The Effectiveness of Pilates Exercise in People with Chronic Low Back Pain: A Systematic Review

Objective To evaluate the effectiveness of Pilates exercise in people with chronic low back pain (CLBP) through a systematic review of randomised controlled trials (RCTs). Data Sources A search for RCTs was undertaken using Medical Search Terms and synonyms for “Pilates” and “low back pain” within the maximal date range of 10 databases. Databases included the Cumulative Index to Nursing and Allied Health Literature; Cochrane Library; Medline; Physiotherapy Evidence Database; ProQuest: Health and Medical Complete, Nursing and Allied Health Source, Dissertation and Theses; Scopus; Sport Discus; Web of Science. Study Selection Two independent reviewers were involved in the selection of evidence. To be included, relevant RCTs needed to be published in the English language. From 152 studies, 14 RCTs were included. Data Extraction Two independent reviewers appraised the methodological quality of RCTs using the McMaster Critical Review Form for Quantitative Studies. The author(s), year of publication, and details regarding participants, Pilates exercise, comparison treatments, and outcome measures, and findings, were then extracted. Data Synthesis The methodological quality of RCTs ranged from “poor” to “excellent”. A meta-analysis of RCTs was not undertaken due to the heterogeneity of RCTs. Pilates exercise provided statistically significant improvements in pain and functional ability compared to usual care and physical activity between 4 and 15 weeks, but not at 24 weeks. There were no consistent statistically significant differences in improvements in pain and functional ability with Pilates exercise, massage therapy, or other forms of exercise at any time period. Conclusions Pilates exercise offers greater improvements in pain and functional ability compared to usual care and physical activity in the short term. Pilates exercise offers equivalent improvements to massage therapy and other forms of exercise. Future research should explore optimal Pilates exercise designs, and whether some people with CLBP may benefit from Pilates exercise more than others.


Introduction
Chronic low back pain (CLBP) is defined as pain for more than twelve weeks in the posterior lumbar region between the twelfth ribs and inferior gluteal folds [1]. CLBP is highly prevalent and associated with significant levels of disability [2][3][4]. As a consequence, CLBP places a large social and economic burden on society [2][3][4].
Pilates exercise is a commonly prescribed to people with CLBP [5][6][7]. Pilates exercise is named after its founder, Joseph Pilates, who developed a series of exercises in the 1920s to encourage physical and mental conditioning [8,9]. Core stability, strength and flexibility are emphasised in Pilates exercise, as is control of movement, posture, and breathing [9]. All of these aspects of Pilates exercise may benefit people with CLBP as exercises with similar features have been successful in reducing pain and improving functional ability [10][11][12].
When treating people with CLBP, it has been suggested in a Delphi survey that supervised Pilates exercise sessions should be undertaken 2-3 times per week for 3-6 months, and be supplemented by home exercises [13]. Individualised supervision has been advised in the first 2 weeks, but thereafter group sessions of up to 4 clients per therapist [13]. The use of specialised Pilates exercise equipment with spring resistance, such as a Reformer, has also been recommended for people with CLBP [13,14].
Despite the popularity of Pilates exercise in treating people with CLBP, its effectiveness in people with CLBP is yet to be established [15]. Six systematic reviews have investigated the effectiveness of Pilates exercise in people with CLBP, and a protocol for a Cochrane review has also been published [6,7,[16][17][18][19][20]. Completed reviews, though, report different findings [6,7,[16][17][18][19][20]. Several reviews report a decrease in pain, but not all report improvements in functional ability [6,[16][17][18]. Other reviews report no improvement in pain or functional ability or inconclusive findings [7,13,19]. The small number and mixed methodological quality of primary studies has made reporting of credible results difficult [15]. Several reviews have also conducted metaanalyses in the presence of significant clinical heterogeneity, resulting in misleading findings [15].
Recently, several randomised controlled trials (RCTs) have been published that are relevant to evaluating the effectiveness of Pilates exercise in people with CLBP [21][22][23][24][25][26]. The majority of these RCTs have not been included in prior reviews, so incorporating this new evidence in an updated systematic review is indicated. Given there is now a moderate volume of evidence available, it is also appropriate that this new systematic review includes only RCTs. This will ensure this review represents a high level of evidence and increases the credibility of results [15,27,28].
The aim of the following systematic review is to provide an update on the effectiveness of Pilates exercise in reducing pain and improving functional ability of people with CLBP based on the highest level and quality of research evidence available [28].

