Does pain influence control of muscle force? A systematic review and meta-analysis

Background and Objective: In the presence of pain, whether clinical or experimentally induced, individuals commonly show impairments in the control of muscle force (commonly known as force steadiness). In this systematic review and meta-analysis, we synthesized the available evidence on the influence of clinical and experimental pain on force steadiness. Databases and Data Treatment: MEDLINE, EMBASE, PubMed, CINAHL Plus and Web of Science databases were searched from their inception to 19 December 2023, using MeSH terms and pre-selected keywords related to pain and force steadiness. Two independent reviewers screened studies for inclusion and assessed their methodological quality using a modified Newcastle–Ottawa risk of bias tool. Results: In total, 32 studies (


| INTRODUCTION
During voluntary submaximal movements, the force exerted by individuals inherently varies, fluctuating around an average value (Enoka et al., 2003).Such force variability stems from neuromuscular noise and other external factors (Oomen & van Dieën, 2017).The ability to generate steady force output during submaximal voluntary contractions is termed force (or torque) steadiness (Tracy & Enoka, 2002).Minimizing these force fluctuations is vital for maintaining physical functionality, and its deterioration could compromise the precision of voluntary actions, affecting joint stability, coordination and overall movement (Deering et al., 2017;Maenhout et al., 2012).
The smooth generation of force relies heavily on the perception of force, a component of proprioception (Ager et al., 2020).Proprioception is commonly compromised in the presence of acute, chronic or experimentally induced pain, likely due to altered afferent information from the affected area and/or central changes (Ager et al., 2020;Lee et al., 2010;Röijezon et al., 2015;Salahzadeh et al., 2013;Sjölander et al., 2008;Treleaven, 2008).Proprioceptive and nociceptive inputs share a complex neurological pathway and may interfere with each other at different levels of processing (Ager et al., 2020).Moreover, musculoskeletal injuries can further impair proprioception by structural damage to tissue (Röijezon et al., 2015).Therefore, pain-induced proprioceptive alterations could potentially degrade our ability to control muscle force.
Assessment of force steadiness typically involves measuring the variability of the force output, quantified by the standard deviation (SD) or coefficient of variation (CoV) of the force or torque signal over time, which serve as indicators of absolute and relative neuromuscular control (Fiogbé et al., 2019), where higher SD and CoV values denote less force steadiness.Additionally, measures such as approximate and sample entropy are sometimes used (SampEn and ApEn), as they offer insights into the complexity and regularity of the force signal.Lower entropy values suggest a more predictable and regular force output, while higher values indicate greater complexity and potential instability in motor control (Smith et al., 2021).
The current literature offers inconsistent findings on the effect of pain, be it clinical or experimental, on force steadiness.Some studies suggest that conditions such as chronic low back pain (CLBP) (Arvanitidis et al., 2022;Miura & Sakuraba, 2014), chronic neck pain (CNP) (Muceli et al., 2011) or patellofemoral pain syndrome (PFP) (Ferreira et al., 2021) are associated with decreased force steadiness in specific isometric tasks, while others report no such changes (Camargo et al., 2009;Crowley et al., 2022;Hirata et al., 2015;Martinez-Valdes et al., 2020).Similarly, studies have shown that experimentally induced pain is associated with reductions in isometric fifth finger abduction (Farina et al., 2012) and knee extension (Poortvliet et al., 2019) force steadiness, while others did not observe changes in dorsiflexion steadiness in the presence of experimentally induced pain (Martinez-Valdes et al., 2021, 2020).Given this variability, the aim of this systematic review is to synthesize the available evidence to understand if pain, whether it be clinical or experimental, influences force steadiness during isometric and dynamic voluntary contractions.

| METHODS
This review was conducted in line with the 2020 guidelines of the preferred reporting items for systematic reviews and meta-analyses (PRISMA) (Page et al., 2021) as evidenced in Data S1.The protocol for this review was pre-registered on the International Prospective Register of Systematic Reviews (PROSPERO) with the registration number CRD42020196479 on 21 July 2020 and has been published (Arvanitidis et al., 2021).While adherence to the published protocol was rigorous, minor deviations were necessary during data synthesis.Specifically, it was not possible to subgroup data based on the muscle/ joint assessed or the implementation of visual feedback.To address the variability among studies, additional heterogeneity measures were evaluated.Furthermore, the standardized mean difference (SMD) was utilized for all studies instead of odds ratios, as it was considered more appropriate for the analysis.

| Eligibility criteria
The inclusion criteria for this systematic review were defined by using a modified version of the PICOS framework, which included population, indicator, comparison, outcomes and study design (Smith et al., 2021).Due to of force steadiness training in alleviating patients' symptoms and enhancing their functional performance.This could potentially lead to the development of innovative therapeutic approaches for individuals suffering from musculoskeletal pain.
the nature of the studies, the original term 'Intervention' was replaced with 'Indicator', in line with previous adaptations for observational studies (Devecchi et al., 2019;Devecchi et al., 2021).

| Population
Studies were eligible for inclusion if they focused on adults (≥18 years) with musculoskeletal pain (clinical), defined as 'pain experienced in muscles, tendons, bones or joints that arises from an underlying disease classified elsewhere' -chronic secondary musculoskeletal pain (Perrot et al., 2019) or 'pain that is characterised by significant emotional distress or functional disability, and cannot be attributed directly to a known disease or damage process' -chronic primary musculoskeletal pain (Nicholas et al., 2019;Perrot et al., 2019).Additionally, studies investigating experimentally induced musculoskeletal pain, that is, pain arising from the sensitization of nociceptors in subcutaneous tissues through electrical or chemical stimulation of the muscle and/or joint (Reddy et al., 2012), were also included.These models were specifically chosen because they can induce deep tissue pain, closely mirroring the characteristics of clinical musculoskeletal pain (Graven-Nielsen et al., 1997).A requirement for inclusion was a control group of asymptomatic individuals or a crossover design where the same participants served as their controls in a pain-free state.There were no restrictions based on gender or ethnicity.The exclusion criteria applied to studies involving pain from exercise-induced soreness (delayed onset of muscle soreness; DOMS) or fatigue, or thermal models, as well as participants with neuromuscular and neurological disorders, systemic inflammatory diseases or a history of surgery/fracture in the area of interest, to eliminate confounding factors.

