Efficacy of Pilates on Pain, Functional Disorders and Quality of Life in Patients with Chronic Low Back Pain: A Systematic Review and Meta-Analysis

Background: Chronic low back pain (CLBP) is a common health problem. Pilates is a unique exercise therapy. This meta-analysis aims to evaluate the efficacy of Pilates on pain, functional disorders, and quality of life in patients with chronic low back pain (CLBP). Methods: PubMed, Web of Science, CNKI, VIP, Wanfang Data, CBM, EBSCO, and Embase were searched. Randomized controlled trials of Pilates in the treatment of CLBP were collected based on the inclusion and exclusion criteria. The meta-analysis was performed using RevMan 5.4 and Stata 12.2. Results: 19 randomized controlled trials with a total of 1108 patients were included. Compared with the controls, the results showed the following values: Pain Scale [standard mean difference; SMD = −1.31, 95%CI (−1.80, −0.83), p < 0.00001], Oswestry Disability Index (ODI) [mean difference; MD = −4.35, 95%CI (−5.77, −2.94), p < 0.00001], Roland–Morris Disability Questionnaire (RMDQ) [MD = −2.26, 95%CI (-4.45, −0.08), p = 0.04], 36-item Short-Form (SF-36) (Physical Function (PF) [MD = 5.09, 95%CI (0.20, 9.99), p = 0.04], Role Physical (RP) [MD = 5.02, 95%CI (−1.03, 11.06), p = 0.10], Bodily Pain (BP) [MD = 8.79, 95%CI (−1.57, 19.16), p = 0.10], General Health (GH) [MD = 8.45, 95%CI (−5.61, 22.51), p = 0.24], Vitality (VT) [MD = 8.20, 95%CI(−2.30, 18.71), p = 0.13], Social Functioning (SF) [MD = −1.11, 95%CI (−7.70, 5.48), p = 0.74], Role Emotional (RE) [MD = 0.86, 95%CI (−5.53, 7.25), p = 0.79], Mental Health (MH) [MD = 11.04, 95%CI (−12.51, 34.59), p = 0.36]), Quebec Back in Disability Scale (QBPDS) [MD = −5.51, 95%CI (−23.84, 12.81), p = 0.56], and the sit-and-reach test [MD = 1.81, 95%CI (−0.25, 3.88), p = 0.09]. Conclusions: This meta-analysis reveals that Pilates may have positive efficacy for pain relief and the improvement of functional disorders in CLBP patients, but the improvement in quality of life seems to be less obvious. Registration: PROSPERO CRD42022348173.


Introduction
Low back pain (LBP) has been a prevalent health problem for adults. The incidence is as high as 84%. In modern society, with the accelerated pace of life and the increased pressure of work, the incidence of LBP is increasing year by year. From 2006 to 2016, the incidence of LBP increased by 18% [1]. Typically, LBP is clinically characterized by pain at the lower costal margin, lumbosacral region, and buttock region, with the potential for radiating pain to the lower extremities [2]. From 1990 to 2015, the years lived with disability caused by LBP increased by 54% [3]. The lifetime prevalence of LBP is approximately 70% to 80%. About 10% to 20% of LBP patients experienced pain lasting at least 3 months and progressed to chronic low back pain (CLBP). Clinically, CLBP is treated in a variety of ways. Major non-surgical treatment methods for CLBP include pharmacotherapy, physiotherapy, and exercise therapy [4,5]. However, pharmacotherapy may cause nausea, constipation, tiredness, and other side effects, and it is difficult for physiotherapy to relieve long-term

Retrieval Strategy
This meta-analysis was planned and implemented according to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines [26]. The protocol was registered on the international prospective register of systematic reviews (http://www.crd.york.ac.uk/PROSPERO, accessed on 1 August 2022) with the registration number CRD42022348173.
In this study, PubMed, Web of Science, CNKI, VIP, Wanfang Data, CBM, EBSCO, and Embase were searched. The search time ranged from the date of database construction to November 2022. The last retrieval date is 20 November 2022. The literature search was conducted using a combination of subject terms and free terms. The search terms included "Pilates", "Pilates training", "low back pain", "back pain", "low back ache", "nonspecific low back pain", "chronic nonspecific low back pain", "chronic nonspecific lumbago", "chronic nonspecific lower back pain", "chronic nonspecific lumbar pain", and "non-specific lower back pain". In order to obtain all the randomized control trials related to Pilates intervention in the treatment of CLBP patients, we also traced the references of the retrieved literature to supplement the relevant literature. The full search strategy for each database is presented in Supplemental Table S1.