Study Design
A systematic review was undertaken to locate, evaluate and summarise findings from RCTs that have investigated the effectiveness of Pilates exercise in people with CLBP. A systematic review was chosen over a narrative review as it limits bias and error in the selection and appraisal of evidence [29,30]. In this systematic review, a comprehensive search of the literature was undertaken to answer a focused question, the methodological quality of primary studies was appraised, and findings were synthesised to address the study aim [29,30].

Search Strategy
A comprehensive literature search for evidence was undertaken on the 1st May, 2014 using 10 databases: Cumulative Index to Nursing and Allied Health Literature, Cochrane Library, Medline, Physiotherapy Evidence Database, ProQuest: Health and Medical Complete, Proquest: Nursing and Allied Health Source, Proquest: Dissertation and Theses, Scopus, Sport Discus, and Web of Science. To ensure relevant trials were not overlooked, the maximal date range available in each database was used. Medical Subject Headings terms of ''Pilates'', ''Pilates method'', and ''Low Back Pain'' and synonyms for low back pain were inputted in the title, abstract, and as able, the keyword fields to identify relevant evidence (Table 1).

Selection of Evidence
Selection of relevant studies was based on the study's title and the abstract, and as required, the full document. Two independent reviewers selected the evidence according to the selection criteria (CW, BH). Any disagreements were resolved through discussion with a third reviewer (AB). To be considered in this systematic review, studies needed to: 1. Be published in the English language, as access to interpreters was not available. 2. Be published in full so that the methodological quality of the study could be assessed alongside results. Abstracts were excluded as they contained insufficient data to enable analysis of methodological quality [31]. 3. Be RCTs to limit the risk of bias in findings regarding efficacy [26]. If studies reported that they were RCTs but did not describe the randomisation procedure they were included in this review. If, however, studies reported that they were RCTs, but described a pseudo-random technique of allocating participants to groups, such as alternative allocation, they were excluded from this review [27,28]. 4. Assess the effectiveness of Pilates exercise where the term ''Pilates'' was used to describe the type of prescribed exercise being investigated. Exercises described as ''motor control'' or ''lumbar stabilisation'' did not suffice for Pilates. This is because Pilates may include other features apart from motor control and lumbar stabilisation [9]. 5. Include participants with CLBP, that is, localised pain in the lumbar region of more than 3 months in duration [1]. If studies only included participants with low back pain of less than 3 months duration, they were excluded. This is because people with CLBP respond differently to treatment compared to those with acute or subacute symptoms [32]. If studies included participants with acute or subacute low back pain and CLBP, the study was included as findings were still considered relevant. 6. Use outcome measures with appropriate psychometric qualities that evaluate pain and/or functional ability in people with CLBP [33]. For example, the Visual Analog Scale and Numerical Rating Scale for pain, and the Oswestry Disability Questionnaire and Roland Morris Disability Questionnaire for functional ability. RCTs with outcome measures for pain and/ or functional ability that did not have sufficient validity, reliability, or responsiveness were excluded to avoid imprecise measurements of treatment effect [33].

Appraisal of Evidence
The methodological quality of included RCTs was evaluated by two independent reviewers using the McMaster Critical Review Form for Quantitative Studies (CW, BH) [34]. This critical appraisal tool was utilised because it is comprehensive in assessing methodological quality of quantitative evidence [35]. This critical appraisal tool also has good inter-rater reliability [36][37][38]. To confirm the reliability of scoring in this review, the percentage agreement and kappa score between the two reviewers was calculated [39]. Any disagreements between reviewers were resolved through discussion with a third reviewer (AB).
The McMaster Critical Review Form for Quantitative Studies directs reviewers to consider 16 items of methodological quality relating to the study's purpose, literature review, design, sample, outcomes, intervention, results and conclusions [34]. Guidelines for the appraisal of evidence in this review were created to assist reviewers in consistently evaluating methodological quality ( Table 2). These were based on guidelines provided by the authors of the McMaster appraisal tool [34,38].
If RCTs met each criterion outlined in the appraisal guidelines, they received a score of ''one'' for that item, or, if they did not meet the criteria, they received a score of ''zero''. Individual item scores were then summated to provide a total score of methodological quality out of 16, with higher scores reflecting greater methodological quality. Once quality scores were calculated, these were divided into five qualitative categories of poor (score = 0-8), fair (score = 9-10), good (score = 11-12), very good (score = 13-14) and excellent (score = 15-16) methodological quality, as defined in previous research [38].