| Indicator
Studies selected included those using dynamometers or similar instruments, such as force or torque sensors, to assess steadiness of force or torque or related measures of variability.Studies employing force platforms for assessing postural steadiness, such as during standing, were not considered.Inclusion was not limited by the provision of visual feedback during trials; studies with and without visual feedback on the force or torque output were eligible for inclusion.All forms of contractions, whether measured at an absolute level or relative to the individual's maximal voluntary contraction (MVC), were considered and there was no restriction for the side of the body assessed (i.e.dominant or non-dominant).

| Comparison
Studies that were selected included those that compared force or torque steadiness in voluntary contractions between painful and non-painful states.Comparisons were conducted either within participants, as in studies of experimentally induced pain, or between groups, such as between pain-affected participants and a control group.Additionally, comparisons encompassed multiple assessments and measurements taken both pre-and post-task.

| Outcomes
The primary outcome assessed was the measurement of force or torque steadiness.For this purpose, the following force steadiness measures were considered: the coefficient of variation (CoV) and SD of force or torque, as well as the root mean square (RMS) of the force or torque signal during force template matching tasks.Studies that evaluated torque complexity through measures such as SampEn or ApEn were also included.The inclusion was restricted to quantitative studies that measured force or torque steadiness or variability.Studies that investigated other aspects of muscle force control, such as force accuracy (i.e. the ability of an individual to produce a force that matches a specified target force as closely as possible; force reproduction tasks, etc.) (Pranata et al., 2017), were excluded.

| Study design
Scoping searches indicated that observational studies primarily explore the research question of this systematic review.Consequently, only observational studies employing quantitative methods, specifically cross-sectional, casecontrol and cohort studies, were included.Non-original works such as systematic and narrative reviews, as well as other study types, were excluded.To reduce potential bias, the search encompassed studies in all languages.However, due to constraints of time and resources, studies not written in English were excluded.Details of any excluded studies are documented in the PRISMA flow diagram (Figure 1).

| Information sources
Electronic databases were searched from their inception to 19 December 2023 and included MEDLINE (Ovid Interface), EMBASE (Ovid Interface), PubMed, CINAHL Plus (EBSCO Interface) and Web of Science (Clarivate Analytics).While the ZETOC database was initially considered for the search, it was ultimately not accessed as the database was closed as of 1 August 2022.Tailored search strategies were formulated for each database, incorporating Medical Subject Headings (MeSH) where applicable to refine and optimize the search (Richter & Austin, 2012).
In parallel with the electronic database search, hand searches of select journals were also performed.Additionally, proceedings from the World Congress of Biomechanics, the International Society of Biomechanics, the International Society of Electrophysiology and Kinesiology and the World Confederation for Physical Therapy were scrutinized, covering the years 2016 to 2023.Authors of studies that appeared to be eligible were contacted to confirm the publication status of their work.To further reduce the chance of publication bias, the reference lists of all included studies were manually reviewed to identify any relevant studies that might have been missed during the initial search.

| Search strategy
The lead author (MA) conducted the search without imposing limitations on date, format, design, geographical area or language.To guarantee both comprehensiveness and reproducibility of the search, it was crafted following initial scoping searches and with guidance from a skilled Health Sciences Librarian, a member of the University of Birmingham's Research Skills Team.The full electronic search approach for the MEDLINE (Ovid Interface) database is detailed in Data S2.This approach combined MeSH terms and keyword searches to optimize the search yield.The search was slightly adapted for different databases, but consistency was ensured.These changes involved alterations in MeSH terms and syntax.For example, the 'ADJ' operator was switched to 'NEAR' in the Web of Science database and to 'N' in the CINAHL Plus (EBSCO Interface) database.The search strategies for all databases are detailed in Data S2.

| Selection process
All search results were imported into EndNote Version 20 (Clarivate Analytics) by one reviewer (MA) for data management.Duplicates were automatically removed by the software.For screening, the search results were also made available in separate folders for each independent reviewer (MA and AS) and were assessed using a pretested screening form.Titles and abstracts were initially screened by two independent reviewers (MA and AS) to categorize studies as definitely eligible, ineligible or doubtful.Any disagreements or doubtful studies were discussed between reviewers, and in cases of disagreement or uncertainty, a third reviewer (EM-V) was consulted.
The two reviewers independently conducted both stages of the selection process, including the screening of titles/abstracts and full-text evaluation.Any disagreement between reviewers was resolved by the third reviewer (EM-V).The level of agreement between reviewers was evaluated using the kappa (κ) statistic.

| Data collection process and data items
Data extraction was conducted by one reviewer (MA) using a tested extraction form, designed to gather essential elements aligned with the review's objectives.The accuracy of the extracted data was verified by a second reviewer (AS).When additional clarification was needed for ambiguous or incomplete data, authors were contacted with a 2-week reply window.Failure to respond led to the study's exclusion for ambiguity.
As detailed earlier, data relevant to each element of our inclusion criteria were extracted.When studies included additional groups not relevant to our review, only data relevant to our specific interest groups were extracted.

| Risk of bias assessment
Two independent reviewers (MA and AS) employed two adapted versions of the Newcastle-Ottawa Scale (NOS) to assess risk of bias; one version for clinical pain studies adapted from the case-control version and another for experimental pain studies adapted from the cohort version.It is important to note that the experimental pain studies had repeated measures or cross-over design and were experimental in nature.We selected to use the adapted cohort version of the NOS for these studies to ensure consistency within the manuscript, to adhere to our previously published protocol (Arvanitidis et al., 2021) and because it was considered to be more appropriate to the case-control version.The changes and justifications for each are included in Data S3.Briefly, in both versions of the tool, the 'exposure' domain in the original NOS was specifically modified to an 'outcome' domain to fit the nature of our studies better while maintaining the essence of the questions.Scoring ranged from 0 to 9, with higher scores indicating lower risk of bias.Quality categorization for each study was then done as 'good', 'fair' or 'poor', according to established thresholds.Further information on our rationale for selecting the NOS tool for the risk of bias can be found in our published protocol (Arvanitidis et al., 2021).

