Comparing the effectiveness of group-based exercise to other non-pharmacological interventions for chronic low back pain: A systematic review

Low back pain (LBP) is the leading cause of disability worldwide with a substantial financial burden on individuals and health care systems. To address this, clinical practice guidelines often recommend non-pharmacological, non-invasive management approaches. One management approach that has been recommended and widely implemented for chronic LBP is group-based exercise programs, however, their clinical value compared with other nonpharmacological interventions has not been investigated systematically.


Introduction
Low back pain (LBP) is the leading cause of disability globally with a substantial financial burden on individuals, families, communities and governments worldwide [1]. At an individual level, LBP diminishes quality of life by limiting activities of daily living, deteriorating mental health, decreasing life span [2] and inducing financial hardships [3]. Therefore, LBP is thought to be the most costly disability of the working-age population [4]. The nature of LBP is highly prevalent and recurrent: the lifetime occurrence is estimated to be 85%, and~50% of people will have at least 10 episodes in their lifetime [1].
In addressing chronic LBP, clinical practice guidelines often recommend non-pharmacological and non-invasive management approaches for chronic LBP [3]. Specifically, these guidelines recommend education and exercise as first-line interventions [5][6][7]. While many randomised controlled trials have provided scientific evidence supporting the benefits of exercise in chronic LBP [8], how to best deliver exercise interventions is less clear. Individual exercise programs are the most widely implemented approach for addressing chronic LBP [9]. In contrast, group exercise-based classes have been found to be beneficial [10][11][12], but are not as widely used. Group exercise may be an equally effective alternative to individual exercise with potentially lower healthcare costs [8]. The potential for social support and better social interaction in groups should also be considered a potential advantage [8]. With this in mind, group exercise approaches have been recommended by the National Institute of Health and Care Excellence [12].
Given the above, we could not identify any prior systematic reviews that compared groupbased exercise to individual non-pharmacological interventions that may include education and/or exercise in people with chronic LBP. Therefore, we conducted this review to evaluate the comparative effectiveness of group-based exercise to other non-pharmacological interventions that may or may not include education and exercise on pain and disability in patients with chronic LBP.
Group-based exercise programs were defined as a group of three or more participants taking part in an exercise class supervised by a health care provider. A non-pharmacological intervention was defined as one-on-one care between a health care provider and their patient that did not involve pharmaceuticals. The intervention programs were identified using the search terms "group exercise"," "GLA:D Back", "group strengthening", "group physical activity", or "group strength training". Low back pain was identified using the search terms "chronic back pain", "persistent back pain", "long-standing back pain", "long-duration back pain", "longstanding lumbar pain", "long-duration lumbar pain", "chronic low back pain", "persistent low back pain", "long-standing low back pain", or "long-duration low back pain".

Eligibility criteria
Only peer-reviewed, randomized, controlled trials comparing group-based exercise including strengthening, physical activity, and strength training with other types of non-pharmacologic interventions for chronic LBP were included. We excluded reports related to conference proceedings, specific low back pain diagnoses, case series of fewer than ten subjects, case studies, systematic reviews, and protocol papers.

Selection of studies
Two investigators (JL and VA) with more than 10 years of cumulative experience in reviewing literature screened all titles and abstracts independently and retrieved the full text of the potentially eligible studies. Disagreements at the titles and abstracts stage were resolved through consensus.

Data extraction
A standard form (S2 Appendix) was developed to extract data based on published guidelines [15][16][17]. Data for each study were extracted and cross-checked by two investigators (JL and VA). Disagreements were resolved by a third investigator (GK). The following information was extracted for each study: 1) characteristics of the participants: sample size, age, gender, height, diagnosis, pain duration, location and intensity; 2) inclusion and exclusion criteria; 3) characteristics of the interventions: the type, length of the program, mode of application, frequency and duration of group and individual exercise based physiotherapy; 4) characteristics of the outcomes: pain and disability outcomes measures, follow-up times.

