A comprehensive search of PubMed, MEDLINE, Embase, CINAHL, Web of Science, PsycINFO, ZETOC, PROSPERO and the first 10 pages of Google Scholar was conducted from inception to 31 January 2022 and updated on 22 June 2023. A search strategy (Supplementary material 1) was developed using the following key words: physical activity, exercise, kinesiophobia, fear of movement, spinal pain, neck pain, low back pain and randomized controlled trial and adapted for each database. Additional filters/limits were added when searching for articles on the databases where they were available. The search strategy was developed by RS and FJ and iteration discussed with DF. The review protocol was registered in PROSPERO (International Prospective Register of Systematic Reviews) (registration number: CRD42021295755) and published (Jadhakhan et al., 2022). Assuming homogeneity between studies, we planned to conduct a random effect meta-analysis with and without low quality studies. This review is conducted and reported in accordance with the Preferred reporting System and Meta-Analysis (PRISMA) 2020 statement (Page et al., 2021) and conducted following the Cochrane Handbook for Systematic Review of Interventions (Higgins et al., 2021) (Supplementary material 2).

Studies were included if they were randomized or non-randomized controlled trials that investigated exercise or physical activity interventions for people with non-specific spine-related pain, with recorded measures of kinesiophobia taken at baseline and post-intervention. The inclusion and exclusion criteria of this review was determined using the Population, Interventions, Comparators, Outcomes and Study design (PICOS) framework (Richardson et al., 1995; Akers et al., 2009).

Screening results from the database searches were exported into a digital library using the Endnote version 20 reference management software (Clarivate Analytics, Philadelphia, PA, US). Duplicate records were automatically and manually removed. Within the digital library, the titles and abstracts of the articles were reviewed by two reviewers (RS and FJ) independently. Both reviewers discussed the included studies along with studies that required discussion for potential inclusion. Studies comparing two or more eligible exercise interventions were discussed between RS and FJ for inclusion only if the effects of the interventions were separately measured. In the event of disagreement between the two reviewers, a third reviewer (DF) adjudicated the eligibility of the article. In the event of full text availability for some of the selected studies, the lead author was contacted via email twice with a follow-up email sent 2 weeks apart. For the selected articles, full texts were acquired and screened using the same process with both reviewers (RS and FJ) to see if they met the inclusion criteria. A PRISMA (Page et al., 2021) flow diagram describes the process of the inclusion and exclusion of studies and reasons for exclusion from the latest screening stage (Figure 1).

The data from each study was extracted by both RS and FJ and organized into a pre-determined data extraction sheet. Data items extracted from the eligible studies were: authors, publication year, title of the study and study design, country/setting, the characteristics of the participants (age, gender, ethnicity, spinal pain diagnosis, length of diagnosis), sample size, duration/frequency of interventions, length follow-ups, outcome measure, statistical methods, results and findings. Any discrepancies were resolved by discussion and re-visiting the relevant study. A third reviewer (DF) was available to mediate any disagreement in data extraction. If any information was missing or incomplete, an initial attempt was made to contact the study authors and a follow-up email sent 2 weeks after to retrieve the missing data. Descriptive data was extracted from the included studies and were summarized in a Microsoft Excel spreadsheet (Supplementary material 3).

Risk of bias was assessed using Version 2 of the Cochrane risk-of-bias tool for randomized trials (RoB 2) (Sterne et al., 2019). The RoB-2 tool consists of five domains of bias: bias arising from the randomization process, bias due to deviations from intended interventions, bias due to missing outcome data, bias in the measurement of the outcome, and bias in the selection of the reported result. The Risk Of Bias in Non-randomized Studies of Interventions tool was planned to be used to assess the risk of bias of non-randomized studies of interventions however all included studies were randomized trials. Two reviewers (RS and FJ) independently assessed each of the included studies. A third assessor (DF) was available if needed. The reviewers (RS and FJ) used the RoB 2 Cribsheet to follow the algorithm to determine assessment of bias for the individual domains and then present the overall judgement of the study as either low risk, some concerns, or a high risk of bias.