| Synthesis methods
Both narrative and meta-analyses methods were used to synthesize the data from the included studies, to explore the influence of pain on force steadiness (Deeks et al., 2019;McKenzie et al., 2019).Studies were grouped by the type of pain (clinical or experimental) for methodological consistency, and the narrative synthesis included the outline and tabulation of study characteristics (McKenzie et al., 2019).This allowed studies, not included in the meta-analysis, to be interpreted and the influence of their findings to be considered, alongside other clinicalor methodological-related information from each study.
In preparation for the meta-analysis, studies were further subgrouped based on the outcome investigated (e.g.CoV or SD of force/torque).Considering our research question and the limited number of studies per muscle/ joint, subgroup analysis by muscle/body region or type of contraction was not performed.Only studies with available or retrievable mean ± SD data on force steadiness were included for meta-analysis.Before conducting the meta-analysis, descriptive statistics were used to appropriately format the data.For example, if a study reported the standard error (SE) instead of the SD, this was calculated by using the following formula.
Additionally, for within-study calculations where multiple observations per group were reported, a weighted average was employed to adjust for any differences in sample size across these observations.A pooled SD was used to pool the individual variances of these observations into a cohesive measure of dispersion.The formulas used are reported below: where for the formula for the weighted mean (x w ), x i is the mean of observation i and n i is the sample size associated with observation i and k the total number of observations.For the SD pooled formula (SD pooled ), SDi is the standard deviation of observation i, n i is the sample size for observation i and k is the number of observations.All meta-analysis procedures and forest plot generation were performed using the R software (4.2.1) (Team, 2022) and the 'meta' package (Balduzzi et al., 2019) by selecting a multi-level random-effects model to adequately accommodate the anticipated heterogeneity across the included studies and adjust the weight of studies that were included more than one time in the model (Deeks et al., 2019).Data pooling was performed using the SMD and the 'metacont' function within the 'meta' package which is designed for continuous outcomes using 95% confidence intervals in line with others (Devecchi et al., 2021;Pethick et al., 2022) and guidelines (Deeks et al., 2019).The Knapp-Hartung adjustment (Knapp & Hartung, 2003) was also used to calculate the 95% CI around the pooled effect estimate (Jackson et al., 2017).
Even though we implemented a random-effects model to address heterogeneity statistically, it is important to recognize that this approach does not eliminate heterogeneity; it merely accounts for it (Deeks et al., 2019).Therefore, as recommended (Deeks et al., 2019;IntHout et al., 2016) we further explored heterogeneity within the data, by calculating and reporting additional measures.Rather than relying solely on the I 2 statistic, which has some limitations in providing insights into the nature of heterogeneity, we also incorporated τ 2 (τ 2 ).τ 2 , which is a quantitative estimate of the extent to which the true effects vary across studies (Deeks et al., 2019).As recommended (Deeks et al., 2019;IntHout et al., 2016;Jackson et al., 2017;Teichert et al., 2023), the between-study variance (τ 2 ) was estimated using the restricted maximum-likelihood estimator.In accordance with recent recommendations (IntHout et al., 2016), we also incorporated the calculation of the prediction interval for the pooled effect size into our analysis.This approach offers a broader perspective on the potential range of true effects within comparable studies (Deeks et al., 2019;IntHout et al., 2016).By including the prediction interval alongside the summary estimate and confidence interval (CI), our reporting illustrates the expected range of true effects in future studies and can simplify its clinical interpretation (Deeks et al., 2019;IntHout et al., 2016).

| Sensitivity analyses
Due to the expected high heterogeneity commonly seen in pain studies, we also conducted a sensitivity analysis (Deeks et al., 2019).This involved excluding each study in turn to check the consistency of our meta-analysis results, as detailed in Data S4.This step confirmed the findings' robustness and assessed the impact of excluding each study on overall heterogeneity.Additionally, considering that the presence of outliers and influential cases may affect the validity and robustness of the conclusions from a meta-analysis, we quantitatively explored potential outliers using deletion diagnostics known from linear regression (i.e.externally standardized residuals, DFFITS values, Cook's distances, covariance ratios, DFBETAS values, estimates of τ 2 and Q when each study is removed in turn, diagonal elements of the hat matrix and the weights (%) given to the observed outcomes during model fitting) that can also be adapted to the context of meta-analysis as described previously (Viechtbauer, 2010;Viechtbauer & Cheung, 2010).This was performed using the 'influence' function from the 'metafor' package in R (Viechtbauer, 2010).Sensitivity analyses were also performed by removing studies that did not use visual feedback during the contractions to maintain consistency, as the presence or absence of visual feedback could have been a confounding factor.This approach allowed us to standardize the analysis across the more commonly reported condition of 'with visual feedback'.It was not possible to perform a sensitivity analysis excluding studies with visual feedback due to the limited number of studies assessing force steadiness without visual feedback.Additionally, we removed studies of poor quality to assess their influence on the overall results.Using this comprehensive approach to identify influential cases and conduct sensitivity analyses, we aimed to ensure the robustness of our meta-analysis results.

| Certainty of evidence
The cumulative evidence from the meta-analyses (for each outcome and type of pain) was also appraised by the two independent reviewers (MA and AS) using the grading of recommendations assessment, development and evaluation (GRADE) method, assessing cumulative strength and quality.This process involved five steps, as described previously (Goldet & Howick, 2013), with the final evidence quality classified as 'high ', 'moderate', 'low' or 'very low'.Observational studies initially received a low-quality rating, which was then adjusted based on certain criteria.Evidence was upgraded for substantial effect sizes and clear dose-response trends.Conversely, risk of bias, study inconsistencies, imprecision, indirectness and publication bias led to downgrades (Balshem et al., 2011).Publication bias was assessed by generating funnel plots and visually inspecting their symmetry, as well as using Egger's regression test to quantify the presence of publication bias (Egger et al., 1997;Sterne & Egger, 2001).We applied this quality assessment across all studies and subgroups in our review, leading to tailored evidence interpretation recommendations in line with guidelines for observational studies (Dekkers et al., 2019;Mueller et al., 2018).

