Impact of isolated lumbar extension strength training on reducing nonspecific low back pain, disability, and improving function: a systematic review and meta-analysis

Abstract

Nonspecific low back pain (LBP), a prevalent condition with a lifetime prevalence of up to 84%, presents a considerable burden on individuals and healthcare systems. Isolated lumbar extension (ILEX) has been studied for its ability to exercise the lumbar region with more controlled activation of the erector spinae and other paravertebral muscles. It aims to serve as a specific resistance training method to improve outcomes related to pain, disability, and physical functionality in adults with nonspecific low back pain (LBP). This systematic review and meta-analysis aimed to evaluate and summarize the effects of ILEX in alleviating pain, reducing disability, and improving physical functionality in adults with chronic LBP. Searches were conducted on October 14, 2024, across key databases, including PubMed, Scopus, and Web of Science. Eligibility criteria included adults (> 18 years old) with chronic LBP participating in resistance training focused on ILEX, with comparators comprising true control or active control groups, all from randomized clinical trials. The RoB2 was used to assess the risk of bias in the studies, while the GRADE scale was employed to evaluate the certainty of the evidence. The meta-analysis calculated Hedges’ g effect sizes (ES) with 95% CIs and PIs for main outcomes, using the DerSimonian and Laird random-effects model to address inter-study variability, I² for heterogeneity, and the extended Egger’s test for publication bias, all performed with SPSS Software. After screening, a total of 8 randomized studies were included, with 381 participants overall. The results indicated a significant favorable effect of ILEX compared to the true control group in pain-related outcomes (ES = − 0.633, p = 0.004). However, there was a non-significant effect of ILEX compared to the true control group in disability-related outcomes (ES = − 0.292, p = 0.190) and isometric strength outcomes (ES = 0.967, p = 0.150). The GRADE scale indicated that the certainty of the evidence is very low.ILEX significantly reduces pain intensity in individuals with low back pain, indicating its potential as an effective intervention, but its impact on disability and physical functionality is less consistent, warranting cautious use alongside pelvic stabilization to optimize rehabilitation outcomes.

Supplementary Information

The online version contains supplementary material available at 10.1038/s41598-025-90699-5.

Introduction

Nonspecific low back pain (LBP) is a common condition that affects a significant portion of the population, with a lifetime prevalence of up to 84%¹. Key concerns associated with LBP include the intensity of pain localized between the costal margin and the inferior gluteal folds, which may occur with or without accompanying leg pain². This pain can cause substantial functional impairments, negatively impacting daily activities, work capacity³, and contributing to a decline in quality of life⁴. Epidemiological studies indicate that chronic LBP is widespread across all age groups, although it is most frequently reported in adults aged 40 to 50 years⁵. Despite its high prevalence, nonspecific LBP generally lacks identifiable structural abnormalities, which complicates both diagnosis and treatment strategies⁶.

While previous studies have suggested that passive treatments—such as ultrasound, heat and cold therapy, and massage—are generally ineffective in reducing pain in adults with LBP^7,8, guideline-endorsed long-term therapeutic approaches emphasize the importance of physical rehabilitation. Exercise and resistance-based training are particularly effective strategies⁹. Evidence indicates that regular physical exercise, encompassing both aerobic and resistance training, significantly alleviates pain and enhances function in individuals with chronic LBP¹⁰. Notably, resistance training, which focuses on strengthening the core and back muscles, is associated with greater functional improvements and pain relief compared to other therapeutic modalities¹¹. While a recent network meta-analysis has highlighted the specific benefits of Pilates for pain relief, it appears that resistance training, along with stabilization and motor control exercises, is also particularly effective¹¹. Furthermore, consistent participation in structured exercise programs is linked to decreased healthcare utilization and reduced reliance on medications¹², underscoring its critical role in the management of chronic LBP.

Among the resistance training, isolated lumbar extension (ILEX) training has gained attention as a targeted approach to low back pain therapy, particularly due to its unique method of restraining the pelvis¹³. This technique focuses on isolating the lumbar spine’s movement while stabilizing the pelvis, which allows for a more controlled activation of the erector spinae and other paravertebral muscles¹⁴. Research indicates that ILEX can effectively enhance lumbar strength and stability without placing excessive strain on the intervertebral discs¹⁵, making it a suitable option for individuals with LBP. Studies have shown that ILEX not only improves muscle endurance but also reduces pain¹⁶ and disability levels¹³ among chronic low back pain patients. Compared to conventional resistance training methods that engage multiple muscle groups, ILEX provides a more specific strengthening stimulus to the lumbar region. This focused approach aligns with current rehabilitation guidelines advocating for individualized exercise interventions tailored to the specific needs of patients suffering from LBP¹⁷.

Although numerous systematic reviews have explored exercise therapy for LBP, the evidence regarding ILEX remains outdated, with the most comprehensive review published in 2015¹⁸ and included studies involving both healthy and non-healthy patients. As a result, it failed to provide a clear summary of the impact on LBP populations. Additionally, this review¹⁸ was narrative in nature and did not adhere to methodological standards like PRISMA guidelines or conduct a meta-analysis. A contemporary meta-analysis was crucial as it synthesized data from various studies to provide a more precise estimate of ILEX efficacy, identified patterns and variability in treatment outcomes, and assessed specific intervention characteristics that enhanced effectiveness. Furthermore, it addressed the quality and consistency of existing research, ultimately contributing to updated clinical guidelines that could improve patient care. Thus, this systematic review and meta-analysis aimed to evaluate and summarize the effects of ILEX in alleviating pain, reducing disability, and improving physical functionality in adults with chronic LBP.

Methods

Our systematic review followed the guidelines set out in the 2020 PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) Statement¹⁹.

