- Top of page
- Supporting Information
While recommendations for the duration, frequency, mode and intensity of exercise programmes for people with chronic obstructive pulmonary disease (COPD) are specified in consensus statements, criteria for exercise session attendance are less clear. The review questions were: (i) how commonly are a priori criteria and attendance rates reported for people with COPD participating in exercise programmes and (ii) what is the strength of association between attendance and improvements in functional exercise capacity. Database searches identified primary studies of people with COPD participating in exercise or pulmonary rehabilitation programmes of at least 2 weeks duration. Primary outcomes were a priori criteria for attendance, reports of attendance at supervised exercise sessions and mean improvements in functional exercise assessments. Data extraction processes were confirmed prospectively (>80% agreement). Variants of exercise attendance data were described. Linear associations between attendance and improvements in exercise outcomes were explored (Pearson r, P < 0.05). Of the 234 included studies, 86 (37%) reported attendance and 29 (12%) provided a priori criteria for attendance. In the small sample of studies which reported attendance and functional exercise data before and after the intervention, there was little to no relationship between improvements in functional exercise capacity and training volume (prescribed r = −0.03, P = 0.88; attended r = −0.24, P = 0.18). Reporting of exercise programme attendance rates is low and of variable quality for people with COPD. Consistent and explicit reporting of exercise attendance in people with COPD will enable calculation of dose–response relationships and determine the value of a priori exercise attendance criteria.
- Top of page
- Supporting Information
Multiple controlled trials have shown positive effects of pulmonary rehabilitation or exercise training for individuals with chronic obstructive pulmonary disease (COPD).[1, 2] While there remains no consensus on the optimal duration and frequency of pulmonary exercise programmes,[3-6] exercise training is typically prescribed three to five times per week for at least eight successive weeks. Current pulmonary rehabilitation guidelines do not request attendance data to be reported, nor state a minimum required attendance rate for successful completion of pulmonary rehabilitation. In principle, non-attendance could impact upon training load (exercise dose) and consequent improvements in functional exercise outcomes.
From a clinician's view point, information concerning the number of people attending supervised exercise sessions is useful to plan recruitment processes and resources (space, equipment and staff) and interpret whether improvements in exercise outcomes reflect all (intention-to-treat (ITT) analysis) or only people with a certain attendance level (per-protocol (PP) analysis). If intervention studies report training prescription, exercise session attendance rates and improvements in functional exercise outcomes, it should be possible to calculate dose–response characteristics and identify whether an optimal criterion level for exercise session attendance exists.
This systematic review aimed to determine: (i) how commonly a priori criteria and attendance rates are reported for people with COPD participating in exercise training programmes; and (ii) the strength of association between exercise session attendance (vs exercise prescription) and improvements in functional exercise capacity.
- Top of page
- Supporting Information
A systematic search of CINAHL (1981–6 October 2011), Medline (1948–6 October 2011), Embase (1974–6 October 2011) and CENTRAL (Cochrane) was undertaken (Table S1 in the supplementary information available online presents search terms and citation yield). Citations were included if: (i) the original study was published in English and included people with COPD; (ii) pulmonary rehabilitation was the primary intervention (with or without adjunctive strategies), was at least 2 weeks in duration, irrespective of the intervention site (e.g. hospital, community, home) and included exercise training as a key component; and (iii) the study design was prospective experimental or observational (controlled and uncontrolled clinical trials, cohort studies). Studies concerning maintenance following pulmonary rehabilitation, associations between physiological and behavioural responses in people with COPD and cost/benefit analysis of rehabilitation were excluded, as were abstracts, study protocols, commentaries and narrative approaches (reviews, editorials, letters).
An a priori strategy including both the search and inclusion/exclusion criteria was developed and piloted by the research team. Studies were included if any statement concerning attendance or participation in exercise training sessions was reported. In the absence of an a priori or specific statement concerning attendance of exercise sessions, studies that reported only the number of participants included in analysis were excluded. Reporting of attendance at exercise sessions could be numeric (number of exercise sessions attended or required to ‘complete’ training or number of people excluded as a result of not attending the exercise intervention) or text (e.g. ‘all participants attended the exercise sessions prescribed’). Where participants were excluded from either the study or analysis for ‘poor compliance’, only studies where the poor compliance was reported to relate to the exercise training sessions were retained.