Methodological quality
The quality of included studies was assessed as outlined by PRISMA, and the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guidelines [18]. The quality appraisal focused on seven categories: subject recruitment, examiners, methodology, outcomes, handling of missing data, statistical analysis, and results (S3 Appendix). Two reviewers (JL, VA) conducted critical appraisal separately on each of the papers and decisions were verified through consensus. Practice appraisals and discussion of five full-text papers occurred for calibration before the full review. Studies with a minimum score of 70% were considered to be of high quality and those with a lower score to be of low quality [19]. moderate, limited, no, and conflicting evidence) for the effect of interventions was determined according to the consistency of the research findings and the methodological quality of the included studies [19]. The level of evidence was considered strong if there was more than 75% agreement between at least two high-quality studies and more than two low-quality studies on the outcome of the interest (Table 1) [19].
The evidence was considered moderate if there was more than 75% agreement between a high-quality study and at least three low-quality studies (Table 1) [19]. The evidence was considered limited if only one high-quality study reported that outcome or at least three out of four low-quality studies (75%) reported the same outcome (Table 1) [19]. The evidence was considered conflicting if there was less than 75% agreement among the studies irrespective of study quality (Table 1) [19].
Summary tables were prepared for participants' descriptions (Table 2), intervention used (Table 3), quality appraisal scores (Table 4), the level of evidence summary statements and outcomes extracted (Table 5).

Studies included
The search identified 639 references after removing duplicates (Fig 1). Following title and abstract screening, 628 papers were excluded. One paper was identified by manual searching. This resulted in a total of 11 papers meeting the selection criteria. The most frequent reason for exclusion was inappropriate study design (e.g. did not carry out between-group comparisons).

Pain information
Of the 11 studies meeting the inclusion criteria, all enrolled participants reported chronic LBP. All but one of the 11 studies reported on pain chronicity [20] (Table 2) Seven of the included studies reported pre-intervention and post-intervention pain intensity [20][21][22][23][24][25][26]. Table 3 summaries the intervention, duration, metric, and data collection time points used in the included studies. From the resulting 11 studies, 27 different outcome measurements were identified (Table 3).

Methodological quality
Five studies met the methodological high-quality threshold of 70% (Table 4) [20,22,25,28,30]. Five studies scored between 60% and 69% [23,24,26,27], and one scored 50% [21]. The major source of bias in the resulting 11 papers was the failure to formulate correlation and Table 1. Levels of evidence for summary statements and description of criteria adopted a priori to determine the level of evidence [19].

Moderate
1 high-quality � study and consistent findings (�75%) in 1 or more low-quality studies

Limited
Findings in 1 high-quality � study or consistent results (�75%) among low-quality studies

No
No study identified

PLOS ONE
Comparing the effectiveness of group-based exercise to non-pharmacological interventions for chronic LBP mean difference-testing hypotheses (i.e. a priori). These studies did not provide any information regarding the expected direction of correlations or if the mean differences met the original hypotheses. All studies clearly described 1) their sample size estimation for each experimental group and 2) their main findings.

Measurement outcomes
From the resulting 11 studies, 47 different outcome measurements were identified with the resulting level of evidence and summary statements described in Table 5.

Primary outcome measures
Self-administered disability measures. Low back pain associated disability was evaluated in 10 studies. Five studies used the Roland-Morris Disability Questionnaire [20,25,26,29,30]; four used the Oswestry Disability Index Questionnaire [22,24,27,28] and one used Quebec back pain disability scale [23]. There was strong evidence of no difference between groups 3-month post-intervention from 3 high-quality studies and a study with moderate quality [20,22,26,30]. Likewise, there was limited evidence of no difference between groups from one study for 9-month and 15-month post-intervention [20] and another study for 6-month postrandomization [24]. Two studies compared the post-intervention disability level with preintervention disability level [23,26]. There was limited evidence of lower disability scores in people who received individual intervention compared to group exercise immediately and 6-month post-intervention. Results indicated limited evidence of no difference between exercise and education vs. education group only at 3-month and 6-month post-intervention compared to the base-line group [26]. The results were inconsistent from two studies 6-month post-intervention [23], from two studies 3-month post-randomization [24,29], and three studies 6-month post-randomization [24,28,29]. There was limited evidence from one study for lower disability scores 4-week post-intervention (Table 5). People in the group exercise (intervention group) had a lower disability score than people in the waiting list (control) 4-week post-intervention [25]. Likewise, there was limited evidence from one study for lower disability scores 6-week post-randomization [29]. In this study, people in the yoga intervention group had a lower disability score than people in the booklet only group 6-week post-intervention   [29]. In this study, the difference was not significant between yoga and conventional therapeutic exercise classes vs. self-care book, and between conventional therapeutic exercise classes vs. self-care book [29]. There was limited evidence from one study for lower disability scores 12-month post-randomisation (Table 5). Cognitive functional therapy led to greater reductions in disability compared with the group exercise intervention [28].

