Comparative efficacy of non-pharmacological interventions on kinesiophobia in patients with chronic non-specific low back pain: a study protocol for a systematic review and network meta-analysis

Abstract

Introduction

Kinesiophobia is a significant factor which influences the prognosis of patients with chronic non-specific low back pain (CNLBP). The restriction of functional training can severely impede functional recovery and contribute to negative emotional states (including anxiety and depression) and a heightened risk of insomnia while exacerbating the economic burden on patients. Although several randomised controlled trials have evaluated the effects of various non-pharmacological interventions on kinesiophobia in CNLBP, their relative efficacy and potential adverse effects remain ambiguous. This study will conduct a systematic review and network meta-analysis to identify which non-pharmacological intervention may represent the most effective treatment for kinesiophobia in patients with CNLBP.

Methods and analysis

Comprehensive searches will be conducted across English and Chinese databases (including The Cochrane Library, PubMed, Scopus, Springer, Embase, Wanfang Data, the Chongqing VIP Database and China National Knowledge Infrastructure) from the date of their inception to 20 November 2024. Only Chinese or English studies will be considered for analysis. The primary outcomes will include a significant reduction in secondary movement phobia associated with CNLBP, pain alleviation and enhancement of the functional status of the lumbar muscles. The Cochrane Bias Risk Assessment Tool will be used to conduct a bias risk assessment.

Pairwise meta-analysis will be performed by Review Manager V.5.3 software, Stata V.16.0 and Open BUGS V.3.2.3 software will be used to conduct a network meta-analysis. The Grading of Recommendations, Assessment, Development and Evaluation framework will be employed to assess the quality of evidence.

Ethics and dissemination

All of the data included in this study will be derived from the literature; therefore, ethical approval is not necessary. The findings will be disseminated via peer-reviewed journals and academic conferences.

PROSPERO registration number

CRD42024605343.

Strengths and limitations of this study.

• This study will exclusively include randomised controlled trials.

• This systematic review protocol will be developed per the reporting guidelines established by the Preferred Reporting Items for Systematic reviews and Meta-Analysis Protocols checklist.

• The outcome indicators of the study will be amalgamated with both subjective assessment scales and objective physiological indices.

• Two reviewers will independently conduct the study selection, data extraction and quality assessment.

• The findings of this study may be constrained by a range of factors, including publication bias, language bias, study heterogeneity, the assessment tools employed and methodological quality.

Introduction

Chronic non-specific low back pain (CNLBP) is defined as low back pain from non-specific spinal or non-spinal causes and occurs between the lower rib margin and the hip fold.¹ It accounts for ~90% of low back pain cases and is a major cause of disability worldwide.² Kinesiophobia—an excessive and irrational fear of movement due to concerns about pain or re-injury—significantly impacts rehabilitation outcomes by causing patients to avoid physical activities.³ Such avoidance behaviour can lead to functional impairment, reduced quality of life and long-term disability.⁴ For example, a high incidence of kinesiophobia has been observed postsurgery, where it challenges the rehabilitation process by limiting a patient’s participation in exercise and exacerbating pain.⁵ Additionally, kinesiophobia is associated with increased anxiety and depression, which can further hinder recovery. Therefore, addressing kinesiophobia via targeted interventions is crucial for improving rehabilitation outcomes and overall patient well-being.⁶

The occurrence of CNLBP is closely related to psychosocial factors, somatosensory function and the ability of the central nervous system to regulate pain.⁷ The catastrophising of pain not only directly affects motor function but may also indirectly affect motor function through the mediating factors of exercise fear or self-efficacy, as well as the chain mediation between exercise fear and self-efficacy.⁸ Therefore, in the chain between pain and motor function, kinetophobia plays a key role by seriously hindering the recovery process of patients.⁹ Additionally, the amygdala, which is a key brain region for fear and threat, plays a pivotal role in phobophobia.¹⁰ The process of this fear is similar to the mechanism of conditioned fear, in which a ‘movement-pain-fear’ chain is formed.¹¹ One study shows that by inhibiting the activation of gamma-aminobutyric acid (GABA) neurons in the central nucleus of the amygdala in a neuropathic pain mouse model, the mouse’s perception of pain can be suppressed, thereby improving the symptoms of phobophobia.¹² Therefore, the occurrence of CNLBP is intertwined with complex neurophysiological mechanisms, encompassing psychosocial factors, somatosensory functions and intricate elements of the central nervous system.¹³ The interplay between these factors has a collective impact on a patient’s pain perception and motor function, which culminates in a complex chain reaction.