Protocol and registration

The protocol for this systematic review was published on the Open Science Framework (registration: osf.io/43xgf) on 13/10/2024. It can be accessed via DOI: 10.17605/OSF.IO/43XGF.

Eligibility criteria

The eligibility criteria followed the PICOS (Participants, Intervention, Comparator, Outcomes, Study Design) framework detailed in Table 1. The inclusion criteria were limited to original research articles published in peer-reviewed journals, with no restrictions on the year of publication²⁰ or the language of the articles.

Table 1.

Inclusion and exclusion criteria based on PICOS.

	Inclusion Criteria	Exclusion Criteria
Population	The study will include both men and women aged 18 years and older who have nonspecific chronic low back pain.	Individuals under 18 years of age and those with non-chronic low back pain will be excluded from the review. Additionally, studies will be excluded if they focus on pain associated with pregnancy, infection, tumors, osteoporosis, fractures, structural deformities (e.g., scoliosis), inflammatory disorders, radicular syndrome, or cauda equina syndrome.
Intervention or	Intervention studies lasting at least 4 weeks that include a minimum of one supervised session per week will be included. The intervention must consist of a specific lumbar training program utilizing a restraint system designed to facilitate isolated lumbar extension resistance training.	Resistance interventions that do not include isolated lumbar extension resistance training will be excluded. Furthermore, other training modalities, such as Pilates, yoga, and those focused on flexibility or aerobic exercise, will also be excluded.
Comparator	Both true control groups (i.e., those not enrolled in any form of exercise or rehabilitation) and other non-exercise control groups will be included. Non-exercise controls may involve rehabilitation methods such as manual therapies (e.g., chiropractic, passive physiotherapy, osteopathy, massage, or acupuncture) or hands-off approaches (e.g., general practitioner management, education, or psychological interventions). Additionally, exercise controls that do not involve isolated lumbar extension training, such as yoga, Pilates, aerobic exercise, flexibility training, or multimodal interventions, will also be included.	Interventions that include isolated lumbar extension as part of a broader resistance training protocol will be excluded.
Outcomes	The review will include outcomes related to pain intensity (e.g., scores on the Visual Analog Scale or Numeric Pain Rating Scale), disability (e.g., scores on the Oswestry Disability Questionnaire, Roland-Morris Disability Questionnaire), and physical functionality (e.g., range of motion, strength levels or flexibility), which will be assessed at two time points: before and after the intervention.	Assessments that report only one time point, as well as other health-related outcomes such as mental health (e.g., depression), biomarkers, cognitive measures, and social-related outcomes (e.g., social interaction or support), will be excluded.
Study design	Two-arm randomized controlled study design.	Non-randomized experimental study designs, descriptive studies, systematic reviews, and grey literature will be excluded.

Information sources

We conducted a comprehensive search for relevant studies using the PubMed, Scopus, and Web of Science (Core Collection) databases, on 14 October 2024, following the protocol registration. Additionally, we manually reviewed the reference lists of the included studies to uncover any additional relevant sources.

To further strengthen the review process, we applied snowball citation tracking via the Web of Science database. To ensure rigor, we sought feedback from two globally recognized experts, identified through Expertscape for their expertise in Low Back Pain (https://expertscape.com/ex/low+back+pain). All studies selected for inclusion were also thoroughly checked for any associated errata or retractions [44].

Search strategy

The search strategy utilized a combination of Boolean operators “AND” and “OR” to maximize the retrieval of relevant studies. No filters or limitations were applied for publication date, language, or study design, ensuring a broad and comprehensive search. This approach aimed to capture as many relevant studies as possible without restricting the scope. The specific search methodology was as follows:

[Title/Abstract] back* OR lumbar OR lumbosacral OR sacral OR spinal

AND

[Title/Abstract] ache* OR pain* OR backache* OR disability*

AND

[Title/Abstract] resistance OR resisting OR exercise* OR strength*

AND

[Title/Abstract] “lumbar exten*” OR “isolated lumbar”

The complete search strategy for each database is detailed in Supplementary Material 1.

Selection process

In the initial stage of the research process, studies were screened based on their titles and abstracts by two independent authors (RT and FMC). The abstracts of the selected studies were assessed according to the established inclusion criteria, and full-text articles were obtained as needed. During the subsequent phase, the full texts of the studies that passed the initial screening were evaluated separately by the same two authors. If any disagreements occurred during either phase, the reviewers engaged in further discussion. Should they fail to reach an agreement, a third reviewer (WM) was brought in to help reach a consensus. To efficiently manage records and remove duplicates, a combination of manual and automated methods was utilized, aided by EndNoteTM software (version 20.5, Clarivate Analytics, Philadelphia, PA).

Data collection process

One of the authors (FMC) initiated the data extraction process, which was subsequently reviewed by two additional authors (RT and WM) to ensure accuracy and completeness. To aid in this endeavor, a specialized Microsoft Excel spreadsheet (Microsoft®, USA) was created to capture all pertinent data.

In cases where the full-text articles lacked certain data, one of the authors (FMC) reached out to the corresponding authors via email and ResearchGate to obtain the required information. If there was no reply from the corresponding author after two weeks, the data from those studies were removed from both the review and the meta-analysis.

The key details extracted from each included study included: (i) sample size; (ii) information regarding physical activity (e.g., active, inactive, athlete status), age, gender, and clinical condition; and (iii) relevant aspects of the study design, including randomization techniques and contextual elements such as blinding procedures.

To summarize the training intervention information, the following data were collected: (i) training duration and frequency; (ii) the overall number of training sessions conducted; (iii) participant adherence; and (iv) the training regimen implemented as specified in each study (e.g., sets, repetitions, exercises, and equipment used).