| Search results and study selection
The results of the search and selection process are outlined in Figure 1.The database searches initially resulted in 5301 records, with nine additional records/abstracts identified through targeted handsearching of conferences.After removing duplicates, 3230 records from databases and 9 from conference hand searching underwent title and abstract screening, with the inter-reviewer agreement reflected by a high kappa coefficient κ = 0.9.Full-text screening was then applied to 78 records from the databases/journal search, excluding the other 9 records for reasons specified in Figure 1.The inter-reviewer agreement for the full-text screening was perfect, with κ = 1.00.The review ultimately included 32 studies.
Some studies that may seem relevant for this review were excluded.Specifically, seven studies (Hollman et al., 2021;Maenhout et al., 2012;Ross et al., 2015;Saccol et al., 2014;Sarah Ward & Bryant, 2018;Zanca et al., 2010;Zanca et al., 2013) were excluded because the patient groups did not experience or report pain during the assessment.Moreover, one study (Dinsdale et al., 2023) was excluded because the assessment of force steadiness was performed during an endurance contraction, and muscle fatigue could have been a confounding factor.Lastly, another study was excluded due to its randomized controlled trial design, which was outside the methodological scope of this review (Mista et al., 2016).Furthermore, this study was quite different from the others included, as it consisted of two groups that attended three experimental sessions (day 0, 2 and 4), with one group receiving nerve growth factor and the other isotonic saline.The only usable data would have been from day 2, but even the control group (i.e. the individuals who received the isotonic saline injection) experienced some minimal pain, and a learning effect could not be excluded, considering that the individuals had already performed the task twice on day 0.

| Characteristics of included studies
This review included 32 studies (19 clinical pain and 13 experimental pain), and their characteristics are summarized in Tables 1 and 2. Most studies used CoV and/or SD of force to quantify force steadiness, while only a few used other measures such as SampEN and ApEn for force complexity and root mean square (RMS) for force steadiness.

SD (N)
The variability of task-related force was not altered by pain.Hypertonic saline (0.5 mL, 5.8%) into TA muscle.The painful sensation lasted for the full set of contractions, reaching its peak:

Martinez
6.3 (1.6) 1 min after the injection and ceasing completely within 500 s.Pain was consistently felt under the electrode grid by all participants (mainly under the 2nd, 3rd and 4th columns of the grid); however, three participants also experienced referred pain at the lateral malleoli.

Characteristics of pain Force steadiness task and contraction intensity
Outcome measure
Isometric knee extension, eccentric and concentric force steadiness contractions against a dynamometer (Biodex Medical Systems Inc, Shirley, New York, USA).For each contraction, the target was set at 10%

MVC.
A monitor was used to show the actual knee extensor force to the participants.

ApEn SampEn
The presence of pain did not influence torque complexity, measured as ApEn or SampEn In Experiment 1, participants were seated with the arm and first four digits secured, and the fifth digit attached to a load cell to measure isometric abduction force.They performed a 60-s isometric abduction of the fifth digit at 10% MVC under three conditions: baseline, isotonic saline and hypertonic saline.In Experiment 2, participants, with their foot in an isometric force brace, executed a 4-min isometric dorsiflexion at 25% MVC.After a 20-min rest, one leg was treated with hypertonic saline (to induce discomfort) and the other with isotonic saline.Following the infusion, the isometric dorsiflexion was repeated, with visual force feedback provided via an oscilloscope in both experiments.

CoV (%)
↓ Force steadiness after the injection of hypertonic saline for both muscles (ADM and TA) when measured as SD of force (p < 0.01 for both) and CoV of force (p < 0.05 for both).
T A B L E 2 (Continued) found no significant differences specifically investigated jaw clenching steadiness in individuals with neck pain; however, when the same research group specifically assessed jaw clenching force steadiness in a cohort of individuals with TMDs, that is, where the pain was task specific, a significant difference was observed.

| Experimental pain studies
Thirteen studies involving a total of 174 participants were evaluated to investigate the effect of experimental pain on force steadiness.For 12 studies, hypertonic saline was the experimental pain model used, and one further study (Del Santo et al., 2007) used ascorbic acid.Four studies assessed the effect of experimental pain on knee force steadiness (n = 61) (Poortvliet et al., 2019;Rice et al., 2015;Salomoni et al., 2013;Smith et al., 2021) and two on the ankle (n = 30) (Martinez-Valdes et al., 2021, 2020).One study assessed the low back (n = 12) (Hirata et al., 2015), one the shoulder (n = 9) (Bandholm et al., 2008), one the elbow (n = 12) (Mista et al., 2015) and one the fifth finger (n = 11) (Farina et al., 2012).Additionally, three of these studies investigated the effect of experimentally induced pain on force steadiness in various regions.One study evaluated this at the elbow and fifth finger (n = 8) (Del Santo et al., 2007), another one at the ankle, elbow and knee (n = 15) (Salomoni & Graven-Nielsen, 2012) and a third one at the fifth finger and ankle (n = 16) (Yavuz et al., 2015).Most of these studies (10 of the 13) (Bandholm et al., 2008;Del Santo et al., 2007;Farina et al., 2012;Martinez-Valdes et al., 2021;Mista et al., 2015;Poortvliet et al., 2019;Rice et al., 2015;Salomoni et al., 2013;Salomoni & Graven-Nielsen, 2012;Yavuz et al., 2015) observed reductions in force steadiness when experimental pain was induced.In contrast, three studies found no differences in force steadiness (Hirata et al., 2015;Martinez-Valdes et al., 2020) or torque complexity (Smith et al., 2021) between conditions with and without pain for the same individuals.However, the SD of torque data provided by Smith et al. (2021), which was included in the meta-analysis, revealed that force steadiness is reduced in the presence of experimental pain when quantified with this variable.The average age of participants ranged from 22 to 53 years, while the peak pain intensity scores reported by the patients during the task ranged from 2.6 to 6.3 out of 10 (NRS or VAS scales).