Consistency of data extraction was confirmed for all authors (n = 12) with 10 randomly selected articles retained from the search (≥80% agreement required for acceptable consistency). Full versions of all retained citations were accessed and uploaded to an internet-based reference management system (RefWorks Version 2.0; RefWorks-COS, Bethseda, MD, USA). Reference lists from included articles were reviewed to further identify eligible articles. For data extraction, authors worked in pairs, with each member extracting data independently and conferring to reach consensus for each data item. When consensus could not be reached, a third independent reviewer was consulted.
Traditionally, systematic reviews of intervention efficacy, diagnosis or prognosis include appraisal of methodological bias. The first aim of this systematic review was to audit whether, and how, authors reported a priori criteria for, and attendance at, supervised exercise sessions in people with COPD. Appraisal of methodological bias was undertaken only for studies providing sufficient data to explore the second aim of this review which was to assess the strength of association between exercise session attendance and improvements in functional exercise capacity. Scores from the publically available Physiotherapy Evidence Database PEDro scale for randomised controlled trials (PEDro http://www.pedro.org.au/) were retrieved. Studies not currently available within PEDro were independently assessed by two members of the research team using the PEDro scale with consensus reached by discussion.
The percentage of included studies reporting attendance rates and/or an a priori criterion for attendance at exercise sessions was calculated. Publications were grouped according to whether metrics for attendance could be extracted: group 1—descriptive statement but no metric available; group 2—percentage, mean/median available; and group 3—all central tendency metrics available (mean (standard deviation), mean standard error and/or median (interquartile range)). For studies with multiple active intervention groups, the category was based on their ‘best’ reporting of attendance.
Where reported by authors, data concerning the analysis approach (ITT, PP or both) were extracted. In studies which reported a priori criteria for attendance at exercise sessions but did not state a specific analysis approach, the results section was reviewed to determine if analyses included all recruited participants (ITT) or only those meeting a priori criteria (PP).
The strength of linear association between mean percentage attendance at supervised exercise sessions (at least a mean or median available (groups 2 and 3)) and prescribed exercise frequency per week or weeks of programme duration were analysed descriptively (Pearson's correlations coefficient (r) where values from 0 to 0.25 represent little or no relationship, 0.26 to 0.50 fair, 0.51 to 0.75 moderate to good and >0.75 good to excellent linear relationship). In studies which reported mean percent attendance at supervised exercise sessions and functional field exercise outcomes (6 min walk test or 12 min walk test (minute walk test with improvement in 12 min walk test divided by 50%)) or incremental shuttle walk test, data for mean pre to post training differences were extracted and converted to percent increase in metres from baseline. The maximum number of prescribed supervised exercise sessions possible in each study was used as an indicator of training volume, where prescribed training volume = prescribed frequency per week (days) × the programme duration (weeks); and actual training volume = reported mean attendance at supervised exercise sessions per week (days) × the programme duration (weeks).
- Top of page
- Supporting Information
The outcome of the search strategy is presented in Figure 1. Of the 234 articles reporting an exercise training/pulmonary rehabilitation programme for people with COPD, 86 (37%) reported attendance of participants in exercise training sessions (summary of included studies in Table S2 in the supplementary information available online). The number of studies reporting attendance increased over time (Fig. 2). Only 29 studies (12%) reported a priori criteria for attendance (Table 1a), which represented 34% of the studies reporting attendance.
Figure 2. Publication pattern for studies of exercise training in people with chronic obstructive pulmonary disease (COPD) including studies that reported attendance. ○, studies of pulmonary rehabilitation / exercise training in COPD; ●, studies reporting attendance at exercise sessions.