Currently, there is a lack of specific drugs for CNLBP, and the treatment relies primarily on non-drug therapies.¹⁴ Many patients use opioid drugs to achieve pain relief; however, such an approach can cause extreme addiction and seriously harm their health.¹⁵ Although there are medications to treat phobias, they do not provide pain relief and can have side effects such as digestive discomfort.¹⁶ Given the considerable adverse effects associated with certain therapeutic agents, there is an urgent necessity for effective non-pharmacological interventions to address kinetophobia in patients with CNLBP. As a first line of treatment, Cognitive Behavioural Therapy (CBT) is highly effective in improving the psychological state of patients; however, patient compliance is low and the direct relief of physical pain is limited.¹⁷ Although conventional exercise therapy can intervene in physical symptoms, it has proven to be limited in improving the psychological state of patients.¹⁸

Several systematic reviews and meta-analyses have summarised this area; however, they have all demonstrated one or more of the following limitations: (1) The included studies are not comprehensive, and some of them do not mention non-drug safety.19,21 (2) The evaluation tool is relatively simple, and the patient’s symptoms cannot be evaluated multidimensionally.22,24 (3) Some of the evidence-based evidence was published some time ago and requires updating25,27 and (4) They contain a lack of evidence concerning direct or indirect comparisons between non-drug modalities.28,30 Given this, this study collected all of the available literature and employed a network meta-analysis methodology to conduct a comprehensive evaluation and ranking of different non-drug ways to improve kinesiophobia in CNLBP. This approach was selected to provide evidence-based evidence for the best non-drug intervention for clinical rehabilitation practice.

Methods

Study registration and design

This protocol will be conducted per the Preferred Reporting Items for Systematic Reviews and Meta-Analyses Protocol (PRISMA-P).³¹ The study has been registered with PROSPERO (Registration No. CRD42024605343). It is scheduled to commence on 15 April 2025, and is anticipated to conclude by 20 September 2025.

Inclusion criteria of studies

Types of studies

The study type will focus on randomised controlled trials (RCTs) containing an examination of the intervention effects of various non-pharmacological therapies for patients with kinesiophobia in CNLBP.

Types of participants

All participants will meet the following criteria:

1. Adhere to the Clinical guidelines for Chronic non-specific low back pain in China and The North American Spine Society’s Evidence Based Clinical Guideline for the Diagnosis and Treatment of Low Back Pain, along with other pertinent diagnostic criteria for CNLBP.^{32 33}

2. The Tampa Scale of Kinesiophobia score (TSK) is above 37 points.

3. Recurrent low back pain with a duration of more than 12 weeks.

4. The age of the participants must be between 18 and 65 years old (with no limitations on gender).

The aforementioned criteria must be met simultaneously during the visit to enable a phobia diagnosis.

Types of interventions

Following a comprehensive consideration of the findings from the literature review, non-pharmacological interventions will encompass (1) CBT. (2) Rehabilitation Training under Virtual Reality Technology. (3) Exercise therapy (such as muscle strength training, movement control, breathing exercises, stability training, Tai Chi, Qigong, yoga, Pilates and therapeutic aquatic exercise). (4) Physical therapy (such as transcranial direct current stimulation, extracorporeal shock wave therapy, magnetic hyperthermia and electromyographic biofeedback). (5) Acupuncture treatment. (6) Kinesio taping. (7) Manual therapies. (8) Mindfulness-Based Cognitive Therapy and (9) Mindfulness-Based Stress Reduction. The experimental group will receive one or more of these non-pharmacological intervention methods.