| Risk of bias
Comprehensive summaries of the risk of bias scores are presented in Tables 3 and 4 respectively.In the assessment of risk of bias across the 19 clinical pain studies, the scores varied from 4 to 9 out of a possible 9.The quality of studies was rated as 'poor' for 10, 'good' for 8 and 'fair' for 1 (Table 3).The most common reasons for 'poor' ratings were from potential biases introduced by not matching participants on age and/or gender as part of the study methodology, and not reporting full participant recruitment information.The assessment was also performed across 13 experimental pain studies with the scores ranging between 5 and 9 out of 9.The quality was rated as 'good' in eight of the studies, 'fair' in three and 'poor' in two (Table 4).

| Results of syntheses and certainty of evidence
Meta-analyses for both clinical and experimental pain studies were conducted using the CoV and SD of force measures.However, since some studies utilized either the CoV or the SD measure, but not both, it was not feasible to incorporate all studies into both meta-analyses.As a result, some studies could be included in either the CoVbased meta-analysis or the SD-based meta-analysis, but not both.
3.4.1 | Force steadiness in the presence of clinical pain The results of the meta-analysis that included 13 of the total 19 studies with clinical pain, involving 382 participants with pain and 322 without, demonstrated that there was a significant overall effect of musculoskeletal pain on the CoV of force, indicating that force steadiness is impaired in the presence of pain (SMD = 0.80 [95% CI: 0.31-1.28],I 2 = 88% [95% CI: 82%-92%], τ 2 = 0.55, Q = 124.07,t 15 = 3.52, p = 0.003).Similarly, the results of the meta-analysis that included 14 of the total 19 studies with clinical pain, involving 355 participants with pain and 297 without, demonstrated that there was a significant overall effect of musculoskeletal pain on the SD of force, indicating impaired force steadiness in the presence of pain (SMD = 0.61 [95% CI: 0.11-1.11],I 2 = 87% [95% CI, 80%-91%], τ 2 = 0.66, Q = 114.18,t 15 = 2.60, p = 0.020).The results of the forest plots with meta-analyses are presented in Figure 2.
One study was excluded from both meta-analyses because related data could not be extracted (Crowley et al., 2022).This study did not observe any differences in plantar flexor steadiness (measured as CoV of force) in people with insertional Achilles tendinopathy compared to controls.In the CoV-based meta-analysis, four studies were not included because they quantified force steadiness using the SD of force.Among these, two observed reductions in force steadiness in people with TMD (Testa et al., 2018) and knee osteoarthritis (Hortobágyi et al., 2004) compared to controls, while the other two found no significant differences in individuals with chronic elbow pain (Mista et al., 2018) or people with CNP (Testa et al., 2015).Conversely, in the SDbased meta-analysis, three studies were excluded because they only used CoV of force as the outcome measure with all observing significant reductions in force steadiness in people with CNP (Falla et al., 2010;Muceli et al., 2011) andPFP (Ferreira et al., 2021), which is in line with the findings of the meta-analysis.
3.4.2| Force steadiness in the presence of experimental pain The results of the meta-analysis that included 10 out of the total 13 studies on experimental pain, involving F I G U R E 2 Meta-analyses on the effect of clinical pain on force steadiness during different submaximal voluntary contractions.The mean ± SD of each outcome measure and sample size for each group are reported, alongside the standardized mean difference and 95% confidence interval (95% CI).The analysis is divided into two segments, (a) focuses on the coefficient of variation (CoV) of force/torque, while (b) is based on the SD of force/torque.All heterogeneity measures and overall effect size tests are reported below each meta-analysis.The prediction interval is also depicted as a red line within the graph.The forest plots are organized in ascending order of their effect sizes.The results of the forest plots with meta-analyses are presented in Figure 3. Three studies were excluded from the CoV-based meta-analysis.Among these, two did not observe any differences in elbow flexion (Mista et al., 2015) and trunk extension (Hirata et al., 2015) force steadiness between the experimental pain and baseline conditions, while the other one showed that knee extension force fluctuations were higher in the presence of experimentally induced knee pain (Poortvliet et al., 2019).Four studies were excluded from the SD-based metaanalysis.Among these studies, three showed that force steadiness is reduced in the presence of experimentally induced knee (Salomoni et al., 2013), fifth finger (Farina et al., 2012) and fifth finger and elbow pain (Del Santo et al., 2007).The last one investigated the effect of experimentally induced pain on the force steadiness of multiple muscles and observed reductions in force steadiness only for the knee at higher forces (Salomoni & Graven-Nielsen, 2012).