After the evidence was collected and summarized, the assessment of certainty in the body of evidence was conducted in accordance with the Grading of Recommendations Assessment, Development and Evaluation (GRADE) guidelines (Guyatt et al., 2011) and performed by one person (FJ) using the GRADE rating guidance presented in the Cochrane Handbook (Balshem et al., 2011; Schünemann et al., 2021). Consistent with GRADE, the quality of the summary evidence was assessed as high, moderate, low or very low. For each study, the following domains were assessed: imprecision, inconsistency, indirectness, risk of bias including publication bias. Applicability of results were categorized by the study interventions and rated when making judgement about the quality of evidence presented in the included studies (Guyatt et al., 2011).

Given the significant clinical and statistical variation between the studies included in this review, it was not possible to pool effect estimates of exercise/physical activity on kinesiophobia using meta-analysis. Variations was reported in extracted effect measures, population and measures used to ascertain kinesiophobia (e.g., TSK and FABQ). There was also difference among studies in sample size, interventions (including their duration and frequencies) length of follow-ups. Instead, we summarized effect estimates (mean difference or standardized mean difference) with 95% confidence intervals (CI) where appropriate in the included studies. Standardized Mean Difference (SMD) was extrapolated from reported (Cohen's d) values, exploring mean difference between groups; an effect size of 0.8 or greater was considered a relatively large effect size between two means in the sample population. In the event of missing data we attempted to contact the author(s) at least twice by email. A final reminder was sent to corresponding authors/co-authors if no response was received following our initial email. We initially planned to quantify heterogeneity using the Cochrane Q-test and the I² with corresponding 95% (CI) where appropriate. Higher I² values (>50%) (Higgins et al., 2021) indicate larger degrees of heterogeneity pertaining to variability in effect size estimates between studies (Higgins et al., 2021). Recorded measures of kinesiophobia in individuals aged (≥18 years) with chronic non-specific spine-related pain (i.e., neck pain, LBP, and thoracic pain) were extracted from each study and a narrative summary of the outcome of the included studies was presented.

Due to different types of spine-related pain reported, studies were grouped by the type of spine-related pain (neck pain, thoracic pain and LBP) for synthesis rather than pooling across all spine-related pain.

The characteristics of the included studies are presented in Table 1. The 17 studies selected were published between the years 2011 to 2023. The 17 selected studies (Nassif et al., 2011; Miyamoto et al., 2013, 2018; Da Luz Junior et al., 2014; Vincent et al., 2014; Cruz-Díaz et al., 2017, 2018; Keane, 2017; Zadro et al., 2019; Galan-Martin et al., 2020; Tagliaferri et al., 2020; Akodu et al., 2021; Vicente-Campos et al., 2021; Martins de Sousa et al., 2022; Cana-Pino et al., 2023; Hernandez-Lucas et al., 2023; Ogunniran et al., 2023) involved 1,354 participants. Variation in chronic non-specific spine-related pain, exercise/physical activity and measures used to assess kinesiophobia contributed to the significant level of heterogeneity between the studies and reported effect estimates. Most studies were performed in Spain (n = 6), followed by Brazil (n = 4), Australia (n = 2), Nigeria (n = 2), England (n = 1), France (n = 1) and the United States (n = 1). Most studies were conducted in a community clinic setting and University Laboratory (n = 8), followed by outpatient units (n = 3), primary care centers (n = 2), physical activity training unit (n = 2), private clinic (n = 1) and participant homes (n = 1). Studies included in this review were mostly randomized controlled trials (n = 9), followed by single blinded RCTs (n = 6), double blinded RCT (n = 1) and one repeated measures RCT (n = 1). Across all the studies, 64% (n = 871) of participants were female and 36% (n = 483) were male. The ethnicity of the participants was not reported in any of the studies. Thirteen studies recruited people with chronic LBP, three studies recruited people with chronic neck pain, and one study recruited people with (any sort of) chronic spine-related pain. The length of follow-up across the studies were between 4 weeks and 12 months. Frequency of exercise/physical activity was between 30 min and 1 h per session.