Data items

Three main categories of data items were collected: (i) pain-related outcomes, (ii) disability-related outcomes, and (iii) physical functionality-related outcomes. The pain-related outcomes included various measures of pain perception, obtained through different questionnaires such as the Visual Analog Scale, Numeric Pain Rating Scale, and McGill Pain Questionnaire. For disability-related outcomes, the data included scores from several questionnaires, including the Oswestry Disability Index, Roland-Morris Disability Questionnaire, Health Assessment Questionnaire, and Quebec Back Pain Disability Scale. Finally, for physical functionality-related outcomes, the data encompassed measures such as range of motion (e.g., assessments using goniometry or specific tests like the Trunk Flexion Test), strength levels (e.g., evaluations from isokinetic and isometric tests, as well as Functional Strength Tests like the Sit-to-Stand Test and Single Leg Stance Test), and flexibility levels (e.g., assessments from tests such as the Sit-and-Reach Test and Thomas Test).

Quality of studies

The quality of the studies was assessed using the Physiotherapy Evidence Database (PEDro) scale. The scale has been used in physiotherapy-based randomized clinical trials and has previously been validated for its reliability and effectiveness. The scale comprises 11 items that evaluate key aspects of study design and reporting, including random allocation, concealed allocation, baseline comparability, blinding of participants and assessors, retention rates, intention-to-treat analysis, and the reporting of outcomes. Each item is scored as either “yes” (1 point) or “no” (0 points), except for items 1 and 2, which are not included in the total score, leading to a maximum score of 10. Scores of 3 or below indicate poor study quality, scores between 4 and 5 reflect fair quality, scores from 6 to 8 represent good quality, and scores of 9 or higher indicate excellent quality²². Each of the included studies was independently evaluated by two of the authors (RT and FMC) using the PEDro scale. If the two authors could not reach a consensus on the assigned score after discussion, a third author (AK) was consulted to help make a final decision. An inter-rater reliability analysis was conducted using Cohen’s kappa statistic, revealing a high level of agreement among the authors (κ = 0.91).

Risk of bias

The risk of bias in the included randomized controlled trials was evaluated using the Cochrane risk-of-bias tool for randomized trials, version 2 (RoB2)²³. This tool assesses five potential sources of bias: (i) the randomization process; (ii) deviations from the intended interventions; (iii) missing outcome data; (iv) the measurement of outcomes; and (v) the selection of reported results. A qualitative synthesis was carried out based on the RoB2 framework. Two authors (FMC and RT) independently conducted the risk-of-bias assessment, with any disagreements resolved through discussion and, when necessary, by consulting a third author (WM).

Summary measures, synthesis of results, and risk of publications bias

Meta-analyses were conducted when at least three separate studies provided effect size data related to the same category of data items (i.e., pain, disability, and physical functionality), regardless of the specific test or variable measured²⁴.

Hedges’ g effect sizes (ES), along with 95% confidence intervals (CI) and 95% prediction intervals (PI), were calculated for main outcomes in both the ILEX training and comparison groups. These effect sizes were based on the mean and standard deviation values from measurements taken before and after the interventions, with standardization using post-intervention standard deviations. In this meta-analysis, we employed the DerSimonian and Laird random-effects model to account for inter-study variability and enhance the reliability of the overall results. This method is particularly effective in addressing differences across studies that might influence small-study effects (SSE)^25,26.

Effect size (ES) values were reported with 95% confidence intervals (CIs) and interpreted according to the following scale: 0.0-0.2 as trivial, 0.2–0.6 as small, > 0.6–1.2 as moderate, > 1.2-2.0 as large, > 2.0–4.0 as very large, and > 4.0 as extremely large²⁷. For studies with multiple intervention groups, the control group sample size was proportionally adjusted to allow for fair comparisons across groups²⁸. To evaluate heterogeneity, we employed the I² statistic, classifying values as follows: <25% indicating low, 25–75% indicating moderate, and > 75% indicating high heterogeneity²⁹.

To evaluate the potential for publication bias in continuous variables (with at least 10 studies per outcome), we used the extended Egger’s test³⁰. To correct for any bias detected, a sensitivity analysis was performed using the trim and fill method³¹ with L0 as the default estimator for identifying missing studies³². All statistical analyses were performed using SPSS (version 28.0, IBM Corp. Released 2021. IBM SPSS Statistics for Windows, Armonk, NY: IBM Corp), with significance defined at p ≤ 0.05.

Certainty assessment

Using the Grading of Recommendations Assessment, Development, and Evaluation (GRADE) approach³³, two authors (RT and FMC) independently assessed the quality of the evidence, reaching agreement through discussion when discrepancies arose. The evaluation focused on four of the five core GRADE criteria^34,35: risk of bias, inconsistency, imprecision, and likelihood of publication bias.

Evidence from randomized controlled trials starts as high-certainty. Depending on these criteria, the certainty of evidence can be downgraded (e.g., for high risk of bias, inconsistency of results, or indirect evidence) or upgraded (e.g., for a large effect size or a dose-response gradient). The final score is categorized into four levels: high, moderate, low, or very low certainty of evidence, reflecting confidence in the estimated effect.

Results

Study selection

A search across Web of Science, Scopus, and PubMed databases initially yielded 757 studies. After removing duplicates (n = 339), the remaining 418 records were screened based on their titles and abstracts. Of these, 359 were excluded at this stage, leaving 59 studies for full-text review. Following this evaluation, 51 studies were excluded for not meeting eligibility criteria: excluded due to not meeting the eligibility criteria for the population (n = 35); excluded due to not meeting the eligibility criteria for the intervention (n = 11); and excluded due to not meeting the eligibility criteria for the intervention (n = 5). A comprehensive list of included and excluded studies, along with the reasons for exclusion, can be found in Supplementary Material 2. Ultimately, 8 original studies were incorporated into the meta-analysis, as illustrated in Fig. 1.