PLOS ONE
Comparing the effectiveness of group-based exercise to non-pharmacological interventions for chronic LBP

Groups compared
Limited 1 [28] A lower score for Cognitive functional therapy

12-month postrandomization
Group-based exercise + education vs. Cognitive functional therapy

Risk of Chronicity
Limited 1 [28] No difference 6-month postrandomization Group-based exercise + education vs. Cognitive functional therapy Limited 1 [28] A lower score for Cognitive functional therapy

12-month postrandomization
Group-based exercise + education vs. Cognitive functional therapy

Coping
Limited 1 [28] No difference 6-month postrandomization Group-based exercise + education vs. Cognitive functional therapy Limited 1 [28] A lower score for Cognitive functional therapy

12-month postrandomization
Group-based exercise + education vs. Cognitive functional therapy

Number of Pain Sites
Limited 1 [28] No difference 6-month postrandomization Group-based exercise + education vs. Cognitive functional therapy Limited 1 [28] No difference 12-month postrandomization Group-based exercise + education vs. Cognitive functional therapy

Short Form Health Survey-Physical Component
Limited 1 [20] No difference 3-month post-intervention Group Exercise vs. Individual Physical Therapy Limited 1 [20] No difference 12-month postintervention Group Exercise vs. Individual Physical Therapy

Short Form Health Survey-Mental Component
Limited 1 [20] No difference 3-month post-intervention Group Exercise vs. Individual Physical Therapy Limited 1 [20] No difference 12-month postintervention

PLOS ONE
Pain. Pain level was measured in three studies using the Visual Analogue Scale [22,23,25] and using the Numeric Pain Rating Scale in four studies [21,24,26,28] (Table 5). There was moderate evidence of no difference between groups for 6-month post-randomization and 12-month post-randomization [24,28]. There was limited evidence of a lower pain score of people in the group exercise and education compared people of the education group 3-month and 6-month post-intervention compared to baseline [26]. There was limited evidence of nondifference between groups for immediately and 6-month post-intervention [21], 9-month and 15-month post-intervention [20], and 3-month post-randomization [24]. There was limited evidence of a lower pain score of people in the group exercise compared to people of the individual intervention group 4 week post-intervention [25].

Secondary outcome measures
Quality of life. Quality of life was evaluated in four studies. Two studies used the EQ-5D quality of life scale [20,30], one used the EQ-5D-5L, one used the EQ-VAS [30] and one study used the short form SF-36 Health Survey [29]. There was strong evidence of no difference between groups in health surveys scores from two high-quality studies [20,30]. Likewise, there was limited evidence of no difference among groups for all measurement time points [20,21,29,30].
Lumbar spine flexibility (flexion, extension, and lateral flexion). There was limited evidence for no difference between groups post-intervention and 12-month post-intervention [23] with respect to group exercise vs. individual intervention on lumbar spine flexibility, however, there was limited evidence for more flexion, extension, and lateral bending range of motion in people of the group exercise group compared to the controls 4-week and 8-week post-intervention [25]. Likewise, there was limited evidence of a higher range of motion for lumbar extension and lateral bending 6-month post-intervention [23]. Differences in the flexion range of motion between these groups were not significant [23].
Fear beliefs. Low back pain associated fear beliefs were evaluated in three studies [24,26,27] with inconsistent results irrespective of the quality of the studies included. One study evaluated pain-related fear with the Tampa Scale of Kinesiophobia-13 (TSK-13, a modified version of the original Tampa scale of Kinesiophobia) [26], one used the Fear-avoidance Beliefs Questionnaire (FABQ) [27] and one used the Fear Avoidance Beliefs Questionnaire-PA subscale and Back Beliefs Questionnaire [24]. There was limited evidence of no difference among groups for fear beliefs 3-month post-intervention [24], 3-month and 6-mont post-randomization [26], either 6-month post-intervention [24] or post-randomisation [28], and either 12-month post-intervention [24,27] or post-randomisation [28].
Other outcome comparisons. Most studies reported outcome measures in addition to those describing disability, quality of life and pain (Table 5). One study showed limited evidence that cognitive functional therapy was superior in pain self-efficacy, risk of chronicity, and coping compared to group-based exercise [28]. The remaining other outcome measures had limited evidence of no difference between the group and individual programs (Table 5).