Types of control groups

The control group will receive standard treatment or non-pharmacological interventions that differ from those provided to the intervention group.³⁴

Types of outcome measures

Primary outcome measures

Based on the preliminary search of relevant literature, the primary outcome measures will encompass:

1. Fear of movement indicators—as determined by the TSK.

   TSK is a validated 17-item self-report questionnaire that is employed to assess the fear of movement/re-injury in chronic pain populations (higher scores signify a greater level of kinesiophobia). It has demonstrated strong internal consistency and construct validity across diverse patient groups.³⁵

2. Pain indicators—as determined by the Visual Analogue Scale (VAS) and the Numeric Rating Scale (NRS).

   VAS and NRS are simple tools for quantifying pain intensity. The VAS uses a continuous line anchored by ‘no pain’ and ‘worst possible pain’, while the NRS uses a 0–10 numerical scale. Both are widely used in clinical and research settings to track pain changes over time and they have demonstrated strong reliability and validity.^{36 37}

3. Functional indicators—as determined by the Roland-Morris Disability Questionnaire (RDQ) and the Oswestry Disability Index (ODI).

   RDQ and ODI are self-report questionnaires that are used to measure functional disability caused by low back pain. The RDQ (24 items, scored 0–24) focuses on daily activities, while the ODI (10 sections, scored 0–100) assesses broader disability. Higher scores indicate greater disability. Both tools are widely used, with strong reliability and validity.³⁸

4. Muscle strength and muscle fatigue indicators: a surface electromyogram (sEMG) instrument will be employed to determine the mean power frequency (MPF) and integrated electromyogram (IEMG) of the lumbar extensor muscle group.

   The sEMG is a non-invasive method that is used for assessing muscle activity by recording electrical signals from the skin surface. Electrodes are placed over target muscles (eg, lumbar extensors) to measure specific parameters (such as MPF and IEMG) to provide objective data concerning muscle strength and fatigue. sEMG is widely used in clinical and research settings due to its reliability and validity.³⁹

Secondary outcome measures

Based on the preliminary retrieval of relevant literature, the secondary outcome measures will comprise:

1. Psychological state—as determined by the Self-Rating Anxiety Scale (SAS) and the Self-Rating Depression Scale (SDS) for assessing the psychological state of patients.

   SAS and SDS are self-report questionnaires that assess anxiety and depression, respectively. Both use a 4-point Likert scale, with higher scores indicating greater severity. They are widely used for their ease of administration and good psychometric properties, including internal consistency and construct validity.⁴⁰

2. Sleep quality—as determined by Polysomnography (PSG) and the Pittsburgh Sleep Quality Index (PSQI).

   PSG objectively assesses sleep by recording physiological parameters (such as brain waves and breathing). PSQI subjectively measures sleep quality over a month via a 19-item questionnaire (scored 0–3). PSG is the gold standard for sleep studies, and PSQI has demonstrated strong reliability and validity.^{41 42}

3. The incidence of adverse events.

Exclusion criteria of studies

1. Literature that is not published in either Chinese or English.

2. Duplicate publications.

3. Review articles, case reports, conference abstracts and other similar publications.

4. Literature for which full-text access is not available.

5. Literature which lacks sufficient data or contains incomplete data.

6. Literature rated as high risk of bias by the Cochrane Risk of Bias tool or classified as low or very low quality of evidence according to the GRADE criteria will be excluded.

Search methods for identification of studies

Electronic searches

Comprehensive literature searches will be conducted in both English and Chinese databases, including The Cochrane Library, PubMed, Scopus, Springer, Embase, Wanfang Data, the Chongqing VIP Database (VIP) and the China National Knowledge Infrastructure (CNKI). Additionally, to enhance the rigour and comprehensiveness of the search, additional relevant articles will be manually retrieved from preprint servers such as bioRxiv and medRxiv.

