5; N=8 and N=3, respectively. Table 1 lists the comorbidities and medical issues of each participant. Participants in this study could not be safely, effectively accommodated in a non-specialist, community-based swimming club.
Transport costs supported: Participants were not independent on public transport and required either a taxi or family member to drive them. Associated expenses were met by research funding and community donations.
The importance of this study is amplified because little is known about exercise training responses in people with CP, GMFCS level IV.7 In the absence of research evidence, some clinicians and researchers have vastly underestimated the physical capabilities of this group. One recent review stated that people at GMFCS IV and V ‘…will struggle performing structured exercise programmes’ and ‘are unable to perform activities greater than 1.0 MET’.30 Note that 1.0 METs is the energy expended during quiet sitting.31 Low rates of physical activity participation and gross under-representation in exercise training trials7 may result from such assertions and are refuted by results from this study. Future studies should include those with CPHSN.
Methodologically, the SCED used was ideally suited to the study aims. The design generated high-level evidence14 and conferred a range of advantages10 including permitting the allocation of time and expertise required to safely supervise participants with severe primary impairments and multiple comorbidities who were at increased risk of serious adverse events (see table 1); providing personalised assistance to alleviate the increased time cost associated with training12 and providing the methodological freedom to individualise training type, duration and intensity without compromising experimental control.
Importantly, the SCED overcame the arguably impossible task of achieving both adequate sample size and satisfactory participant homogeneity in relation to key prognostic variables for a group-level study design. Specifically, we posit that the absence of RCTs investigating responses to sport and exercise training interventions in people with CPHSN may be due, at least in part, to the infeasibility of recruiting a sample that is both large enough to adequately power the trial and also sufficiently homogenous with respect to key prognostic variables (age, sex, neurological subtype, functional effects and comorbidities). Wider use of the SCED may facilitate generation of high-quality Para sport and exercise training evidence in other heterogeneous populations, including people with acquired brain injuries and spinal cord injuries.
Limitations
This study has several limitations. First, the small number of participants and use of the SCED enhanced internal validity in this study but limited external validity. This necessitates cautious interpretation of the generalisability of the results. Second, the age range within the sample was narrow (all aged 15 years on enrolment), and it is possible that children of different ages may respond differently. Further longitudinal studies throughout the known period of decline (from age 7 years to 21 years) are required. Finally, free-living physical activity was not measured during the baseline or withdrawal periods. Although people with CPHSN typically accumulate low volumes of daily activity6 and no training was conducted during these periods, we did not control for this effect.
Conclusion
This study demonstrated that performance-focused swimming training provided a context for adolescents with CPHSN to accumulate health-enhancing volumes of physical activity, improve their swimming performance and their gross motor function during a life stage when population-based modelling predicts gross motor decline. However, this is a clinically complex population. In order to permit their effective participation in sports, priority should be placed on the development of procedures and programmes that can be delivered by a multiprofessional team. Further research employing SCED methodology is required, with emphasis on replication in this population and in other Para sports.
Data availability statement
Data are available on reasonable request.
Ethics statements
Patient consent for publication
Ethics approval
This study involved human participants and two ethical approvals were provided by the University of Queensland Ethics Committee: UQ approval #2015000831, which applied to the initial 16-month MB-SCED; and UQ approval #2018001472 which applied to the 30-month follow-up study. Participants gave informed consent to participate in the study before taking part.
References
Supplementary materials
Supplementary Data
This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.
Footnotes
Contributors IMD is the overall content guarantor and contributed to project design, data collection, analysis and manuscript preparation. MC contributed to project design. He conducted the analysis of the MB-SCED and longitudinal dataset and contributed to manuscript preparation. EB was a lead researcher on the project and contributed to project design, protocol development and manuscript preparation. LJ was a lead researcher on the project and contributed to project design, recruitment, data analysis and manuscript preparation. PW contributed to project design, data collection and analysis of the MB-SCED and manuscript preparation. AM contributed to project design, data collection and analysis of the MB-SCED and manuscript preparation. JO’S contributed to project design, data collection and analysis of the longitudinal dataset and manuscript preparation. ST was the lead investigator. He oversaw the ParaSTART programme and contributed to all aspects of the project.
Funding We gratefully acknowledge the funders of the work reported in this manuscript: (1) Queensland Academy of Sport; (2) Paralympics Australia; (3) Swimming Australia; (4) Sporting Hasbeens; (5) Gregory Terrace: St Joseph’s College; (6) Pat Rafter Cherish the Children. We also gratefully acknowledge the following individuals/groups who contributed to and facilitated the work reported in this manuscript: (1) The participants and their families; (2) Dr Gaj Panagoda—medical doctor; (3) Dr Jacki Walker—dietician; (4) Minnie Ma—physiotherapist; (5) Jean-Michel Lavalliere—swimming coach; (6) Nathan Seefeld —sport psychologist; (7) UQ Swim Club—community organisation.
Competing interests None declared.
Patient and public involvement Patients and/or the public were involved in the design, or conduct, or reporting, or dissemination plans of this research. Refer to the Methods section for further details.
Provenance and peer review Not commissioned; externally peer reviewed.
Supplemental material This content has been supplied by the author(s). It has not been vetted by BMJ Publishing Group Limited (BMJ) and may not have been peer-reviewed. Any opinions or recommendations discussed are solely those of the author(s) and are not endorsed by BMJ. BMJ disclaims all liability and responsibility arising from any reliance placed on the content. Where the content includes any translated material, BMJ does not warrant the accuracy and reliability of the translations (including but not limited to local regulations, clinical guidelines, terminology, drug names and drug dosages), and is not responsible for any error and/or omissions arising from translation and adaptation or otherwise.