Data Extraction and Syntheses
The number of included RCTs and their methodological quality were summarised using descriptive statistics. The author(s), year of publication, and details regarding participants, interventions, comparison treatments, outcome measures, were extracted from RCTs by the primary author (CW), and tabulated. To determine whether a meta-analysis of study findings could be performed, the clinical and statistical heterogeneity of RCTs was assessed [15,[40][41].
Clinical heterogeneity of RCTs was assessed by comparing differences in the population, intervention, comparison treatments, outcome measures, and timing of reassessment of individual studies [15]. Statistical heterogeneity was assessed by calculation of an i 2 statistic of studies with similar comparison treatment groups using the Cochrane Review Manager (version 5.2) software [41][42][43]. If i 2 was greater than 75%, studies were considered to have substantial heterogeneity [41]. If substantial clinical and statistical heterogeneity were present, pooling results in a meta-analysis was deemed inappropriate. Key findings of RCTs were expressed in terms of betweengroup mean differences and 95% confidence intervals. If betweengroup mean differences and 95% confidence intervals were not provided by RCTs, these were calculated from post-treatment mean values and standard deviations using the Physiotherapy Evidence Database calculator [44]. For the randomised cross-over trial, the between-group mean difference and 95% confidence interval was calculated from the first comparison time period between the Pilates exercise and control group, as the carryover effect in relation to time of treatment and pain intensity was statistically significant [21].
Results for each outcome measures were considered to be statistically significant if the 95% confidence interval of the between group difference did not cross ''zero'' [45]. If a 95% confidence interval was unable to be calculated from results given, a p value less than 0.05 for the between group comparison was considered to indicate statistical significance [45]. Results were considered to be clinically significant in this review if the mean between-group difference was greater than the minimal clinically important difference reported in the literature [46,47].

Search Results
A total of 267 ''hits'' were obtained with database searching, and an additional RCT was identified when reviewing reference lists of previous systematic reviews ( Figure 1) [48]. The majority of studies identified by this search strategy were excluded due to being duplicates (n = 115) or not being an RCT (n = 95). Other studies were excluded as they were not published in the English language (n = 16), did not assess the effectiveness of Pilates exercise in people with CLBP compared to other treatment (n = 17), were published in an abstract format (n = 8).

Methodological Quality
There was 95% agreement between the two reviewers regarding item scores of the McMaster Critical Review Form for Quantitative Studies. This represents ''almost perfect'' inter-rater reliability (kappa score = 0.88, p = 0.00) [39]. Discussion with the third reviewer was required to reach consensus regarding the adequacy of the description of Pilates exercise [27], presence of assessor bias [21], documentation of informed consent and ethical approval [51], use of valid and reliable outcome measures [51], and discussion of the statistical significance and clinical importance of results [48,49]. One of the RCTs was published by an author of this review (PM) [23]. To avoid bias, this author was not involved in the stages of review of this RCT [23].
The methodological quality of studies ranged from 4 to 16, representing ''poor'' to ''excellent'' methodological quality (Table 3). RCTs published in the past 2 years were generally of higher quality, as were those published in journals compared to academic theses. According to the McMaster Quantitative Review Form, strengths in the methodological quality of most RCTs related to the provision of a clear purpose (Item 1), description of participants (Item 5), and documentation of ethic approval and consent (Item 7) [34]. The majority of RCTs also provided results in terms of statistical significance (Item 11) and conducted appropriate statistical analyses (Item 12) [34].
Several RCTs, however, did not ensure assessor blinding (item 4), recruit an adequate sample size (Item 6), or document the validity and/or reliability of outcome measures (Item 8, 9) [34]. Other RCTs did not provide adequate detail of Pilates exercise programs for replication (Item 10), or discuss the clinical importance of results (Item 14) ( Table 3) [34].

Description of Included Studies
A summary of the population, intervention, comparison, and outcome measures for each RCT is provided in Table 4 and 5. The number of participants per RCT ranged from 12 to 83, while the mean age of participants across RCTs ranged between 21 to 49 years of age. The ratio of female to male participants ranged from 5:1 through to 1:1, except in one study where only females were recruited [54].
In terms of the Pilates interventions, most RCTs described supervised exercise programs delivered in 30 to 60-minute sessions, 1-3 times per week, for 4-15 weeks ( Table 4, 5). Home exercises were incorporated in 6 RCTs as part of the Pilates exercise intervention [25,26,48,52,54,55]. The supervision ratios of clients per therapist for supervised sessions ranged from 11:1 to 1:1, although not all RCTs provided this information. Use of specialised Pilates exercise equipment, such as a Reformer, was reported in 5 RCTs [21,23,26,49,55].
Variable outcome measures were used to investigate the effectiveness of Pilates exercise in reducing pain and improving functional ability in people with CLBP ( Table 4, 5). These included the Visual Analog Scale and Numerical Rating Scale (11 and 101 point scales) for pain, and the Roland Morris Disability Questionnaire, Oswestry Disability Questions, Quebec Score, and the Miami Back Index for functional ability [33,47]. Treatment outcomes were measured at different time periods, ranging from 4 to 24 weeks.