Download figure to PowerPoint
Table 1a. Studies including a priori criteria for exercise attendance (ordered by percentage of the intervention population meeting the a priori criterion for attendance)
|Group code||Study||Programme duration (week) : frequency (week) : maximum number of exercise sessions||Criterion for session attendance n (%)||Reference provided for criterion||Participants meeting criterion/participants recruited to active groups (%)||ITT or PP reported or inferred|
|3||||12:2:24||24 (100)|| ||13/18 (72%)||PPI|
|3||||10:2:16||16 (100)|| ||20/27 (74%)||PPI|
|1||||9:3:21||21 (100)|| ||307/316 (97%)||PPI|
|3||||8:3:24||24 (100)|| ||29/38 (76%)||PPI|
|3||||8:3:24||24 (100)|| ||24/32 (75%)||PPI|
|3||||8:3:24||24 (100)|| ||41/48 (85%)||PPI|
|3||||8:2:16||16 (100)|| ||40/54 (75%)||PPI|
|3||||8:1–2:12||12 (100)|| ||57/57 (100%)||PPI|
|3||||6:3:18||18 (100)|| ||25/28 (89%)||PPI|
|3||||6:1 to 2:6 or 12||6 or 12 (100)|| ||66/91 (73%)||PPR|
|3||||NS:3:20||20 (100)|| ||25/25 (100%)||PPR|
|2||||4:2:8||7 (88)|| ||22/23 (96%)||PPI|
|1||||7:2:14||12 (86)|| ||44/55 (86%)||ITTR|
|1||||12:3:36||25.25 (85)|| ||35/47 (74%)||PPI|
|1||||6:2:12||10 (83)|| ||49/54 (91%)||ITTR|
|1||||8:3:14||20 (83)||Yesa||22/30 (73.3%)||ITTR|
|1||||52:5:260||208 (80)|| ||23/27 (86%)||PPI|
|3||||16:2:32||24 (75)|| ||60/100 (60%)||PPI|
|1||||8:2:16||12 (75)||Yesb||19/47(40%)||ITTR and PPR|
|1||||8:2:16||12 (75)|| ||42/63 (67%)||PPI|
|3||||4:5:20||15 (75)|| ||98/104 (92%)||PPR|
|3||||8:2:16||12 (75)|| ||40/45 (89%)||PPI|
|2||||8:2:16||12 (75)|| ||20/30 (67%)||ITTR and PPR|
|1||||8:3:24||15 (62)|| ||240/252 (95%)||ITTR and PPR|
|2||||6:2:15||>7.5 (50)|| ||21/43 (49%)||PPI|
|1||||8:2:16||8 (50)|| ||23/30 (77%)||ITTR|
|2||||12:3:36||18 (50)|| ||24/30 (80%)||ITTR and PPR|
|1||||8:2:16||8 (50)|| ||55/63 (87%)||PPI|
|1||||8:1–2 per day: group 1 = 55 group 2 = 28||‘6 weeks of uninterrupted training’|| ||28/45 (62%)||PPI|
The majority of studies providing a priori criteria for exercise session attendance used this as a definition for completion of training/pulmonary rehabilitation[4, 10-19, 21, 25, 26, 28, 30, 31, 34, 36, 37] or adherence.[20, 22, 23, 27, 32, 33, 35] Three studies did not provide a specific purpose for the a priori criteria[18, 24, 29] (Table 1a). While not reported as a priori criteria, five studies reported retrospective attendance as an indicator of the degree of adherence[39-43] (Table 1b). Attendance at exercise sessions was reported in a variety of ways, with 45% (n = 38) of studies reporting or providing sufficient data to calculate measures of central tendency and variability (Table 2).