Fig. 1.

Prisma flowchart.

Study characteristics

Table 2 presents the key characteristics of the included studies. Among the included studies, a total of 381 participants were involved, with the individual studies ranging from a minimum of 24 participants¹⁶ to a maximum of 107 participants³⁶. All the studies focused on adults, as the average age of participants was approximately 41.5 years, with a pooled standard deviation of 11.5 years. While five studies^16,37–40 included both men and women, two studies^36,41 focused exclusively on men, and none concentrated on women. The remaining studies¹³ did not specify the sex of the participants.

Table 2.

Characteristics of the included studies.

Study	N	LBP duration	Age	Sex	Control group	Pain-related tests/outcomes	Disability-related tests/outcomes	Physical functionality-related tests/outcomes
Bruce-Low et al. 37	41	≥ 6 months	45.5 ± 14.1	Women & men	True control	Visual analogue scale#	Oswestry disability index#	Schober’s test (cm) and ROM (º); Isometric lumbar extension strength tests (Nm)
Fortin et al. 38	50	≥ 3 months	41.4 ± 11.1	Women & men	Active control	Numerical Pain Rating Scale	Oswestry Disability Index	Multifidus and erector spinae muscle cross-sectional area (cm2)
Harts et al. 41	39	≥ 3 months	42.4 ± 9.6	Men	True control	-	Roland-Morris Disability Questionnaire	Isometric back extension strength (Nm)
Helmhout et al. 36	107	≥ 4 weeks	36.1 ± 11.0	Men	Active control	-	Roland-Morris Disability Questionnaire; Patient-Specific Functional Scale	Isometric (net) muscular torque in lumbar extensors (Nm)
Risch et al. 39	54	≥ 1 year	45.0 ± ND	Women & men	True control	West-Haven Yale Multidimensional Pain Inventory	-	Isometric strength of the lumbar extensor (Nm)
Smith et al. 42	42	≥ 6 months	42.9 ± 10.8	ND	True control and active control	Visual analogue scale	Oswestry Disability Index	Maximal voluntary isometric torque in lumbar extension (Nm)$
Steele et al. 35	24	≥ 3 months	43.5 ± 14.7	Women & men	True control	Visual analogue scale	Oswestry Disability Index	Maximal isometric lumbar extension strength (Nm); Lumbar ROM (º); Schobers standing ROM (cm)
Steele et al. 40	24	≥ 3 months	45.5 ± 13.6	Women & men	True control	Visual analogue scale	Oswestry Disability Index	Maximal isometric lumbar extension strength (Nm); Stride to stride waveform pattern (CV)

ND: not described; ROM: range of motion; LBP: low back pain

^#The available data could not be extracted due to the absence of baseline information, as it only reported the mean differences.

^$The available data only reported means, without including standard deviations. True control: those not enrolled in any form of exercise or rehabilitation; Active control: parallel groups exposed to an alternative physical exercise intervention or non-exercise rehabilitation methods.

Regarding pain-related tests and outcomes, the Visual Analog Scale was the most used, featured in four studies^13,16,37,40. For disability-related tests and outcomes, the Oswestry Disability Index was the most prevalent, utilized in five studies^{13,16,37,38,40}. Finally, concerning physical functionality tests, maximum isometric strength was the most frequently measured, appearing in seven studies^{13,16,36,37,39–41}.

Table 3 provides a concise overview of the training protocols implemented in the included studies. The training duration ranged from a minimum of 8 weeks⁴¹ to a maximum of 12 weeks^{13,16,37,38,40}, with a mode of 12 weeks. The training frequency ranged from 1^13,16,40 to 2^36,38 sessions per week, with some studies initially scheduling 2 sessions and then reducing to 1 session as the intervention progressed^37,39. The total number of sessions varied between a minimum of 12³⁷ and a maximum of 24 ^13,16,38,40. The MedX machine was the most used equipment for promoting ILEX in the included studies^{13,16,38–40}.

Table 3.

Characteristics of the training protocols.

Study	Duration (weeks)	Frequency (days/week)	Total sessions (n)	Adherence (%)	Machine	Training regimen
Bruce-Low et al. 37	12	1–2	12–24	ND	Lumbar extension machine	One group trained once a week, the other twice. Each session included 8–12 reps at approximately 80% of maximum functional torque, lasting 70–105 s.
Fortin et al. 38	12	2	24	92.5	MedX machine	Two sets of lumbar extensions: 15–20 reps at 55% of 1 RM at a 24° angle. Increase load by 5% after completing 15–20 reps.
Harts et al. 41	8	1–2	10	ND	Modified lower back machine: secures the pelvis and hips to isolate the lower back	High intensity group: 15 to 20 repetitions at 50% of maximal isometric lumbar extension. Low intensity group: 15 to 20 repetitions at 20% of maximal isometric lumbar extension.
Helmhout et al. 36	10	2	14	ND	Total Trunk Rehab machine	15 to 20 reps per session on the lower back machine at approximately 50–70% of 1-RM.
Risch et al. 39	10	1–2	14	ND	MedX machine	12 repetitions per session; 2 sessions per week for the first 4 weeks, then reduced to 1 session per week.
Smith et al. 13	12	1	24	ND	MedX machine	One set of 8–12 repetitions through the participant’s full ROM on the lumbar extension machine to volitional fatigue, performed slowly with a 2-second lift and 4-second lower per repetition.
Steele et al. 16	12	1	24	ND	MedX machine	The Full range of motion group used their full range of motion, while the Lim range of motiongroup used the mid 50%. Both groups performed one set at 80% of maximal torque to volitional failure, with a 2-second concentric phase, 1-second hold, and 4-second eccentric phase.
Steele et al. 40	12	1	24		MedX machine	A single set to momentary muscular failure using a load at 80% of max tested torque.