Key features affecting the methodological quality of each reviewed study are presented in Table 2. There was significant risk of bias detected across the studies, with the overall risk of bias considered high or with some concerns of bias. This was largely because of blinding, allocation sequence concealment and deviation from intended intervention. Five studies (Nassif et al., 2011; Miyamoto et al., 2013, 2018; Zadro et al., 2019; Galan-Martin et al., 2020) were rated as having high risk of risk, ten (Vincent et al., 2014; Keane, 2017; Cruz-Díaz et al., 2018; Tagliaferri et al., 2020; Akodu et al., 2021; Vicente-Campos et al., 2021; Martins de Sousa et al., 2022; Cana-Pino et al., 2023; Hernandez-Lucas et al., 2023; Ogunniran et al., 2023) with some concerns of bias and two (Da Luz Junior et al., 2014; Cruz-Díaz et al., 2017) deemed low risk. Most studies adequately described the follow-up period, but attrition rate was poorly defined. Only two studies had a low risk of bias in all criteria of the checklist (Da Luz Junior et al., 2014; Cruz-Díaz et al., 2017). For all studies, the analytical approach utilized was considered appropriate. In seven studies (Vincent et al., 2014; Keane, 2017; Tagliaferri et al., 2020; Akodu et al., 2021; Vicente-Campos et al., 2021; Hernandez-Lucas et al., 2023; Ogunniran et al., 2023) consisted of a small sample size which is a major limitation and demonstrate a lack of adequate sample size calculation.

For all 17 studies, the result of the GRADE was very low suggesting the true effect is likely to be substantially different from the estimated effect (Schünemann et al., 2021) (Table 3), thus suggesting very little confidence in the effect estimates. Three studies (Da Luz Junior et al., 2014; Cruz-Díaz et al., 2017, 2018) used two types of exercise intervention (Mat and equipment based Pilates) and a GRADE ranking was undertaken. The GRADE ranking indicates low certainty that the Pilates interventions are effective at reducing fear of movement with relatively small sample size RCT's indicating larger fully-powered trials are needed to adequately test this assumption. Another study (Akodu et al., 2021) investigated the effect of neck stabilization and Pilates exercise on kinesiophobia measured by TSK. The GRADE ranking indicated low certainty that there is no effect on kinesiophobia following Pilates and neck stabilization exercise. Two studies (Miyamoto et al., 2013, 2018) investigated the effect of the frequency of Pilates and education on fear of movement measured by TSK; the GRADE ranking was very low for both interventions, meaning that we have little certainty that the observed effects are the true effects of these interventions. Two further studies (Keane, 2017; Galan-Martin et al., 2020) tested the effectiveness of land based stretching and Aqua stretch exercise and these were rated as very low. Four studies (Nassif et al., 2011; Vincent et al., 2014; Tagliaferri et al., 2020; Vicente-Campos et al., 2021) examined the effect of general strength and conditioning, muscle strengthening, motor control exercise with manual therapy Hypopressive Abdominal Gymnastics and Lumbar extension resistance exercise on fear of movement measured by either the TSK, TSK-11 (Spanish version) and FABQ; the GRADE rating was either low or very low, suggesting little evidence that the observed effects are the true effects of these interventions. Another study (Zadro et al., 2019) assessed the effect of video game-based exercise (Wii Fit U) on fear of movement measured by TSK; the GRADE rating was also very low. Another study (Cana-Pino et al., 2023) examined the effect of supervised/laser guided exercise and neuroscience education (PNE) on fear of movement measured by TSK-11 (Spanish version), the GRADE ranking was very low suggesting little evidence that the observed effects are the true effects of these interventions. Ogunniran et al. (2023) examined the effects of kinesiology taping and core stability exercise on clinical variables in patients with non-specific chronic low back pain. Fear of movement/Kinesiophobia was measured by TSK, the GRADE rating was very low. Another study (Hernandez-Lucas et al., 2023) tested the effects of the Back School-based intervention on non-specific neck pain in adults. The Back schools are an educational and exercise programs with lessons given to patients or workers by a therapist with the aim of treating or preventing low back pain. Fear of movement/Kinesiophobia was measured by TSK-11 (Spanish version). The GRADE ranking indicated low certainty that there is no effect on treating/preventing low back pain following the Back School intervention. Additionally, in a study (Martins de Sousa et al., 2022) evaluating the effect of high frequency (HF) or low frequency (LF) transcutaneous electrical nerve stimulation (TENS) in a specific therapeutic exercise program for the treatment of patients with chronic neck pain; were rated as low. Fear of movement/Kinesiophobia - measured by TSK.

Findings from the included studies have been pooled to describe the effect exercise has on kinesiophobia for people with different types of spine-related pain.