| Alternative measures of force control
There was inadequate data to conduct meta-analyses on studies involving other measures of force control.Specifically, in the studies with clinical pain, only two used ApEn as a measure of force complexity with one of them showing that people with sub-acromial pain syndrome had lower ApEn values during shoulder abduction and adduction tasks (i.e. a smoother force pattern with less variability) compared to the controls F I G U R E 3 Meta-analyses on the effect of experimental pain on force steadiness during different submaximal voluntary contractions.The mean ± SD of each outcome measure and sample size for each group are reported, alongside the standardized mean difference and 95% confidence interval (95% CI).The analysis is divided into two segments, (a) focuses on the coefficient of variation (CoV) of force/torque, while (b) is based on the SD of force/torque.All heterogeneity measures and overall effect size tests are reported below each meta-analysis.The prediction interval is also depicted as a red line within the graph.The forest plots are organized in ascending order of their effect sizes.(Overbeek et al., 2020), while the other did not observe any differences for ApEn values during a jaw clenching task in people with and without jaw pain (Wang et al., 2018).These studies were included in our metaanalysis as they also assessed steadiness using the CoV or SD of torque.However, another study with clinical pain was not included in any meta-analysis because it only used RMS and SampEn to quantify changes in force control between the two groups.This study showed that RMS was larger in people with CLE, while no differences were observed in terms of the complexity of the force signal (measured using SampEn) (Chen et al., 2023).
From the experimental pain studies, only two used alternative measures of force control to assess the effect of experimental pain on force control.One used both ApEn and SampEn to assess changes in isometric knee extension force control (Smith et al., 2021), while the other only the latter during elbow flexion contractions (Mista et al., 2015), with both studies not observing significant differences in these variables during the painful and nonpainful states. 3.4.4| Certainty of evidence -Sensitivity analyses and GRADE Influence analyses on the main meta-analyses for both clinical and experimental studies, focusing on force/ torque CoV and SD, did not identify any potential outliers, except for the SD-based meta-analysis in experimental pain studies, where one study (Rice et al., 2015) was detected as a potentially influential (Data S4).This study was removed as part of the sensitivity analysis to explore its influence on the overall result as described below.Additionally, the sensitivity analyses included: (i) removal of one study at a time, (ii) removal of studies that did not provide visual force feedback and (iii) removal of studies classified as having poor quality.

Clinical pain studies
Sensitivity analyses indicated that the overall results of the meta-analyses for both force/torque CoV and SD were not influenced by the removal of any single study or by excluding data from studies that did not provide force visual feedback (Data S4).However, removing studies of poor quality altered the overall result, showing no differences in force steadiness in the presence of clinical pain.This change is likely due to the removal of the majority of studies from the meta-analysis (eight and six studies, respectively).This likely does not provide an accurate representation, as many studies were rated poor, primarily due to a lack of information on the comparability of groups within the methodology, which was only reported in the results section.

Experimental pain studies
Removing the study (Rice et al., 2015) that could potentially influence the SD-based meta-analysis resulted in a similar outcome to the meta-analysis that included all studies.Removing each study one at a time had a significant impact on the CoV-based meta-analysis, usually showing no changes in force steadiness in the presence of experimentally induced pain.For the SDbased meta-analysis, the result of no differences was maintained most of the time, but the overall effect became significant when two studies were removed (Martinez-Valdes et al., 2021, 2020).Excluding force/ torque SD data from studies that did not provide visual feedback did not alter the overall result of the main meta-analysis.However, the CoV-based meta-analysis result changed when these data were removed, suggesting that experimentally induced pain does not influence force steadiness.Further exploration identified one study as potentially influential in the model; when this study was removed, the same result as the main metaanalysis was observed.Lastly, removing two studies of poor quality (Del Santo et al., 2007;Farina et al., 2012) did not alter the overall result for the CoV-based metaanalysis, while no studies could be removed from the SD-based meta-analysis since none were classified as having poor quality.
The evaluation of the certainty of evidence with GRADE was conducted independently for studies with clinical and experimental pain, based on the outcome used.This assessment was performed on studies using the CoV and SD of force as measures of force steadiness, due to the limited number of studies using alternative metrics.Any sources of publication bias were assessed by visually inspecting funnel plots and using Egger's regression tests, as shown in Figure 4.
The results showed that for the clinical pain studies, funnel plots were quite symmetrical, and Egger's regression test did not reveal any potential sources of publication bias (torque CoV: p = 0.591, intercept = 0.0427, t = 0.553; torque SD: p = 0.633, intercept = 0.0563, t = 0.491; Figure 4a, b, respectively).In contrast, for the experimental pain studies, the funnel plot for Torque CoV (Figure 4c) suggested potential sources of publication bias, as indicated by the clustering of dots around the 0.4 level of standard error and confirmed by Egger's regression test (p < 0.01, Intercept = −1.96,t = 4.06).However, the funnel plot for torque SD (Figure 4d) did not reveal any significant publication bias (p = 0.12, intercept = −2.14, t = 1.77).
Considering the above and the other domains assessed with GRADE, the findings from the meta-analysis indicate, with moderate and low levels of evidence strength, that force steadiness is impaired in the presence of clinical pain when measured using force CoV and SD respectively.Additionally, it indicates with very low strength of evidence that force steadiness is impaired in the presence of experimental pain when quantified as the CoV of force, but not when quantified using the SD of force.The summary of findings is presented in Table 5.

| DISCUSSION
We investigated the influence of clinical and experimentally induced musculoskeletal pain on force steadiness.Integrating data from 32 studies, our analysis indicated that clinical musculoskeletal pain is associated with a decrease in force steadiness, as indicated by large and moderate effects on the CoV and SD of force.Experimental pain was associated with a reduction in force steadiness only as measured by the CoV of force (moderate effect), and not in the SD of force.

| Does clinical pain influence force steadiness?
Our findings confirm that clinical musculoskeletal pain is associated with reduced force steadiness and align with a previous systematic review's findings indicating that peripheral musculoskeletal conditions are also associated with increased force CoV (Pethick et al., 2022).The high heterogeneity observed is likely attributed to differences in participant characteristics, the diverse nature, severity and duration of pain, methodological differences and the spectrum of musculoskeletal conditions studied.However, sensitivity analysis confirmed our findings' robustness, showing that no single study's exclusion changed the overall result (Data S4).
While most studies support that clinical pain is associated with reduced force steadiness, some exceptions  were observed.Two studies (Testa et al., 2015(Testa et al., , 2017) ) on jaw clenching force steadiness in people with CNP found no significant between-group differences.However, assessments of people with TMD by the same authors showed reduced force steadiness in the patient group (Testa et al., 2018).This indicates that experiencing pain at the assessed location is crucial for such observations.Additionally, variations in force steadiness could be attributed to the assessed force level.Some studies examined a broad range of low-to-high submaximal forces, whereas some examined a single submaximal level.Deficits reported at specific force levels by certain studies (Bandholm et al., 2006;Miura & Sakuraba, 2014) imply that force steadiness deficits could be force-level specific.