Table 1b. Studies including retrospective criterion for exercise attendance
|Group code||Study||Programme duration (week) : frequency (week) : maximum number of exercise sessions||Criterion for session attendance n (%)||Reference provided for criterion||Participants meeting criterion/participants recruited to active group/s (%)||ITT or PP reported or inferred|
|1||||4:5:20||85%|| ||37/40 (93%)||PPI|
|1||||6:2:12||83%|| ||ns/44 (ns)||PPR|
|1||||48:3:118||70%|| ||51/77 (66%)||ITTR and PPR|
|1||||9:3:25||67%|| ||89/99 (90%)||ITTR|
|2||||52:2-1:27||63%|| ||20/26 (77%)||ITTR and PPR|
Table 2. Reporting styles for attendance at exercise sessions (further detail available in Table S2 in the supplementary information available online)
|Reporting style for attendance at exercise sessions||Percentage attendance not calculable (group 1)||Percentage reported or calculable (group 2)||Percentage reported as mean SD, median IQR (group 3)|
|n = 26||n = 22||n = 38|
|Generic statements||2|| || |
|Categorical data for attendance||1|| || |
|Proxy data||1|| || |
|Number failing to adhere to exercise training (without reporting percentage of exercise training sessions attended for those that were retained in study)||9|| || |
|Number excluded due to poor adherence||1|| || |
|Number/percentage of participants meeting a priori criteria||7|| || |
|Number/percentage meeting retrospective adherence criterion||5|| || |
|Number/percentage of sessions attended|| ||12|| |
|Number/percentage and range of sessions attended|| ||4||32|
|Median percentage ± range of sessions attended|| ||4||1|
|Percentage of sessions attended reported as mean across active groups|| ||1||1|
|Incomplete attendance of sessions reported across all groups|| ||1||3|
|Other|| || ||1a|
Associations between attendance rates, prescribed exercise volume and improvements in exercise outcomes
Of the 36 studies where mean percentage attendance could be extracted (group 2—percentage, mean/median available[20, 31, 33, 36, 44-52] and group 3—all central tendency metrics available[29, 30, 53-71]), there were 54 individual intervention groups which included supervised exercise. Attendance rates ranged from 48% to 100% (mean ± standard deviation of 89 ± 14%). There was fair association between mean percentage attendance and prescribed frequency of exercise per week, with higher attendance rates for programmes with more frequently prescribed exercise sessions (r = 0.26, P = 0.06). Programmes of longer duration had significantly lower attendance rates (r = −0.42, P = 0.002), although the percentage of variance in attendance explained by programme duration was low (r2 < 20%).
Twenty studies (23%) involving 33 active intervention groups reported both prescribed and attended supervised sessions (training volume; frequency of supervised exercise per week × programme duration) and data which enabled improvement in exercise outcomes to be calculated as a percentage change from baseline (rather than metres).[11, 13-15, 20, 28-30, 36, 44, 45, 53, 54, 57, 59, 65, 66, 68, 69, 72] Seventeen studies used the 6 min walk test, one study used the 12 min walk test and two studies used the incremental shuttle walk test.[36, 65] Risk of methodological bias was moderate across studies (median score 6/10, range 2–8, where scores approaching 10 indicate low risk of methodological bias) (Supplementary Information Table S3).
On average, the difference between the prescribed maximum number of supervised exercise sessions and the mean number of exercise sessions attended was 3 (± 5) (range +6 to −19; prescribed training volume = 25 ± 8 vs attended training volume = 22 ± 6). This equates to a mean of 7% of prescribed exercise sessions not being attended. There was little to no relationship (r = 0–0.25) between improvements in functional exercise capacity and training volume (prescribed r = −0.03, P = 0.88 or attended r = −0.24, P = 0.18 Fig. 3). Improvements in exercise outcomes ranged from a mean increase of 8% (mean number of sessions attended 30% greater than the maximum number of sessions prescribed) to a mean increase of 15% in exercise outcome (mean number of sessions attended 52% lower than the maximum number of session's prescribed).
Figure 3. Relationship between improvements in functional exercise capacity and training volume (prescribed or attended). ●, prescribed; —, r = −0.03, P = 0.88; ○, attended; - - -, r = −0.24, P = 0.18.
Download figure to PowerPoint
- Top of page
- Supporting Information
This review clearly demonstrates low levels of reporting attendance data with a wide variety of metrics described in published trials of exercise training programmes for people with COPD. After reviewing 234 eligible full-text articles, only 37% reported attendance of participants in exercise training sessions, and of these, 12% reported a priori criteria for attendance (range 50 and 100%). Fair associations were calculated between exercise session attendance, exercise frequency and programme duration. No significant association was found between improvements in exercise performance and prescribed or attended training volume.