A combination of subject terms and free-text keywords will be used to identify relevant studies, trace published systematic reviews and review their references. The search will cover the period from the inception of each database to 20 November 2024. Details of the search strategy for PubMed, Embase and Cochrane Library will be provided in online supplemental material 1.

Data collection and analysis

Selection of studies

Per the objectives of the study and the established inclusion and exclusion criteria, two investigators (TL and MT), both with extensive experience in systematic reviews and meta-analyses, will independently screen and cross-verify the literature. If a consensus cannot be reached via discussion, the issue will be referred to a third investigator (QC) for resolution. The complete literature screening workflow is detailed in figure 1. The specific steps will be as follows:

Figure 1. Flowchart of study selection.

1. The literature will be imported into EndNote V.X9 reference management software, and any duplicate and irrelevant entries will be excluded.

2. The title and abstract will be preliminarily reviewed to eliminate literature that is not pertinent to the research topic.

3. The abstract and the full text will be thoroughly reviewed, the literature will be screened according to the established inclusion and exclusion criteria, any relevant information will be extracted, and the justifications for exclusion will be comprehensively documented.

Data extraction

Two investigators (TL and MT), both with extensive experience in systematic reviews and meta-analyses, will independently extract data in a double-blind manner. To maintain blinding during the extraction process, the study information provided to the investigators will be anonymised via the removal of the authors’ identities and publication details. In the event of any missing data, the authors will be contacted via email (where possible) to obtain supplementary information. Up to three attempts will be made to contact the authors (over 4 weeks), and sufficient time will be allowed to receive a response. If, following these attempts, the data cannot be obtained, a sensitivity analysis will be conducted to assess the potential impact of missing data on the overall results. Additionally, imputation methods will be employed (where appropriate and feasible) according to the recommendations contained in the Cochrane Handbook for Systematic Reviews.

The following information will be extracted using an Excel sheet:

1. Basic information: primary author, publication date and country.

2. Key characteristics of the subjects: age, gender distribution, sample size, duration of illness and intervention strategies (methodology, duration and frequency).

3. Noteworthy results: relevant outcome indicators, adverse event reporting, follow-up time, literature bias risk assessments and other similar results.

The data will be independently cross-checked by two investigators and any discrepancies that cannot be resolved via discussion between the two investigators will be resolved by consulting a third investigator (QC).

Assessment for risk of bias in included studies

Two investigators (TL and MT) will evaluate the quality of the included studies according to the Cochrane Handbook³⁸ and in cases of disagreement, a decision will be made in consultation with a third investigator (QC). Two investigators (TL and MT) will independently evaluate the quality of the included studies using the Cochrane Risk of Bias tool (as recommended in the Cochrane Handbook for Systematic Reviews of Interventions).³⁸ In cases of disagreement, a consensus will be reached via discussion and, if necessary, a third investigator (QC) will be consulted to make the final decision. The risk of bias assessment will cover six essential aspects: (1) Random sequence generation. (2) Allocation concealment. (3) Blinding of participants and personnel. (4) Blinding of outcome assessment. (5) Completeness of outcome data. (6) Selective reporting and (7) Other potential biases. Each aspect will be classified as either low risk of bias, high risk of bias or unclear risk of bias.

To further enhance the transparency of the quality assessment process, a grading system will be employed to categorise the overall risk of bias for each study. Studies that fully meet the criteria for low risk of bias across all domains will be classified as Grade A (indicating a low potential for bias). Studies that meet most of the criteria but have some domains with unclear or high risk of bias will be classified as Grade B (indicating a moderate potential for bias). Studies that fail to meet the criteria for low risk of bias in multiple domains will be classified as Grade C (indicating a high potential for bias).