| Does experimental pain influence force steadiness?
This review also revealed that experimental musculoskeletal pain is associated with reduced force steadiness when measured as the CoV of force.The meta-analyses showed lower heterogeneity and smaller, more consistent effect sizes.This consistency is likely due to the use of similar pain models and the assessment of comparable anatomical regions in many studies.One study had a notably larger effect size, likely due to its unique use of ascorbic acid to induce pain (Figure 3a).However, the p-values from both the CoV-and SD-based metaanalyses, along with insights from the sensitivity analyses in Data S4, suggest that the removal of some of the  Note: Studies highlighted in bold were excluded from the meta-analysis for the specified outcome.
individual studies significantly influenced the overall effect.Therefore, these findings should be interpreted with caution.
The absence of significant differences in force SD may be due to the characteristics of the force steadiness outcome measures.The CoV quantifies variability relative to the mean force output, reflecting proportional changes, while SD, as an absolute measure, is inherently sensitive to individual strength differences.This distinction is more apparent in data synthesis across multiple studies, where variability in baseline participant strength can substantially influence SD outcomes.Moreover, the measurement unit heterogeneity (e.g.Newtons vs. percentages) across studies might confound comparisons, although this issue was mitigated in our analysis by using SMD for meta-analytic calculations.These factors, coupled with the sample sizes and the number of studies included, could explain the observed discrepancies between CoV and SD findings.Interestingly, although in experimental pain the effect size for torque SD was not significant, the effect sizes for both torque SD and CoV were very similar.This similarity suggests that the overall trend towards impaired force steadiness in the presence of experimental pain is consistent across these measures.However, interpreting CoV results necessitates caution; sensitivity analysis indicated that excluding certain studies substantially alters outcomes, underscoring careful result interpretation and advocating for using both CoV and SD to assess force steadiness in people with pain (Data S4).

| Differences between clinical and experimental pain
The findings suggest that the force steadiness deficits are less marked during experimental pain than in clinical pain conditions, a disparity likely stemming from the fundamental differences between these two pain types.Clinical pain, characterized by its multidimensional and persistent nature, encompasses physical, psychological and emotional elements, along with extensive neurophysiological changes.As clinical pain becomes chronic, significant modifications occur in the nervous system, including alterations in brain structure and function, particularly in areas involved in emotional and sensory processing (Sibille et al., 2016).These alterations include reductions in grey matter volume and white matter integrity, changes in neurotransmitter activity and reduced descending inhibition (Sibille et al., 2016).Such comprehensive changes across multiple levels of the nervous system, coupled with alterations in muscle structure, including atrophy, fatty tissue infiltration and fibre type changes, documented in people with chronic pain, are more likely to alter neuromuscular control than experimental pain (Matheve et al., 2023).
Experimentally induced pain, associated with transient sensitization of nociceptive pathways, differs from clinical pain due to its temporary nature and individuals' knowledge that the discomfort will cease.This understanding may affect their pain perception and response, contrasting with the unpredictability of clinical pain.Previous work further supports this, indicating that experimental pain does not replicate the changes in motor neuron excitability observed in clinical pain conditions across the neuromuscular pathway (Sanderson et al., 2021).In acute experimental pain, individuals might employ more strategies, such as changing movement and muscle patterns, to compensate for pain, which could explain why force deterioration appears less pronounced in experimental studies (Devecchi et al., 2023).

| Physiological explanation of the observed findings
Pain likely impairs muscle force control through altered sensory input from the affected area and/or central changes (Ager et al., 2020).Structural alterations in brain areas essential for proprioceptive input processing or muscle composition changes may also contribute to impaired muscle force control (Pijnenburg et al., 2014;Sterling et al., 2001).These sensory integration changes can alter the efferent response, thereby influencing muscle recruitment strategies employed by individuals to perform muscle contractions.
Alterations in force steadiness may be linked to changes in motor unit behaviour.Reduced modulation in the discharge rates of motor units in the sternocleidomastoid muscle in women with CNP has been observed, indicating a change in neural drive to muscles in the presence of pain (Falla et al., 2010).Additionally, increased motor unit recruitment and firing rates have been observed in early knee osteoarthritis (Ling et al., 2007) and in women with PFP (Gallina et al., 2018) respectively.Experimentally induced knee pain influences force steadiness in healthy individuals (Rice et al., 2015), and notably, changes in motor unit discharge patterns have also been reported in the presence of experimental pain within the same region (Poortvliet et al., 2019).
Recent studies have indicated that the effective neural drive predominantly operates in the low-frequency band (<10 Hz), reflecting the common synaptic input to the motor unit pool, essential for force generation (Farina et al., 2014;Negro et al., 2009).Since muscle force control relies on the amplitude of these oscillations, increased amplitude fluctuations observed in pain conditions may indicate greater variability in synaptic input to α motor neurons.Given the α motor neuron's role in integrating complex inputs and the interplay between proprioceptive and nociceptive signals, the physiological mechanism by which musculoskeletal pain influences force control may vary among different pain types and individuals (Hug et al., 2023).

| Clinical relevance
Impaired force steadiness may lead to suboptimal tissue loading, where a disrupted sense of body position may result in the application of excessive forces in unfavourable positions during movement.This abnormal mechanical stress on tissues can activate nociceptors and cause tissue strain, potentially contributing to the development of additional pain or injuries and the perpetuation of existing symptoms.These impairments likely have clinical relevance and could become targets for treatment for people experiencing musculoskeletal pain.

| Limitations
This review's limitations include the predominance of observational studies in the meta-analyses, which hinders causal inference between musculoskeletal pain and decreased force steadiness.Factors such as reduced activity, medication, sleep and psychological conditions may also influence force steadiness.The relatively small sample sizes of the included studies, particularly in experimental pain studies, also pose a challenge to the robustness of the evidence.Nonetheless, within-participant comparisons mitigate interindividual variability, enhancing statistical power.