Eligibility Criteria and Outcome Indicators
The eligibility criteria were as follows: • Participants-CLBP patients (disease duration more than 3 months/12 weeks, aged 18-64 years), regardless of race and nationality, whose physical examination showed tenderness or pain in the lumbosacral region but no positive result in the straight leg raising test and the strengthening test, while excluding low back pain caused by other diseases, such as fracture, infection, and tumor, and ensuring no structural lesion in the lumbar spine in the imaging examination;

Literature Screening and Data Extraction
Step 1-import the retrieved literature to the literature management software Endno-tex9 (www.endnote.com).
Step 3-perform the first round of screening by reading titles and abstracts.
Step 4-after downloading full texts, conduct the second round of screening to determine if the inclusion criteria were met.
Two independent reviewers, ZY and YY, conducted the literature screening and data extraction. Then, cross-checking was performed. When a possible disagreement occurred, we solved it through discussion or negotiation with a third independent reviewer, XZ. In the literature screening, we first read the title to exclude the irrelevant literature. Then, we further read the abstract and the full text to determine whether to include it. If necessary, we would contact the author of the original research by email or telephone to obtain the unconfirmed information.
The extracted data were as follows: • General information of the included literature, namely the title, the first author, and the year of publication; • General characteristics of the patients, namely the number of cases in each group, the age, and the duration of the disease; • Treatment specifics and the follow-up time; • Key elements of bias risk assessment; • Focused outcome indicators.

Quality Assessment
Two independent reviewers used the Cochrane Collaboration tool to examine the risk of bias for the included studies [42,43], and cross-checking was conducted. A grading of the literature quality was performed according to the Jadad scale. A score of 1 to 3 was considered low quality, and a score of 4 to 7 was considered high quality. The grading was also conducted by two independent reviewers, with the opinions of a third independent reviewer being consulted in the event of any disagreement.

Statistical Analysis
The statistical analysis was based on RevMan5.4 (the Review Manager software 5.4, The Nordic Cochrane Center, The Cochrane Collaboration). If the results included in the literature were continuous variables and from the same assessment method, we used the mean difference (MD) and 95% confidence interval (CI) for statistics. If the results were not from the same assessment method, the standard mean difference (SMD) and 95% confidence interval (CI) were conducted. The p-value and the I 2 index were used as indicators to assess the heterogeneity among studies. There was no heterogeneity between studies when p ≥ 0.10, while p < 0.10 indicates that there was heterogeneity between studies. The I 2 index represented the degree of heterogeneity between studies. If I 2 < 50%, it indicated that there was slight heterogeneity between the studies, and the fixed effect model was used for analysis. If I 2 ≥ 50%, there was heterogeneity in the study, and the random effect model was used for analysis [44]. The α value was set at 0.05. Stata 12.0 software was used to conduct the publication bias analysis and sensitivity analysis of Begg's test for the studies with more than five included outcome indicators. The threshold for statistical significance was set at p < 0.05. The safety analysis was conducted to confirm the safety of Pilates.