Heterogeneity of Included Studies
Significant clinical heterogeneity was noted across RCTs in terms of the study population, intervention, and outcome assessment ( Table 4, 5). Though all RCTs studied people with CLBP, some included people with acute and subacute symptoms, or other diagnoses [21,49,55]. Pilates exercise interventions also Comparison treatments also varied considerably across RCTs (Table 4, 5). When Pilates exercise was compared to usual care and physical activity, participants could access variable therapies depending on the RCT [24,48,55]. Usual physical activity also varied with some studies focusing on participants who were highly active [55,56], while the others did not. When Pilates exercise was compared to other forms of exercise, comparison exercise regimes were either similar to Pilates exercise [26,51], or quite distinct [23]. Also, in some studies, participants could access other interventions as well [26,51].
Significant statistical heterogeneity was also observed when comparing outcomes achieved with Pilates exercise versus usual care and physical activity for pain (i 2 = 90%, p,0.001), and functional ability (i 2 = 87%, p,0.00001) between 4 and 12 weeks. Similarly, significant statistical heterogeneity was noted when comparing pain relief achieved with Pilates exercise versus other forms of exercise (i 2 = 83%, p = 0.0006) between 4 and 8 weeks. Pooling individual study findings in a meta-analysis, then, was deemed inappropriate for these variables in the short term, given the clinical and statistical heterogeneity of RCTs [40,41].
Moderate statistical heterogeneity was noted when comparing improvements in functional ability with Pilates exercise and other forms of exercise in the short term (4-8 weeks) (i 2 = 44%, p = 0.17). At 24 weeks, mild statistical heterogeneity was evident when Pilates exercise was compared with other forms of exercise across the 2 RCTs for pain (i 2 = 25%, p = 0.25) and functional ability (i 2 = 9%, p = 0.29) [23,26]. Nevertheless, the clinical heterogeneity of RCTs comparing Pilates exercise to other forms of exercise suggested a meta-analysis would be of limited benefit [40,41].