ND not described, 1-RM one repetition maximum.

Quality of the studies

The assessment of study quality for the included studies (Table 4) showed that three studies received 5 points on the PEDro scale, while four others scored between 6 and 8 points. Additionally, one study was classified with scores ranging from 9 to 10 points. The most common sources of bias were the lack of blinding for all subjects (n = 8) and the absence of blinding for the therapists administering the therapy (n = 7).

Table 4.

Risk of bias assessment using the Physiotherapy evidence database scale (PEDro).

Study	C1	C2	C3	C4	C5	C6	C7	C8	C9	C10	C11	Score, quality
Bruce-Low et al. 37	1	1	0	1	0	0	0	1	1	1	1	6, good
Fortin et al. 38	1	1	1	1	0	0	1	1	1	1	1	8, good
Harts et al. 41	1	1	1	1	0	0	1	0	1	1	1	7, good
Helmhout et al. 36	1	1	1	1	0	1	1	1	1	1	1	9, excellent
Risch et al. 39	1	1	0	1	0	0	0	1	1	1	1	6, good
Smith et al. 13	1	1	0	0	0	0	0	1	1	1	1	5, fair
Steele et al. 16	1	1	0	0	0	0	0	1	1	1	1	5, fair
Steele et al. 40	1	1	0	0	0	0	0	1	1	1	1	5, fair

C1: eligibility criteria were specified; C2: subjects were randomly allocated to groups; C3: allocation was concealed; C4: the groups were similar at baseline regarding the most important prognostic indicators; C5: there was blinding of all subjects; C6: there was blinding of all therapists who administered the therapy; C7: there was blinding of all assessors who measured at least one key outcome; C8: measures of at least one key outcome were obtained from more than 85% of the subjects initially allocated to groups; C9: all subjects for whom outcome measures were available received the treatment or control condition as allocated, or, where this was not the case, data for at least one key outcome were analyzed according to “intention to treat”; C10: the results of between-group statistical comparisons are reported for at least one key outcome; C11: the study provides both point measures and measures of variability for at least one key outcome.

The risk of bias assessment

The risk of bias in the included studies is summarized in Fig. 2. In the analysis of strength and power performance studies, 6 out of 8 exhibited an overall high risk of bias. Similarly, studies assessing muscle hypertrophy presented a high risk of bias in 5 out of 8 cases. Among studies focusing on endurance performance, 4 out of 6 had an overall high risk of bias. Recovery intervention studies showed mixed results, with 4 out of 5 displaying a high risk of bias. The high risk of bias was mainly concerning dimensions 1 (randomization and allocation concealment), 4 (blinding of assessors), and 5 (selective reporting of results). The main issue was the lack of sufficient information on the randomization procedures, leading to concerns over the potential for bias in group allocation. Studies assessing muscle strength and power demonstrated significant risks in dimensions 1 and 4, with inadequate randomization and blinding^13,41. Furthermore, studies examining muscle hypertrophy raised concerns regarding the selective reporting of results (dimension 5)^16,37. Considering endurance performance, the study by Fortin et al.³⁸ had lower concerns in randomization and blinding and showed issues with missing data (dimension 3). Regarding recovery interventions, Steele et al.⁴⁰ showed a high risk of randomization and assessor blinding.

Fig. 2.

Assessment of risk of bias for the randomized trials (RoB2).

Overall, the majority of the studies included in this review had significant methodological limitations, particularly in randomization, blinding, and selective reporting, which can influence the outcomes related to strength, hypertrophy, endurance, and recovery. These limitations suggest a need for caution when interpreting the findings and highlight the importance of improving methodological rigor in future research to enhance the reliability and validity of the results.

Summary of the individual studies

Table 5 summarizes the main results obtained from the individual studies regarding pain-related outcomes, disability-related outcomes, and physical functionality. Among the studies focusing on pain-related outcomes, a favorable tendency for improvement in ILEX was observed, ranging from − 14.7³⁹ to −64.8%¹⁶. In contrast, the control group exhibited a range of changes from enhancements of − 31.5 ³⁸ to an increase in pain up to 34.9%⁴⁰. In the studies focused on disability, the ILEX results showed improvements ranging from − 26.3%⁴¹ to − 59.0%³⁶ in scores, while the control groups exhibited enhancements of − 55.7%³⁶ to declines in condition of approximately 3.4%¹³. Regarding improvements in physical functionality, specifically maximal voluntary isometric torque, the enhancements in ILEX ranged from strength gains of 5.6%⁴¹ to 78.2%³⁷. In contrast, the control groups exhibited improvements of 16.5%³⁶ to declines of up to − 2.4%³⁹.

Table 5.

Results of the individual studies on pain-related outcomes, disability-related outcomes and physical functionality.