Study Characteristics
A total of 1108 patients were included in the 19 RCTs. The average disease duration ranged from 86 days to 11.6 years. The sample size of each study ranged from 17 to 101 patients. The average intervention cycle is 6.8 weeks (ranging from 4 to 13 weeks), with training 3.1 times a week on average (ranging from 1 to 6 times per week). A total of 16 articles were published in the past 10 years (2013 to 2022), accounting for 84%. In 12 RCTs, the treatment method in trial groups was Pilates alone. In the remaining 7 RCTs, the treatment methods in trial groups were Pilates respectively combined with usual care, home exercise, physical therapy treatment, a standardized education program, infra-red radiation and back care, tendon puncture, the suspension training method, and non-steroidal anti-inflammatory drugs. For the controls, the treatment methods were conducted after the removal of Pilates in seven included RCTs, and there were six RCTs using the method of no treatment. Furthermore, usual care, home exercise, physical therapy treatment, as well as infra-red radiation and back care were, respectively, conducted in one, three, one, and one of the included RCTs. The details of the research characteristics are shown in Tables 1 and 2. of no treatment. Furthermore, usual care, home exercise, physical therapy treatment, as well as infra-red radiation and back care were, respectively, conducted in one, three, one, and one of the included RCTs. The details of the research characteristics are shown in Tables 1 and 2.    12 w, 24 w Miyamoto, G.C. 2013 [23] Pilates X 2 t/w, 6 w 1 3 6 m Notarnicola, A. 2014 [46] Pilates X 60 min/t, 5 t/w, 6 m 2 3 4 -Mostagi, F.Q. 2015 [47] Pilates III 2 t/w, 8 w 6 3 m Natour, J. 2015 [48] Pilates and IV IV 50 min/t, 2 t/w, 90 d 1 3 4 5 180 d Akodu, A. 2016 [49] Pilates [50] Pilates and III III 2 t/w, 6 w 1 2 1 y Cruz-Díaz, D. 2017 [36] PMG/PAG X 50 min/t, 2/w, 12 w 1 3 -Lopes, S. 2017 [22] Pilates X 20 min/t 1 -Mazloum, V. 2018 [12] Pilates X 3 t/w, 6 w  [12,[19][20][21]46,50,52,53,55] and eight [20,23,36,46,48,49,51,54] RCTs, respectively. In addition, SF-36, the sit-and-reach test, and QBPDS were separately performed in four [45,46,48,53], 3 [21,48,53] and two [47,53] RCTs.
The risk of bias assessment was performed using RevMan5.4 software, according to the Cochrane Handbook for Systematic Reviews. The results are shown in Figures 2 and 3. The quality of the literature was graded according to the Jadad scale, with two studies judged to be of low quality and the remaining studies considered to be of high quality. The details are presented in Supplemental Table S2.  [12,[19][20][21]46,50,52,53,55] and [20,23,36,46,48,49,51,54] RCTs, respectively. In addition, SF-36, the sit-and-reach test QBPDS were separately performed in four [45,46,48,53], 3 [21,48,53] and two [47,53] R The risk of bias assessment was performed using RevMan5.4 software, accordi the Cochrane Handbook for Systematic Reviews. The results are shown in Figures 2 and 3 quality of the literature was graded according to the Jadad scale, with two studies ju to be of low quality and the remaining studies considered to be of high quality. The d are presented in Supplemental Table S2.

Results of Meta-Analysis
According to the results of the heterogeneity assessment, the random effects model meta-analysis was conducted to analyze the efficacy of Pilates in the context of the Pain Scale, ODI, and RMDQ of CLBP patients. The results showed that there were statistically significant differences between trial groups and control groups, which indicated that Pi-

Results of Meta-Analysis
According to the results of the heterogeneity assessment, the random effects model meta-analysis was conducted to analyze the efficacy of Pilates in the context of the Pain Scale, ODI, and RMDQ of CLBP patients. The results showed that there were statistically significant differences between trial groups and control groups, which indicated that Pilates              in CLBP patients were conducted. The results showed that there was a significant difference in the sit-and-reach test, but not QBPDS, between trial groups and control groups, which indicated that Pilates had a positive effect on the sit-and-reach test but not QBPDS in CLBP patients. The results of the meta-analysis are presented in Figures 8 and 9. in CLBP patients were conducted. The results showed that there was a significant difference in the sit-and-reach test, but not QBPDS, between trial groups and control groups, which indicated that Pilates had a positive effect on the sit-and-reach test but not QBPDS in CLBP patients. The results of the meta-analysis are presented in Figures 8 and 9.