Findings of Included Studies
Pilates exercise versus usual care and physical activity. Four high quality RCTs reported a statistically significant difference in pain relief with Pilates exercise in the short term (4-15 weeks) [21,24,25,55]. Two high quality RCTs, and one poor quality RCT, however, disagreed with these findings [48,50,52]. At 24 weeks, no statistically significance difference in pain relief with Pilates exercise and education versus education alone was reported (Table 6) [24].
The statistically significant improvements in pain reported in the short term were clinically significant in 3 out of 4 RCTs [21,24,25]. This is because mean difference scores exceeded the minimal clinically important difference for their respective outcome measures. For example, Borges et al. (2013) and Quinn, Barry, and Barry (2011) described a mean reduction (and 95% confidence interval) on the Visual Analog Scale of 4.1 (1.8 to 6.3) and 1.5 (0.9 to 2.1) points respectively [21,25]. The minimal clinically important difference for the Visual Analog Scale in people with CLBP has been reported in the literature as between 1.5 to 2 points [46,47].
Similarly, Miyamoto, Costa, Glavanin, and Cabral (2013) reported a mean reduction (and 95% confidence interval) of pain on the 11 point Numerical Rating Scale of 2.2 (1.1 to 3.2) points where the minimal clinically important difference was 2 points [47]. Meanwhile, Rydeard, Leger, and Smith (2006) did not report a clinically significant improvement in pain, that is a mean reduction (and 95% confidence interval) of 15.6 (13.4 to 17.8) points on the 101 point Numerical Rating Scale [55]. This is lower than the estimated minimal clinically important difference of 20 points [46].
With regards to functional ability, 3 high quality RCTs reported statistically significant improvements with Pilates exercise in the short term (4-12 weeks) [24,48,55]. In contrast, 2 high quality RCTs and 3 poor quality RCTs did not [25,50,52,53,56]. Meanwhile, at 24 weeks, no statistically significance difference in improvement in functional ability with Pilates exercise with education versus education alone was reported (Table 7) [24].
Statistically significant short-term improvements in functional ability with Pilates exercise were not clinically significant.  1 to 1.4) points on the Oswestry Disability Questionnaire which is below the minimal clinically important difference of 10 points [46,47].
Pilates exercise versus massage therapy. Only one RCT of fair quality compared the effectiveness of Pilates exercise to massage therapy [49]. This RCT did not report any statistically significant differences in pain or functional ability between groups at 6 weeks ( Table 8, 9). As a consequence, there were no clinically significant differences noted between Pilates exercise and massage therapy.
Pilates exercise versus other forms of exercise. When Pilates exercise was compared to other forms of exercise, conflicting results in terms of pain relief were reported in the short term (4-8 weeks) [23,26,51,54]. One high quality and low quality RCT reported statistically significant improvements [23,54], while another high quality and low quality RCT did not [26,51]. At 24 weeks, though, there was agreement in 2 high quality RCTs that Pilates exercise resulted in equivalent improvements in pain as other forms of exercise (Table 8) [23,26].
The improvement in pain suggested in one high quality and low quality RCT in the short term was not clinically significant [23,54]. For example, Marshall, Kennedy, Brooks, and Lonsdale (2013) reported a mean decrease (and 95% confidence interval) of 1.1 (0.1 to 2.1) points on the Visual Analog Scale, and Rajpal, Arora, and Chauhan (2009) reported a mean decrease (and 95% confidence interval) of 1.4 (0.7 to 2.1) points on the Visual Analog Scale [23,54]. These scores were less than the minimal clinically important difference of 1.5-2 points on the Visual Analog Scale [46,47].
With regards to functional ability, one high quality RCT reported a statistically significant improvement with Pilates exercise over other forms of exercise in the short term [23]. In contrast, another high quality and low quality RCT did not report a statistically significant difference in improvement [26,51]. At 24 weeks, however, 2 high quality RCTs agreed that Pilates exercise offered similar improvements in functional ability as other forms of exercise (Table 9) [23,26].
The statistically significant improvement in functional ability reported by Marshall et al. (2013) in the short term was not clinically significant [23]. This is because the mean improvement (and 95% confidence interval) in functional ability was 6.5 (1.1 to 11.8) points on the Oswestry Disability Index which is less than the minimal clinically important difference of 10 points [23,46,47].