Study	Outcome	N	ILEX pre (mean)	ILEX pre (SD)	ILEX post (mean)	ILEX post (SD)	% difference post-pre	N	Control pre (mean)	Control pre (SD)	Control post (mean)	Control post (SD)	% difference post-pre
Pain
Fortin et al. 38	NPRS	25	5.23	0.34	2.80	0.38	−46.5	25	5.20	0.42	3.56	0.46	−31.5
Risch et al. 39#	WHYMPI	31	3.4	1.6	2.9	1.7	−14.7	23	3.7	1.6	4.1	1.5	10.8
Smith et al. 13	VAS	15	30.10	17.20	13.40	10.80	−55.5	13	26.80	9.00	26.50	10.20	−1.1
Steele et al. 16@	VAS	10	46.73	25.53	16.43	36.1	−64.8	7	19.2	15.51	25.91	21.5	34.9
Steele et al. 40	VAS	17	47.26	24.09	23.61	32.34	−50.0	7	19.2	15.51	25.91	21.5	34.9
Disability
Fortin et al. 38	ODI	25	29.54	2.05	19.08	1.95	−35.4	25	27.52	2.19	18.19	2.08	−33.9
Harts et al. 41$	RDQ	21	7.6	4.6	5.6	7.3	−26.3	21	6.5	3.9	4.7	5.7	−27.7
Helmhout et al. 36	RDQ	61	8.3	4.8	3.4	4.6	−59.0	46	7.9	4.4	3.5	4.2	−55.7
Smith et al. 13	ODI	15	39.2	14.7	27.30	11.60	−30.4	13	32.70	5.90	33.80	6.30	3.4
Steele et al. 16@	ODI	10	36.18	11.12	17.98	12.93	−50.3	7	26.2	7.27	23.20	9.99	−11.5
Steele et al. 40	ODI	17	34.71	12.69	17.65	14.35	−49.2	7	26.2	7.27	24.49	10.78	−6.5
Physical functionality
Bruce-Low et al. 37&	MVIT	31	121.0	13.2	215.6	15.3	78.2	21	151.6	18.3	165.0	18.2	8.8
Harts et al. 41	MVIT	21	215	52	227	69.5	5.6	21	213	64	209	72	−1.9
Helmhout et al. 36	MVIT	61	214	64	244	66	14.0	46	212	65	247	73	16.5
Risch et al. 39€	MVIT	31	70.9	43.9	100.7	63.4	42.0	23	74.0	54.5	72.2	58.7	−2.4
Steele et al. 40	MVIT	17	177.80	83.80	219.29	89.1	23.3	7	192.21	67.60	202.50	69.96	5.4

ILEX: isolated lumbar extension; VAS: Visual analogue scale; NPRS: Numerical Pain Rating Scale; WHYMPI: West-Haven Yale Multidimensional Pain Inventory; ODI: Oswestry Disability Index; RDQ: Roland-Morris Disability Questionnaire; MVIT: Maximal voluntary isometric torque in lumbar extension.

^#Data was collected from the pain subscale.

^@Data was collected from full-range group.

^$Data was collected from the low-intensity training program.

^&Data was collected from the group that trained once a week, with scores obtained at an angle of 0º.

^€Data was collected from scores at an angle of 0º.

Meta-analysis

A meta-analysis was conducted to compare the ILEX group with true control groups to standardize the study conditions. True controls were defined as those participants who were not exposed to any intervention or treatment during the study period. There were not enough studies with active control groups (parallel groups exposed to an alternative physical exercise intervention or non-exercise rehabilitation methods) to conduct the meta-analysis.

Results (Fig. 3) showed significant favoring effect for the ILEX compared to the true-control group in pain-related outcomes (ES = − 0.633, 95% CI − 1.06 to − 0.20, p = 0.004, I² = 19%, total participants n = 123, Egger test two-tailed = 0.601).

Fig. 3.

Forest plot illustrating changes in pain-related outcomes after isolated lumbar extension (ILEX) in comparison to true-control groups. Forest plot values are shown as effect sizes (ES [Hedges’ g]) with 95% confidence intervals (CI). Blue squares: individual studies. Green rhomboid: overall summary value.

Results (Fig. 4) showed non-significant favoring effect for the ILEX compared to the true-control group in disability-related outcomes (ES = − 0.292, 95% CI − 0.73 to 0.14, p = 0.190, I² = 20%, total participants n = 111, Egger test two-tailed = 0.324).

Fig. 4.

Forest plot illustrating changes in disability-related outcomes after isolated lumbar extension (ILEX) in comparison to true-control groups. Forest plot values are shown as effect sizes (ES [Hedges’ g]) with 95% confidence intervals (CI). Blue squares: individual studies. Green rhomboid: overall summary value.

Results (Fig. 5) showed non-significant favoring effect for the ILEX compared to the true-control group in isometric strength outcomes (ES = 0.967, 95% CI − 0.35 to 2.28, p = 0.150, I² = 93%, total participants n = 172, Egger test two-tailed = 0.617).

Fig. 5.

Forest plot illustrating changes in maximal voluntary isometric strength outcomes after isolated lumbar extension (ILEX) in comparison to true-control groups. Forest plot values are shown as effect sizes (ES [Hedges’ g]) with 95% confidence intervals (CI). Blue squares: individual studies. Green rhomboid: overall summary value.

Certainty of evidence

Table 6 presents the certainty assessment based on the GRADE analysis. The certainty of evidence regarding pain, disability, and physical functionality outcomes has been deemed very low. This is primarily due to the high risk of bias observed in most of the studies included. Furthermore, the small number of participants contributes to the imprecision in the reported effects on physical performance, which further weakens the evidence. The small sample sizes, along with the lack of a consistent direction of effects in the comparisons between the ILEX and control groups, collectively lead to the very low certainty of the evidence.

Table 6.

GRADE analysis.