Statistical analysis

Pairwise meta-analysis

Review Manager V.5.3 software will be leveraged to conduct a conventional meta-analysis of the incorporated literature. For count data, the OR will be used as the statistical measure for evaluating efficacy and safety. Measurement data will be reported as Mean Difference (MD) or Standardised Mean Difference (SMD), with 95% CI employed to illustrate effect sizes. If all of the included studies use the same scale (such as the Tampa scale to assess kinesiophobia) and the units/ranges are consistent, MD will be prioritised to preserve clinical significance. For any study that uses different scales (such as the Tampa scale and FABQ scale) or has significant differences in measurement range, SMD (Cohen’s d) will be used to standardise the effect size. Heterogeneity tests will be conducted using the statistic I² in accordance with the Cochrane systematic review manual.⁴³ If p≥0.05 and I²≤50%, a fixed-effect model will be employed for meta-analysis. Conversely, if p<0.05 and I²>50%, a random-effects model will be used for meta-analysis. In the event that data conversion is deemed necessary, any relevant calculations will be conducted using the calculator provided in Review Manager V.5.3.

Network meta-analysis

Stata V.16.0 and Open BUGS V.3.2.3 software will be used to conduct a network meta-analysis and delineate the interventional relationships within the network. The Markov Chain-Monte Carlo random effects model will be employed to conduct the necessary calculations and statistical analyses. Initially, four Markov chains are established for simulation, with a total of 50 000 iterations; the first 20 000 iterations are subjected to annealing to mitigate the influence of initial values. To ensure the robustness of the results, this research will establish four independent Markov chains. Convergence of the chains will be assessed using the Gelman-Rubin diagnostic potential scale reduction factor (PSRF), with convergence considered satisfactory when PSRF values for all parameters are below 1.05. Additionally, graphical assessments of trace plots and posterior density plots will be conducted to confirm adequate mixing and convergence of the chains.

In the presence of a closed loop within the evidence network, the node-split method will be employed to evaluate the concordance between direct and indirect comparisons. Specifically, the treatment effect for each comparison will be split (when both direct and indirect evidence are available) into two separate parameters. This approach facilitates a statistical evaluation of whether the estimates derived from direct and indirect evidence agree. If the p value exceeds 0.05, indicating no significant inconsistency, it will be concluded that the direct and indirect evidence are consistent, and the consistency model will be employed for further analysis. Conversely, if significant inconsistency is detected (p≤0.05), the potential sources of discrepancy (such as differences in study design, population characteristics or intervention implementation) will be investigated. Based on this investigation, some specific studies may be excluded or meta-regression techniques may be employed to adjust for potential confounders. Additionally, this study will conduct a global assessment of inconsistency across the network (if feasible) to provide a comprehensive evaluation of the overall consistency of the network meta-analysis.

When the outcome is anticipated to be a continuous variable, the corresponding MD or SMD, together with the 95% CI, will be used for analysis in accordance with the measurement tool employed. This study will visually represent the likelihood of each non-pharmaceutical intervention being identified as the optimal intervention through the Surface Under the Cumulative Ranking (SUCRA) and subsequently rank these interventions based on their SUCRA values.⁴⁴ Additionally, this research will employ the PSRF to assess the convergence of the model (values close to 1–1.05 indicate a satisfactory convergence). Furthermore, Stata V.16.0 software will be used to draw a funnel plot, and this plot will be used to evaluate publication bias.

Assessment of reporting bias

For any included RCTs containing a minimum of 10 studies, Stata V.16.0 software will be employed to generate funnel plots for the assessment of publication bias.⁴⁵ Asymmetric or incomplete funnel plots may suggest the presence of publication bias. If funnel plot asymmetry is detected using Egger’s test, the trim-and-fill method will be applied to correct for any potential publication bias. Considering that the sensitivity of Egger’s test exceeds that of Begg’s test (when fewer than 20 RCTs are included), this study will use Egger’s linear regression with Stata V.16.0 software to perform a more comprehensive quantitative evaluation of publication bias. When the intercept obtained from Egger’s linear regression approaches 0 and p>0.05, it indicates a negligible risk of publication bias.