Follow-Ups Analysis
To determine the long-term effects of Pilates on chronic low back pain, we performed a meta-analysis for Pain Scale, ODI, RMDQ, and SF-36 in follow-ups. A total of six RCTs followed the Pain Scale [12,23,45,48,50,54]; a random effects model meta-analysis showed that Pilates was superior to the controls in long-term pain relief in CLBP patients, the difference was statistically significant [MD = −0.70, 95%CI (−1.38, −0.03), p = 0.04]. In two RCTs [12,50], the follow-ups involved ODI. A random effects model meta-analysis showed that the decrease in ODI scores was more pronounced in CLBP patients after the treatment of Pilates compared with the controls, with a statistically significant difference [MD = −6.66, 95%CI (−13.12, −0.86), p = 0.03]. Furthermore, there were three RCTs [23,48,54] which reported RMDQ in follow-ups, and a fixed effects model meta-analysis showed that Pilates had a more positive effect on the decrease in RMDQ scores in CLBP patients in the comparison with the controls, with a statistically significant difference [MD = −1.97, 95%CI (−3.53, −0.40), p = 0.01]. For RCTs with follow-ups on SF-36, we included a total of two [45,48]. A fixed effects model meta-analysis showed that there was no statistically significant difference in all dimensions of SF-36 between the trial groups and control groups. The details are presented in Supplementary Figures S1-S4.

Publication Bias and Sensitivity Analysis
Begg's test was conducted to analyze publication bias for the outcome indicators which were involved in five or more RCTs. The results showed that there was a publication bias risk for Pain Scale (t = −2.98, p = 0.008, p < 0.05), and there was no publication bias for ODI (t = −0.89, p = 0.405, p > 0.05) and RMDQ (t = −0.42, p = 0.686, p > 0.05). The details are presented in Supplementary Figures S5-S7.
We conducted a sensitivity analysis through one-by-one elimination for Pain Scale, ODI, and RMDQ in the included RCTs. After excluding any RCTs, there was no significant change in the pooled results and the results were stable. The details are presented in Sup- in CLBP patients were conducted. The results showed that there was a significant difference in the sit-and-reach test, but not QBPDS, between trial groups and control groups, which indicated that Pilates had a positive effect on the sit-and-reach test but not QBPDS in CLBP patients. The results of the meta-analysis are presented in Figures 8 and 9.

Follow-Ups Analysis
To determine the long-term effects of Pilates on chronic low back pain, we performed a meta-analysis for Pain Scale, ODI, RMDQ, and SF-36 in follow-ups. A total of six RCTs followed the Pain Scale [12,23,45,48,50,54]; a random effects model meta-analysis showed that Pilates was superior to the controls in long-term pain relief in CLBP patients, the difference was statistically significant [MD = −0.70, 95%CI (−1.38, −0.03), p = 0.04]. In two RCTs [12,50], the follow-ups involved ODI. A random effects model meta-analysis showed that the decrease in ODI scores was more pronounced in CLBP patients after the treatment of Pilates compared with the controls, with a statistically significant difference [MD = −6.66, 95%CI (−13.12, −0.86), p = 0.03]. Furthermore, there were three RCTs [23,48,54] which reported RMDQ in follow-ups, and a fixed effects model meta-analysis showed that Pilates had a more positive effect on the decrease in RMDQ scores in CLBP patients in the comparison with the controls, with a statistically significant difference [MD = −1.97, 95%CI (−3.53, −0.40), p = 0.01]. For RCTs with follow-ups on SF-36, we included a total of two [45,48]. A fixed effects model meta-analysis showed that there was no statistically significant difference in all dimensions of SF-36 between the trial groups and control groups. The details are presented in Supplementary Figures S1-S4.

Publication Bias and Sensitivity Analysis
Begg's test was conducted to analyze publication bias for the outcome indicators which were involved in five or more RCTs. The results showed that there was a publication bias risk for Pain Scale (t = −2.98, p = 0.008, p < 0.05), and there was no publication bias for ODI (t = −0.89, p = 0.405, p > 0.05) and RMDQ (t = −0.42, p = 0.686, p > 0.05). The details are presented in Supplementary Figures S5-S7.
We conducted a sensitivity analysis through one-by-one elimination for Pain Scale, ODI, and RMDQ in the included RCTs. After excluding any RCTs, there was no significant change in the pooled results and the results were stable. The details are presented in Supplementary Table S3.