Outcomes (ILEX vs. control)	Studies and PSS	Risk of bias in studies	Risk of publication bias	Inconsistency	Imprecision	Certainty of evidence
Pain-related outcomes	4, n = 123	Downgrade by two levels (high-risk of bias)	Not applicable	Downgrade by one level (I2 = 26%)	Downgrade by two levels: (i) < 800 participants; (ii) no clear direction of effect.	⊕, Very low
Disability-related outcomes	4, n = 111	Downgrade by two levels (high-risk of bias)	Not applicable	No downgrading (I2 = 2%)	Downgrade by two level: (i) < 800 participants; (ii) no clear direction of effect.	⊕, Very low
Maximal voluntary isometric strength	4, n = 172	Downgrade by two levels (high-risk of bias)	Not applicable	Downgrade by one level (I2 = 92%)	Downgrade by two levels: (i) < 800 participants; (ii) no clear direction of effect.	⊕, Very low

(i) Risk of bias in studies: downgraded by one level if some concerns and two levels if high-risk of bias; (ii) Indirectness: considered low due to eligibility criteria; (iii) Risk of publication bias: not assessed, as all comparison had < 10 studies available; downgrade one level if Egger’s test < 0.05; (iv) Inconsistency: downgraded by one level when the impact of statistical heterogeneity (I²) was moderate (> 25%) and by two levels when high (> 75%); (v) Imprecision: downgraded by one level when < 800 participants were available for a comparison or if there was no clear direction of the effects⁴⁶; accumulation of both resulted in downgrading by two levels.

GRADE: Grading of Recommendations Assessment, Development and Evaluation; ILEX: isolated lumbar extension; PSS: pooled sample size.

Discussion

The current systematic review and meta-analysis aimed to evaluate the effects of ILEX on pain, disability, and physical functionality in individuals with non-specific LBP. Despite the inclusion of various study designs, including those comparing ILEX with active controls and true controls (those receiving no treatment), the meta-analysis revealed that ILEX significantly reduced pain intensity compared to true controls. However, this improvement was not observed for disability or isometric strength. Nevertheless, although there was no statistically significant effect on isometric strength of the back extensors, the magnitude of the difference moderately favored ILEX.

Effects of ILEX on pain-related outcomes

The meta-analysis results revealed that ILEX significantly improved pain intensity, typically measured using the visual analogue scale and numerical pain rating scale, compared to true controls. These interventions often involved a limited number of training sessions per week (usually just one)^13,16,40, with minimal training consisting of a single set of 8 to 12 repetitions¹³ performed to failure. The significant reduction in pain intensity observed in the meta-analysis for ILEX exercises can be attributed to the improvement in spinal stability and posture. Strengthening these muscles through controlled, high-effort contractions likely increased local muscular endurance and hypertrophy. This aligns with Fortin’s et al.³⁸, which reported significant increases in the cross-sectional area of the erector spinae and multifidus muscles after ILEX exercises. These improvements may potentially reduce mechanical strain on spinal structures, such as intervertebral discs and facet joints³⁸. Additionally, the minimal training frequency (one session per week) is sufficient to stimulate neuromuscular adaptations, such as increased motor unit recruitment and improved coordination, without overloading the muscles or exacerbating pain symptoms⁴⁰. This type of high-intensity, low-frequency regimen also promotes favorable biochemical changes, including a reduction in pro-inflammatory cytokines and an increase in endorphin production, which can directly influence pain perception⁴².

On the other hand, when comparing ILEX to active control groups, it was not possible to perform a meta-analysis due to the limited number of studies. However, the available evidence suggests that the effect of ILEX may not be significantly different from other forms of exercise. For instance, Fortin et al.³⁸ found no significant differences in pain intensity between the ILEX group and a general exercise group (performing multiple resistance training exercises) after treatment. In contrast, Smith et al.¹³ reported that ILEX was significantly more effective, with participants experiencing lower pain intensities compared to a training group that did not include pelvic stabilization. The reduction in pain intensity following treatment with ILEX, compared to treatment without pelvic stabilization, could be attributed to the fact that without proper stabilization of the pelvis, the joint movement is not effectively isolated¹³. In such cases, the resistance applied during exercise may not primarily target the lumbar muscles, as intended. Instead, larger muscle groups, such as the gluteal muscles and hamstrings, which assist in moving the pelvis during extension, may generate most of the force. This reduces the workload on the lumbar muscles, potentially leading to less effective strengthening and pain relief in the lumbar region¹³. By stabilizing the pelvis, ILEX ensures that the lumbar muscles are more directly engaged, leading to better outcomes in pain reduction¹³.

Effects of ILEX on disability-related outcomes

The meta-analysis comparing the effects of ILEX to true controls on disability-related outcomes indicated that there was no statistically significant difference between the two groups. Furthermore, the observed magnitude of differences was small, suggesting that ILEX may have a limited impact on these outcomes. However, the study by Bruce-Low et al.³⁷ revealed significant benefits of ILEX in improving scores on the Oswestry Disability Index, demonstrating that both one and two sessions per week over twelve weeks were effective in enhancing disability outcomes. Similarly, the research conducted by Smith et al.¹³ also found improvements in disability levels compared to the control group. In contrast, the study by Harts et al.⁴¹ did not reveal any significant enhancements compared to the control group. However, the study⁴¹ noted a limitation that may affect the generalizability of the results: the mean baseline Roland-Morris Disability Questionnaire score was relatively low. This low initial disability level may restrict the applicability of the findings to populations with more severe disabilities⁴¹.

Theoretically, strengthening the lumbar extensors would enhance spinal stability and posture⁴³. Increasing muscle strength in this region would provide better support for the vertebral column, alleviating strain on the intervertebral discs and surrounding soft tissues⁴⁴. This, in turn, may reduce pain, as observed in our meta-analysis and improve overall physical function³⁷. Additionally, the psychological benefits of increased strength and reduced pain can enhance confidence in one’s physical abilities. This is illustrated in the study by Risch et al.³⁹, which found that ILEX helped patients improve their perceptions of physical and psychosocial functioning. These factors may help explain the findings of individual studies^13,16,37 supporting ILEX.

On the other hand, when compared with active controls, the study by Fortin et al.³⁸ did not find significant benefits of ILEX compared to general exercise based on resistance training in terms of its impact on the Oswestry Disability Index. Similarly, the study by Helmhout et al.³⁶ also reported comparable disability levels when comparing ILEX to regular physical therapy. The core elements of both ILEX and general resistance training exercises—such as increased muscular strength and improved range of motion—are likely essential for functional recovery and the reduction of disability, regardless of the specific exercise modality used¹¹.