Sensitivity analysis

To evaluate the robustness of this systematic review and meta-analysis, a sensitivity analysis will be performed by systematically excluding each study in turn and conducting a combined analysis of the remaining studies.⁴⁶ This approach aims to examine the variations in results following the exclusion of each study, thereby assessing the stability and reliability of the meta-analytic findings.

Heterogeneity management

This study will systematically address heterogeneity to ensure robust and interpretable results. When the I² statistic exceeds 50% (indicating significant heterogeneity among study findings), a random-effects model for meta-analysis will be employed to account for both within-study and between-study variability.⁴⁷ Additionally, subgroup analyses will be conducted to investigate any potential factors that may modify the treatment effect. These factors will include subject characteristics (such as gender, age, stage of CNLBP and region) as well as intervention parameters (including intervention duration, intensity, frequency and follow-up period). Each subgroup analysis will be pre-specified with a clear rationale, and potential confounding factors will be carefully considered to ensure an accurate interpretation of the results. Additionally, in the context of network meta-analysis, node-splitting methods will be employed to assess inconsistency between direct and indirect evidence, which can provide further insights into potential sources of heterogeneity. If significant inconsistency is detected, the underlying causes will be determined, and the analysis will be adjusted accordingly.

Grading the quality of evidence

Two researchers (TL and MT) will employ GRADEprofiler software V.3.6—accessible at www.gradeworkinggroup.org—to evaluate the quality of outcome measures using five downgrading factors: study limitations, inconsistency, indirectness of evidence, imprecision and reporting bias. They will categorise the quality of evidence as high, moderate, low or very low. The findings will be summarised in an evidence summary table, which will include details such as the specific outcomes assessed, the quality of evidence for each outcome according to the GRADE criteria, the direction and magnitude of effects, and any additional relevant information (such as subgroup analyses or the identification of any potential biases). Any disagreements arising from the aforementioned process will be addressed via discussion between the investigators, or by consulting a third investigator (QC).

Patient and public involvement

None

Ethics and dissemination

Ethics committee approval is not required for this review, as the data presented herein will be extracted from existing literature. The intention is to submit the study findings for peer review and select an appropriate academic publishing platform for dissemination, thereby ensuring the broad accessibility of its results.

Discussion

This protocol will rigorously adhere to PRISMA-P guidelines by employing double-blind screening and data extraction to minimise bias. Additionally, heterogeneity analysis, publication bias evaluation (via funnel plots and Egger’s test), and sensitivity analyses will ensure methodological robustness. Subgroup analyses will specifically explore kinesiophobia-related outcomes in CNLBP populations by assessing the modulation effects of demographic factors (gender, age, CNLBP stage and region) and intervention parameters (duration, intensity, frequency and follow-up). These stratified explorations will elucidate whether fear-avoidance behaviour changes are contingent on individual characteristics, while Chinese database inclusions (CNKI/Wanfang) will reduce regional bias.

While this network meta-analysis provides critical insights concerning non-pharmacological interventions for kinesiophobia in CNLBP, several constraints warrant consideration. First, substantial heterogeneity in intervention protocols (eg, exercise dosage, session frequency and treatment duration) may compromise the comparability of pooled effect estimates. Second, the absence of double-blinding in the majority of included studies, coupled with variations in cross-cultural adaptations of kinesiophobia assessment tools (such as the TSK), may contribute to measurement bias. Additionally, the inclusion of only Chinese and English–language studies introduces a risk of language bias, potentially limiting the generalisability of findings across diverse populations. Future studies should strengthen the standardisation of intervention protocols and the consensus–building of core outcome indicators to address these challenges.

This network meta-analysis aims to methodically review the latest and most comprehensive literature in the field and evaluate the effectiveness of several non-pharmacological intervention strategies for CNLBP patients with fear avoidance. Its primary objective is to determine the optimal intervention model and provide a robust scientific basis for clinical diagnosis and treatment while helping patients and therapists make informed choices concerning the most effective approach to treatment. Furthermore, these research results will contribute to the development of relevant treatment guidelines.