One study (Akodu et al., 2021) investigated the effect of neck stabilization exercise and Pilates on kinesiophobia and found that both neck stabilization exercise (Z = −3.077; p = 0.002) and Pilates (Z = −2.994; p = 0.003) significantly reduced kinesiophobia levels at 8 weeks following the intervention compared to dynamic isometric neck exercises over the same period. This study indicates Pilates and neck stabilization exercises are effective in reducing kinesiophobia over 8 weeks in people with chronic neck pain. Another study (Martins de Sousa et al., 2022) evaluating the effects of high and low frequency transcutaneous electrical nerve stimulation (TENS) and a series of therapeutic exercises which include performing flexion, extension, inclination, and rotation movements of the cervical spine for the treatment of patients with chronic neck pain. The study found no significant clinical difference between participants allocated to the therapeutic exercise and placebo-TENS (group 1) and participants in the low TENS group (group 2) and kinesiophobia [mean difference between the groups−0.40 (−4.64, 3.84), p > 0.05], mean difference placebo TENS and high TENS was also not significant [mean difference between the groups −0.31 (−5.68, 5.06), p > 0.05]. The Back schools-based intervention, an educational and training programs with lessons given to patients or workers by a therapist with the aim of treating or preventing back pain was used by Hernandez-Lucas et al. (2023). This study investigated the effects of a Back schools-based intervention on non-specific cervical pain in an adult population. The study found a significant treatment effect between the experimental group (people attending the Back school program) compared to control (People who did not attend the Back school program) on TSK-11; [mean difference 7.0 (95% CI: −8.3 to −5.4), p < 0.001, g = 2.04].

Pilates was an intervention for people with chronic LBP in five studies, with four of them showing beneficial effects in reducing kinesiophobia. One study found that both mat Pilates and equipment-based Pilates significantly reduced kinesiophobia over 6 weeks and 12-week intervention period, and both differences were significant (p < 0.05) when compared to the control group (Cruz-Díaz et al., 2017). A significant reduction in TSK scores were observed at 12 weeks following Mat Pilates, with mean scores of 31.7 (+/−3.2) at 12 weeks compared to 34.5 (+/−4.1) at baseline. Equipment based Pilates also showed a significant improvement in TSK scores, with mean scores of 32.0 (+/−3.6) at 12 weeks compared to 36.5 (+/−3.9) at baseline. In contrast Da Luz Junior et al. (2014) compared a six-week equipment-based and mat Pilates interventions and found that after 6 month there was a statistically significant improvement in kinesiophobia [mean difference = 4.9 points (95% CI 1.6 to 8.2)] following equipment-based Pilates. These results suggests that equipment-based Pilates may have a longer-term effect on kinesiophobia levels than mat Pilates.

A study (Miyamoto et al., 2018) investigating the frequency of Pilates administered (once, twice and three times per week) in three groups of patients compared to advice/education, found that Pilates significantly reduced kinesiophobia at a 6 weeks follow-up. Participants having Pilates twice a week showed the greatest reduction in kinesiophobia, baseline mean 40.8 (+/−7.5) compared to 37.4 (+/−8.7) at 6 weeks. However, no significant difference was found for any of the three groups at 6- and 12-months follow-up. Additionally, Cruz-Díaz et al. (2018) found that after a 12-weeks Pilates exercise programme, the exercise group showed a significant improvement in kinesiophobia when compared to the control group at six [mean change 5.5 (+/−0.7), p < 0.001] and 12 weeks follow-up [mean change 5.0 (+/−0.8), p < 0.001]. Conversely, another study (Miyamoto et al., 2013) found that Pilates combined with an educational booklet resulted in no significant between group differences (booklet group v/s Pilates) at a 6 months follow-up [adjusted mean difference 0.6 (95% CI: −1.8 to 3.1), p = 0.61].