It is important to note that our analysis included only those studies published in English that reported attendance rates at supervised exercise training sessions. On review of a random sample (15%) of studies excluded due to the language of publication, it appears unlikely that inclusion of studies published in a language other than English would have improved the reporting rate for attendance. It is more than likely that many of the studies excluded from this review collected data on attendance in exercise sessions which was not reported in the publication. Where attendance was low or modest, authors may have perceived this as a limitation. Alternatively, where attendance was high, or the a priori criteria of 100% were required, authors may have elected not to report attendance on the assumption that the reporting of prescribed exercise frequency and programme duration would suffice.
Theoretically, exercising more frequently per week should have a positive relationship with improvement in exercise outcomes. In an earlier synthesis of exercise frequency and exercise outcomes, Ringbaek et al. plotted mean differences in the 6 min walk test or 12 min walk test between control and active intervention groups against prescribed frequency of exercise per week for six controlled trials. We based our descriptive analysis upon this model but included both prescribed and attended supervised sessions (training volume; frequency × programme duration) and calculated improvement in exercise outcomes as a percentage change from baseline (rather than metres). It should be noted that this descriptive analysis did not account for mode or intensity of exercise and included only a small sample of 20 studies. No relationship between training volume (attended or prescribed) and improvement in functional exercise outcomes was detected. If the reporting of attendance is a random phenomenon and our sample is representative of all relevant studies, then there may truly be little, if any association between attendance, training volume or exercise capacity. However, due to the low number of studies reporting attendance to date, the variable reporting of attendance metrics and the lack of ITT analysis, the associations presented in this paper are highly censored and likely to reflect only a small portion of the actual dose–response curve. Other factors that might explain the wide variety in exercise responses are differences in study population, training exercises, unsupervised exercise in addition to supervised sessions, (unknown) biases caused by dropouts and in some cases lack of randomisation.
There are a number of alternative ways in which the relationship between frequency of training and improvements in exercise outcomes could be explored. This review included studies which prospectively aimed to compare the effect of different exercise frequencies[4, 19, 31, 46] or reported exercise attendance–response associations.[27, 74-76] In addition, where studies had a priori criteria for attendance and included both ITT and PP analysis, exercise outcomes could be compared.[31, 41, 43, 53, 68, 77] Overall, there were a greater number of studies which provided evidence to support a significant relationship between increased frequency of exercise training leading to greater improvements in exercise outcomes[4, 41, 43, 45, 46, 53, 68, 74, 76, 77] compared with those which did not support this association.[19, 27, 31, 75]
In most studies, there are participants who attend very few supervised exercise sessions, leading to negligible improvements in exercise outcomes and reduction in the group effect. If these people are excluded (PP analysis), this is likely to inflate both exercise outcomes and attendance rates. Studies which reported an a priori criterion for attendance appear to preferentially report PP analysis (Table 1a) and rarely reported ITT analysis[21, 23, 24, 34] or both ITT and PP analysis.[27, 31, 32, 35]
In conclusion, this systematic review demonstrated inconsistent and incomplete reporting of attendance data in studies investigating exercise training programmes for people with COPD. To investigate this cost defining feature, we recommend that all studies that evaluate exercise programmes for people with COPD should: (i) report attendance at exercise sessions, preferably using mean and standard deviation (median and interquartile range for skewed data) or use graphic representation; (ii) provide a statement concerning a priori requirements for exercise attendance and why the criterion has been set at a particular level; and (iii) report the dose–response relationship between the primary health outcome and exercise session attendance rate for the entire sample, irrespective of programme completion (ITT analysis). Consistent and explicit reporting of exercise session attendance would be useful for clinicians planning to develop or set up specific exercise training programmes based on previous literature, as well as for researchers evaluating efficiency and effectiveness trials. These reporting standards would permit not only more rigorous secondary analyses of dose–response characteristics of exercise training in this population, but also enable decisions about the level and value of a priori criteria for attendance.