Thus, the physiological benefits of muscle strengthening and improved range of motion, which are key to reducing disability, are not unique to ILEX but also occur with other forms of exercise. This suggests that the core mechanisms driving improvement—muscle strength, range of motion, and neurophysiological adaptations—are shared across different interventions. Thus, while ILEX may provide some benefits, its impact may be comparable to other exercise modalities, and its specific effects on disability may be less pronounced when compared to diverse treatment protocols.

Effects of ILEX on physical functionality related outcomes

Among the included studies, physical functionality was primarily assessed through maximal isometric strength in the lumbar extensors. A couple of studies^16,37 also reported on range of motion, but analysis was not possible due to the limited number of studies addressing this outcome. The meta-analysis revealed non-significant differences between the ILEX and true-control groups; however, the magnitude of the observed difference was moderate.

One possible explanation for the moderate effect size observed in the meta-analysis may be the results of Bruce-Low et al.³⁷ which clearly provided evidence of benefits in enhancing muscle strength, with the one single session a week presenting similar enhancements of two sessions a week. Several individual studies, such as the one by Steele et al., revealed significant enhancements in maximal strength compared to the true-control group⁴⁰. On the other hand, the study of Harts et al.⁴¹ did not reveal significant benefits of either high or low intense ILEX in comparison to control group in improvement of strength torque. Risch et al.³⁹ also did not find significant differences between the ILEX and true-control groups regarding mean torque at angles of 0 and 72 degrees. However, significant benefits of ILEX were reported at angles between 12 and 60 degrees.

In comparison to active controls, such as those in the study by Helmhout et al.³⁶ which examined physical therapy training, the strength gains promoted by ILEX were similar to those of the active control group. However, in the study by Smith et al.¹³, which compared ILEX to lumbar extension exercises without pelvic stabilization, ILEX demonstrated a significant advantage at angles ranging from 0 to 27 degrees. The lack of significant improvements or the contradictory findings in maximal strength following ILEX may stem from various factors related to the specificity of strength adaptations and the distinctions between isometric and dynamic contractions. Isometric strength gains are typically attributed to neural adaptations, which can be influenced by the training modality and intensity⁴⁵. ILEX primarily targets the lumbar extensor muscles through a limited range of motion, potentially failing to recruit the necessary motor units or activate synergistic muscles required for significant isometric strength improvements across a broader range of joint angles. Furthermore, most exercises were performed at sub-maximal intensities, typically ranging from 50 to 70% of one-repetition maximum, often consisting of a single set of repetitions, and frequently conducted only once a week^13,16,40. This training frequency and intensity may not provide the necessary stimulus to elicit substantial improvements in maximal isometric strength.

Study limitations, future research and practical applications

Among the studies included in our review, the primary limitations relate to the lack of blinding for participants, evaluators, and therapists. This is a common issue in physical therapy research that needs to be addressed in future studies. Additionally, most of the research focused on middle-aged populations with low back pain, with few studies examining younger or older individuals, or specific groups such as those who are overweight, who could benefit from ILEX.

Most studies relied on subjective measures of pain-related outcomes rather than employing objective measurements. Furthermore, there was a limitation in the physical tests analyzed, as many did not assess strength under dynamic conditions. Moreover, most studies implemented minimal interventions with low load and volume, making it difficult to determine whether a greater stimulus would lead to more significant positive adaptations. Future research should also focus on analyzing the individualization of training, specifically by adjusting the dosage based on shorter evaluation periods. Additionally, it should examine the impact of baseline characteristics on the heterogeneity of responses.

As practical applications, given the significant reduction in pain intensity associated with ILEX, practitioners can consider incorporating these exercises into rehabilitation programs to enhance patient outcomes, particularly for those with persistent pain. The evidence suggests that even low-frequency, high-intensity training can lead to meaningful improvements in spinal stability and muscular endurance, which are crucial for managing LBP. Clinicians should also recognize the importance of pelvic stabilization during exercises to ensure proper engagement of the lumbar muscles, potentially enhancing the effectiveness of treatment. However, it is important to recognize that the impacts of ILEX on disability and functionality may be limited compared to other forms of resistance training or active physical therapy. Clinicians should keep this in mind during the treatment process.

Conclusions

ILEX was found to significantly reduce pain intensity compared to true control groups, highlighting its potential as an effective intervention for alleviating pain. However, the evidence regarding ILEX’s impact on disability and physical functionality outcomes is less robust. While some studies indicated improvements in disability scores, the overall meta-analysis suggested no significant differences compared to control groups, implying a limited effect of ILEX on these outcomes. Furthermore, while ILEX showed moderate effects on maximal strength, this was not unanimously supported across all studies. Clinically, the significant pain reduction associated with ILEX underscores its value in rehabilitation programs, particularly when combined with a focus on pelvic stabilization to enhance lumbar muscle engagement. Overall, while ILEX shows promise in managing pain in individuals with LBP, clinicians should approach its application with an awareness of its limited effects on disability and functional strength compared to other exercise modalities.

Electronic supplementary material

Below is the link to the electronic supplementary material.

Author contributions

RT and FMC conceived the study idea, designed the protocol, conducted searches and screening, handled data analysis, contributed to writing, and approved the document. RT and FMC conducted searches, performed screening, extracted data, authored the article, and provided approval. WM, SM, BB, MBP, and IR contributed to risk of bias assessment, writing, and approval of the manuscript. MG: conducted the RoB2, writing, and approval of the manuscript.

Data availability

All data is available by per request to the corresponding author.

Declarations

Competing interests

The authors declare no competing interests.