Keane (2017) assessed the effect of land and water-based stretching compared to a control group in people with chronic LBP and found that after 12 weeks, only water-based stretching significantly (baseline mean = 37.1 v/s 12 weeks mean = 28.8, p = 0.03) reduced kinesiophobia. Nassif et al. (2011) investigated the effect of major muscle group training in people with LBP compared to a control group that had no direct intervention. Strengthening exercises significantly improved kinesiophobia post intervention [baseline mean 46.7 (+/−6.8)] at 2 months follow up [mean 41.6 (+/−6.93), p < 0.001]. Additionally, a general strength conditioning exercise programme reduced kinesiophobia [mean difference −6.6 (−9.9, −3.2, p < 0.001)] compared to a motor control (combined with manual therapy) exercise programme at the 6 months follow-up (Tagliaferri et al., 2020). On the other hand, Vicente-Campos et al. (2021) found no significant improvement in kinesiophobia following Hypopressive Abdominal Gymnastics programme compared to a control group [mean difference −2.00 (95% CI: −4.75 to −0.75, p = 0.15] after 8 weeks. Furthermore, in a study of obese older adults with chronic LBP, the effect of two exercise protocols [total body resistance training (TOTRX) and a lumbar extension resistance exercise training (LEXT)] were compared to a control group (standard care) for 4 months on fear of movement or re-injury and avoidance behavior. At the four-month follow-up, there was no statistical difference between the two exercise groups: TSK: TOTRX: mean baseline 24.5 (+/6.6) at 4 months mean 21.0 (+/−6.9); LEXT mean baseline 25.2 (+/−6.7) at 4 months mean 20.9 (+/−5.9). FABQ: TOTRX: mean baseline 13.2 (+/−14.2) at 4 months mean 8.3 (+/−10.5); LEXT mean baseline 11 (+/−5.9) at 4 months 9.1 (+/−7.2). Both exercise modalities shows improvement in fear of movement and avoidance behavior Vincent et al. (2014). Zadro et al. (2019) investigated the effect of a video game-based exercise programme against a control group for 8 weeks on fear of movement/re-injury. The results showed that there were no significant between group difference in fear of movement/re-injury (β = −2.97, 95% CI = −6.14 to 0.21, p =0 0.07) post-intervention at 8 weeks. Cana-Pino et al. (2023) examined the effect of two exercise modalities: Supervised Exercise (SE)/Laser Guided Exercise (LGE) and Pain Neuroscience Education (PNE) on fear of movement/kinesiophobia in participants with Non-Specific Chronic Low Back Pain (NSCLBP). The result showed a significant between-group difference post-intervention scores at 20 weeks in terms of kinesiophobia (TSK-11) (p < 0.05) and a high effect size (d = 0.81). Another study (Ogunniran et al., 2023) examined the effects of Kinesiology taping (KT) and Core-stabilization exercises (CSE) separately and in combination on Kinesiophobia in patients with NSCLBP. The results showed that there was statistically significant difference in kinesiophobia (p < 0.001) across the three groups at the end of 8 week post-intervention. KT and CSE showed the greatest improvement in kinesiophobia at 8 weeks (21.08 ± 3.75 95% CI: 18.70 to 23.47), CSE (34.79 ± 7.89 95% CI: 29.81 to 39.73) and KT (36.40 ± 8.40 95% CI: 30.39 to 42.41) compared to baseline KT and CSE (42.54 ± 7.37 95% CI: 37.47 to47.20), CSE (42.24 ± 7.64 95% CI: 37.98 to 48.33) and KT (41.31 ± 9.10 95% CI: 34.32 to 46.28).

There were no studies that specifically investigated people with chronic thoracic pain. However, a study by Galan-Martin et al. (2020) investigated the effect of an 11-week physical exercise programme (combined with education) on people with chronic spinal (cervical, thoracic, LBP, and combined) pain and found a significant improvement in kinesiophobia levels in the exercise group at 6 months when compared to baseline scores and the control group.

The current review has several strengths. The review protocol was registered in PROSPERO, published (Jadhakhan et al., 2022) and was conducted and reported in accordance with the PRISMA 2020 statement (Page et al., 2021). Two independent reviewers were involved in study selection, data extraction and quality assessment, and a third reviewer was available to ensure overall methodological consistency and to resolve any disagreements. To ensure an exhaustive review of the available literature, a comprehensive search strategy was implemented with broad inclusion criteria. However, there are some limitations which should be noted. Significant heterogeneity was found between the included studies, particularly pertaining to methods of ascertaining exercise/physical activity, variation in chronic non-specific spine-related pain and measures used to assess kinesiophobia which precluded meta-analysis. Furthermore, due to limited data on ethnicity and gender, sub-group analyses to explore the effect of these variable was not possible. The generalisability and applicability of these findings may be reduced, because of the setting, population and criteria to ascertain kinesiophobia. Additionally, there were limitations in the study selection process such as the search was restricted to English language publications.