Pilates exercise versus usual care and physical activity
Pilates exercise results in statistically significant improvements in pain and functional ability in the short term in people with CLBP. This conclusion is based on the balance of evidence, where more high quality RCTs have reported these findings [21,24,25,55]. In addition, short term improvements in pain may be clinically significant, but not improvements in functional ability [45][46][47].
Another conclusion of this review is that superior improvements with Pilates exercise compared to usual care and physical activity is unlikely at 24 weeks. This is based on research evidence of one high quality RCT that has investigated the longer term effect of Pilates exercise [24]. In this RCT, though, participants had ceased Pilates exercises at 6 weeks, and it is not known if a longer lasting effect may have been present if the intervention was continued for more than 6 weeks, as is recommended in the literature [13].
These systematic review findings are similar to those of another review in that a statistically significant reduction in pain with Pilates exercise was achieved when compared to no Pilates exercise [17]. This current review clarifies, though, that this improvement may only be in the short term, and that this change may be clinically significant. In relation to functional ability, these review findings contrast with other systematic reviews as a statistically significant improvement in functional ability in the short term was identified [6,7,17]. This difference may be due to inappropriate meta-analyses in some reviews and variable grouping of comparison treatments [13]. The size of functional improvements in RCTs in this review, however, do not appear to be clinically significant [46,47].
It should be acknowledged that not all RCTs in this review agreed regarding the effectiveness of Pilates exercise compared to usual care and physical activity [48,50,52,53]. Different results may be explained by the variable methodological quality of RCTs. For example, the majority of high quality RCTs reported statistically significant findings (4/6), while lower quality RCTs did not (2/2) [50,53].
Different results may also be due to small sample sizes or cointerventions within RCTs. Three of the 4 RCTs that did not find statistically significant findings were under-powered with small sample sizes, meaning that treatment changes may have been less easily detected [50,52,53]. The other RCT had a large sample size, but allowed the comparison group to access other interventions, such as physiotherapy and medications [48]. This may have led to the effectiveness of Pilates exercise being under estimated as the between group difference in outcome may have been reduced.
In addition, different RCT outcomes may have related to variable Pilates exercise regimes. For example, RCTs with statistically significant results prescribed supervised exercise sessions more than once a week, often with the use of specialised equipment [21,24,25,55]. It is therefore recommended that clinicians replicate Pilates exercise programs contained within RCTs with statistically significant results to maximise treatment outcomes.  Pilates exercise versus massage therapy Only one RCT compared Pilates exercise to massage therapy [49]. No statistically significant difference in improvements in pain or functional ability was noted at 6 weeks. More high quality RCTs, though, are required to confirm these findings due to the ''fair'' methodological quality of this RCT [34,38].     Pilates exercise versus other forms of exercise Based on current evidence, it is difficult to conclude on the short-term effectiveness of Pilates exercise in people with CLBP compared to other forms of exercise. Statistically significant improvements in pain and functional ability have been reported in one high quality RCT [23], but not in other high quality RCTs [26]. The clinical significance of reported statistically significant improvements is also unlikely [23,46,47]. There is consensus across high quality RCTs, though, that people with CLBP will experience equivalent improvements in pain and functional ability with Pilates exercise or alternative forms of exercise at 24 weeks [23,26].
Authors of this review therefore suggest that Pilates exercise is unlikely to provide superior improvements in pain and functional ability compared to other forms of exercise, at least in the long term. Findings of this review are similar to those of previous systematic reviews in that improvements in pain and functional ability with Pilates exercise compared to other forms of exercise have not been reported as statistically significant [6,7,17]. This  review is different, however, in that it acknowledges that there could be differences in the short term. There are two reasons why authors of this review have not ruled out the possibility of Pilates exercise offering superior short-term benefit over other forms of exercise. First, one of the two RCTs that reported no difference in the short term was of ''fair'' methodological quality [34,38]. This meant that findings were likely to be more biased than that of higher quality RCTs [38]. Second, comparison exercise treatments were variable and it could be possible that Pilates exercise is more effective than some types of exercise, but not others. When Pilates exercise was compared to a distinctly different form of exercise, cycling, there was a statistically significant difference in outcome [23]. When compared to exercises involving lumbar stabilisation, however, no difference was noted [26,51]. Future research should investigate the relative effectiveness of Pilates exercise to different forms of exercise.

Limitations
Limitations of this systematic review relate to the inclusion of only RCTs published in the English language and consequent language bias. Of the 16 studies excluded based on their language, however, only 2 appeared to be potentially relevant RCTs when reviewing titles and abstracts translated into the English language [58,59]. Another limitation was the focus of this review on outcomes of pain and functional ability in people with CLBP. Other outcomes may have also been clinically important, such as quality of life [33,60]. In addition, the methodological quality of RCTs was summarised by a total score out of 16 using the McMaster Quantitative Review Form criteria [34]. This approach can lead to oversimplification of methodological quality as all items of the scale are weighted evenly [61].
The strength of the review findings was also influenced by the availability and diversity of primary evidence. The limited number of RCTs that had compared Pilates exercise to massage therapy and other forms of exercise lessened the certainty of results [15]. The small sample sizes and short term follow up of many RCTs also affected the precision of findings [62]. Moreover, the heterogeneity of study populations, interventions, comparison treatments, outcome measures, and timing of reassessment prevented conduction of meaningful meta-analyses of RCTs [40,41].

Conclusion
According to this systematic review, Pilates exercise results in statistically significant improvements in pain and functional ability in the short term compared to usual care and physical activity in people with CLBP [21,24,25,55]. Changes in pain are more likely to be clinically significant than improvements in functional ability. At 24 weeks, though, improvements with Pilates exercise and education may be equivalent to those achieved with education alone [24]. When Pilates exercise is compared to massage therapy or other forms of exercise, equivalent improvements in pain and functional ability have been reported in people with CLBP [23,26,49].

Implications
This systematic review provides an update on the effectiveness of Pilates exercise in people with CLBP that may be used to assist clinical decision-making. Future research should investigate optimal Pilates exercise regimes for people with CLBP, including appropriate frequencies and length of programs, supervision ratios, use of home exercises, and specialised equipment [13]. Future RCTs should also investigate the long term efficacy of Pilates exercise to other treatments, such as massage, and confirm if there is any difference in effectiveness between Pilates exercise and various forms of exercise, such as aerobic exercise versus lumbar stabilisation [15]. Research into whether some people with CLBP may benefit from Pilates exercise more than others may also assist in clinical decision-making on whether Pilates exercise is suitable for individual clients [63,64].