Main findings
The present systematic review identified strong evidence of no difference in disability level and pain scores 3-month post-intervention in people with chronic low back pain group-based exercise compared with controls that underwent other non-pharmacologic interventions. We also identified moderate evidence of no difference between group exercise and cognitive functional therapy for 6-month post-randomization and 12-month post-randomization. We could not find any strong or moderate evidence for or against the use of group-based exercise in the rehabilitation of people with chronic LBP for other time-points and health measurement outcomes.
These findings are consistent with findings of a recent systematic review conducted by O'Keeffe et al. [8] that compared individual exercise to group exercise for all musculoskeletal conditions including LBP. O'Keeffe et al. [8] found that for disability and pain, no clinically significant differences were found between the group and individual physiotherapy including exercise for all musculoskeletal conditions. They also found seven studies that specifically related to LBP that also noticed no clinically significant differences in disability and pain when comparing group and individual physiotherapy involving exercise [8].
While our results suggest there is no difference between group exercise and non-pharmacological interventions, there was one study that demonstrated limited evidence that cognitive functional therapy was superior in self-administered disability measures 6 and 12-month postrandomization compared to baseline. The same study indicated that cognitive functional therapy was superior in pain self-efficacy, risk of chronicity, and coping compared to group-based exercise 12-month post-randomization compared to 6-month post-randomization [28].
Some secondary outcomes demonstrated interesting findings but were not frequently used in the included studies. These included fear-avoidance, QoL and cost. Based on one study investigated here, group-based exercise reduced fear-avoidance scores [32], improved quality of life measures compared to usual general practitioner care [20] and lowered costs [23]. Based on these studies, further exploration of these outcomes in relation to group-based exercise performance is warranted.

Study limitations
This review solely included studies published in English, and no search was conducted of the grey literature. These two factors may have caused a potential bias in selecting relevant studies. As discussed previously, the papers identified here were highly heterogeneous which prevented meta-analysis. Unfortunately, the literature was not sufficiently rich to focus our review on head-to-head comparisons of group-based exercise with individual-based exercise and other specific interventions.
Further, in terms of our specific summary statements, some of these studies conflicted with each other depending on the time-points compared ( Table 5). The majority of conflicts were observed for timepoints with two or three studies (each study weighted 50% or 33.33% in the summary statement, respectively). This indicates that even a different observation from a lowquality study could drastically change the level of evidence for a specific summary statement. The limited evidence summary statements often showed no difference among interventions. The studies compared were heterogeneous in terms of the population studied (different ages, different time points, different pain and disability level among participants) or because of other methodological considerations, which may have contributed to the frequent conflicting evidence summary statements and limited our ability to observe consistent effects of group-based exercise.

Conclusion
We identified strong evidence of no difference between group exercise and other non-pharmacological LBP interventions for disability level, quality of life, and pain. The remaining evidence was not of sufficiently high quality to permit further conclusions. With this equivocal finding, group-based exercise may be a preferred choice given potential advantages in other domains not reviewed here such as motivation and cost. Further research in this area is needed to evaluate this possibilty.