Follow-Ups Analysis
To determine the long-term effects of Pilates on chronic low back pain, we performed a meta-analysis for Pain Scale, ODI, RMDQ, and SF-36 in follow-ups. A total of six RCTs followed the Pain Scale [12,23,45,48,50,54]; a random effects model meta-analysis showed that Pilates was superior to the controls in long-term pain relief in CLBP patients, the difference was statistically significant [MD = −0.70, 95%CI (−1.38, −0.03), p = 0.04]. In two RCTs [12,50], the follow-ups involved ODI. A random effects model meta-analysis showed that the decrease in ODI scores was more pronounced in CLBP patients after the treatment of Pilates compared with the controls, with a statistically significant difference [MD = −6.66, 95%CI (−13.12, −0.86), p = 0.03]. Furthermore, there were three RCTs [23,48,54] which reported RMDQ in follow-ups, and a fixed effects model meta-analysis showed that Pilates had a more positive effect on the decrease in RMDQ scores in CLBP patients in the comparison with the controls, with a statistically significant difference [MD = −1.97, 95%CI (−3.53, −0.40), p = 0.01]. For RCTs with follow-ups on SF-36, we included a total of two [45,48]. A fixed effects model meta-analysis showed that there was no statistically significant difference in all dimensions of SF-36 between the trial groups and control groups. The details are presented in Supplementary Figures S1-S4.

Publication Bias and Sensitivity Analysis
Begg's test was conducted to analyze publication bias for the outcome indicators which were involved in five or more RCTs. The results showed that there was a publication bias risk for Pain Scale (t = −2.98, p = 0.008, p < 0.05), and there was no publication bias for ODI (t = −0.89, p = 0.405, p > 0.05) and RMDQ (t = −0.42, p = 0.686, p > 0.05). The details are presented in Supplementary Figures S5-S7.
We conducted a sensitivity analysis through one-by-one elimination for Pain Scale, ODI, and RMDQ in the included RCTs. After excluding any RCTs, there was no significant change in the pooled results and the results were stable. The details are presented in Supplementary Table S3.

Safety Analysis
Only one RCT [36] reported on safety and adverse events in patients after Pilates. In this study, the patient's symptoms in the lumbar spine showed no aggravation, but we cannot determine the safety of Pilates.

Discussion
In this meta-analysis, we included 19 RCTs, with 1108 CLBP patients. The results showed that Pilates had a positive effect on Pain Scale, ODI, RMDQ and the sit-and-reach test, but that it had no obvious improvement on most dimensions of SF-36 and QBPDS. This suggested a beneficial effect of Pilates on the relief of pain and improvement of functional disability in patients with chronic low back pain, with little effect on the quality of life. Furthermore, the results of the follow-up analysis revealed that the effect of relieving pain and improving functional disability was still maintained in the future period after Pilates treatment. The results on the relief of pain in CLBP patients by Pilates were approximately consistent with those of three previous systematic reviews on Pilates in the treatment of CLBP mentioned in the text. However, one of the systematic reviews summarized that Pilates is no better than other types of exercise in reducing pain in the short term [23]. This negative result, which differs from this meta-analysis, may be caused by the difference in the included studies due to the difference in the inclusion and exclusion criteria. Moreover, this past systematic review may have included a smaller number of studies.
Usually, patients with acute low back pain would experience better improvement and relief in pain as well as dysfunction within 6 weeks [56]. However, acute low back pain would develop into chronic low back pain in approximately 40% of patients, and back pain, as well as functional disability, will persist for more than 12 weeks [6,57]. Furthermore, CLBP patients usually suffered from muscle atrophy in core stable muscle groups (erector spinae, quadratus lumborum, rectus abdominis, internal and external oblique, transversus abdominis and multifidus, etc.) and decreased muscle strength and muscle endurance. Over time, muscle group coordination to maintain strength and stability was dysregulated, with spinal motor flexibility decreased and postural control dysfunction appearing. In exercise, unnecessary compensatory phenomena occurred to maintain spinal stability, and the threshold of muscle fatigue was reduced, leading to rapid fatigue [58,59]. In addition, due to the neurological control disorder of CLBP patients, the local stabilizing muscles started relatively late during the body movement, and the muscle recruitment was delayed, which leads to a further decrease in spinal stability [60]. Long-term low back pain would also lead to core reflex inhibition and nervous system control disorders, resulting in lumbar stability decline [61][62][63]. In this way, patients with chronic low back pain fell into a vicious circle of "muscle atrophy-pain-activity restriction".
Pilates is a unique training method in that it follows six important training principles, consisting of centering (i.e., activation of core muscle groups), concentration (i.e., cognitive attention when performing exercises), control (i.e., postural control management during exercises), precision (i.e., exercises with few repetitions but precise movements, emphasizing the quality of the exercise), breathing (i.e., the coordination of movements and breathings in exercises, promoting the activation of deep trunk muscles), and flow (i.e., smoothness during exercises and flowing transition between consecutive exercises) [14,15,64]. Pilates focuses on the activation of deep core muscles, spinal stabilization exercises and body postural controls [65], which in turn enhances the muscle strength and muscle endurance of deep muscle groups, increases the interaction between them and the whole muscle, reduces joint compression and changes pelvic inclination, improves postural control, and improves the stability of the spine, so that the spine is upright and neutral in the pelvis, and decreases the perceived force of the lumbar spine on the stimuli sent by nociceptors, thereby achieving pain relief [20,55]. It may be precisely, therefore, that Pilates may relieve the pain in patients with chronic low back pain. Furthermore, Pilates emphasizes both the local stability movement of the spine and the overall movement of the body, as well as strengthening the nerve control of the core muscle group and sending out the command of voluntary movement through the vertebral bundle to control the musculoskeletal system to correct body deviation in real time to maintain stability, gradually form the correct sensorimotor ability, and improve lumbar dysfunction [62,66,67]. Probably because of these, Pilates promotes the improvement of functional disorders in patients with chronic low back pain.

Conclusions and Suggestions
In view of the results of this meta-analysis, Pilates seems to have positive efficacy for pain relief and the improvement of functional disorders in patients with chronic low back pain, and the effects of treatment may be maintained for a period. However, there appears to be no positive effect on quality of life in CLBP patients. Furthermore, the safety of Pilates could not be determined.
It is suggested that a more uniform and standardized study design and treatment protocol should be established in future studies, while the doses of Pilates should also be intensively studied. Furthermore, measurements of objective instruments, such as electromyography and nuclear magnetic resonance, should also be added, in order to explore changes in neuromuscular regulation and imaging, and further verify the authenticity of the efficacy.

•
In this meta-analysis, the study languages of the included RCTs were only Chinese and English, and the study sample sizes were relatively small, which may have biased the results.

•
In the included RCTs, the exercise methods of Pilates were not all consistent, and the treatment methods of the controls were not all consistent.

•
There was high heterogeneity among small parts of the literature, which may have caused some influences on the reliability of the meta-analysis.

•
The bias risk in Pain Scale, according to the publication bias analysis, may be due to the difference in the pain assessment methods among studies, which may lead to an effect on the reliability of study results.
Supplementary Materials: The following supporting information can be downloaded at: https: //www.mdpi.com/article/10.3390/ijerph20042850/s1, Figure S1: Meta-analysis of the effect of Pilates on Pain Scale follow-up; Figure S2: Meta-analysis of the effect of Pilates on ODI follow-up; Figure S3: Meta-analysis of the effect of Pilates on RMDQ follow-up; Figure S4: Meta-analysis of the effect of Pilates on SF-36 follow-up; Figure S5: Pain Scale publication bias graph of the included RCTs; Figure S6: ODI publication bias graph of the included RCTs; Figure S7: RMDQ publication bias graph of the included RCTs;   Data Availability Statement: The data are not publicly available for privacy reasons.

Conflicts of Interest:
The authors declare no conflict of interest.