| Recommendations for future research
Future research should explore force steadiness at various force levels and during isometric and dynamic contractions in people with pain, to understand changes in force control across different conditions.Both the CoV and the SD of force should be used for assessing force steadiness since they offer unique insights into force variability.Additionally, calculating SD in Newtons based on a submaximal percentage MVC can be problematic due to the non-linear relationship between force output and MVC percentage, especially when considering inter-individual differences in muscle strength.Thus, studies should express SD as a percentage or ensure both SD and the target are on the same scale.
Moreover, although most studies use the term force steadiness and report results in Newtons (for the SD), the measurements actually pertain to joint moments rather than muscle forces.Often, force transducers are attached at positions defined relative to anatomical landmarks, which can create differences in moment arms between individuals.These variations in moment arms can influence the calculated SD, as the lever arm length directly affects the torque produced by a given force.However, this variability should not affect the CoV, as it normalizes the SD by the mean force, thereby reducing the impact of individual differences in moment arm lengths.It is relevant to acknowledge this issue as it introduces an additional source of variability that future studies need to account for when measuring and interpreting force steadiness.
Lastly, given the significant role of visual feedback in force modulation (Limonta et al., 2015), future research should also investigate the underexplored influence of relying solely on proprioceptive input, in the absence of visual feedback, on force steadiness in individuals with musculoskeletal pain.

| CONCLUSION
This review demonstrates that individuals with clinical pain exhibit decreased force steadiness, quantified by both the CoV and SD of force.It also reveals that force steadiness, quantified as CoV, is reduced in individuals subjected to experimentally induced pain, primarily induced through hypertonic saline injections.However, the results of the experimental pain studies should be interpreted with some caution.Future studies should explore the relationship between enhancements in force steadiness and improvements in patients' symptoms and functional performance.

AUTHOR CONTRIBUTIONS
This review was conceived by Michail Arvanitidis, Deborah Falla and Eduardo Martinez-Valdes; searches, study selection, data extraction and data handling were conducted by Michail Arvanitidis and Andy Sanderson.Michail Arvanitidis had a primary role in preparing the manuscript which was edited by Deborah Falla, Andy Sanderson and Eduardo Martinez-Valdes.All authors have approved the final version of the manuscript and agree to be accountable for all aspects of the work.
These journals encompassed PAIN, Journal of Physiology, Journal of Neurophysiology, Acta Physiologica, Journal of Electromyography and Kinesiology, Clinical Biomechanics, Muscle & Nerve, Journal of Orthopaedic & Sports Physical Therapy, Journal of Science and Medicine in Sport, Isokinetics and Exercise Science and Journal of Applied Physiology.Efforts were made to identify any unpublished but relevant studies by directly contacting subject matter experts in the field.To mitigate the risk of publication bias, grey literature was also reviewed, accessed through the British National Bibliography for Report Literature, OpenGrey database, ProQuest Dissertations & Theses Global and EThOS.

F
I G U R E 1 PRISMA flow diagram, adapted from Page et al.(2021).PRISMA, preferred reporting items for systematic reviews and metaanalyses.T A B L E 1 Characteristics of studies examining whether people with clinical pain show differences in force steadiness with respect to asymptomatic controls.0001 for both).This was evident during both the eccentric and concentric contractions and at both low and high torque levels.↓Trunk flexion torque steadiness in people with CLBP (CoV and SD of torque, condition effect, 0001 for both).This was evident during both the eccentric and concentric contractions and at both low and high torque levels.
Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/ejp.4716by Manchester Metropolitan University, Wiley Online Library on [27/08/2024].See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions)on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License and adduction using a one-dimensional force transducer at the wrist.Isometric shoulder ab-and adduction contractions from a standing position, facing a computer for force feedback, with the target arm in external rotation at the side attached to a one-dimensional force transducer at the wrist.The target force was similar for both abduction and adduction and equal to 60% MVC (defined as the lowest absolute value of the MVC in abduction or adduction Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/ejp.4716by Manchester Metropolitan University, Wiley Online Library on [27/08/2024].See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions)on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/ejp.4716by Manchester Metropolitan University, Wiley Online Library on [27/08/2024].See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions)on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/ejp.4716by Manchester Metropolitan University, Wiley Online Library on [27/08/2024].See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions)on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License

F
Funnel plots of the four meta-analyses performed alongside the results of the Egger regression test.Funnel plots are presented for (a) torque CoV for clinical pain studies, (b) torque SD for clinical pain studies, (c) torque CoV for experimental pain studies and (d) TORQUE SD for experimental pain studies.The blue dots in each scatterplot represent the values from individual studies, with effect size on the x-axis and standard error on the y-axis.The middle dotted line represents the overall effect size, and the side dotted lines form a triangle indicating the 95% confidence interval for the expected distribution of studies in the absence of publication bias.The dotted red line represents Egger's regression line, used to test for potential sources of publication bias.Egger's test values of p < 0.05 indicate potential publication bias.

Study Patient population characteristics & self-reported measures Control population characteristics & self-reported measures Force steadiness task and contraction intensity Outcome measure Results
SD (N), CoV (%) ↓ Reduced force steadiness in people with SIS at 35% MVC during concentric contractions for both SD and CoV of force (p < 0.05 and p = 0.03).No differences during isometric contractions.The SIS group was divided into two groups: (1) SIS with the dominant involved side and (2) the nondominant involved side SD (N), CoV (%) No changes in shoulder abduction torque steadiness in individuals with SIS (SD, CoV; p > 0.05).± 6.1 cm); pain intensity (VAS: 0-10): N/A; DASH score: N/A.Isometric wrist extension against a force sensor (model: MB-100, Interface Inc., Scottsdale, AZ, United States).Visual feedback was provided by a monitor.Target force: up to 75%MVC (8 s ramp-up and 8 s rampdown).Each participant performed three trials.SampEn, RMS ↓ Reduced force steadiness (RMS, ramp-up phase), in people with CLE RMS (p = 0.001).No differences when measured as SampEn during the ramp-up phase (p = 0.226).

Study Patient population characteristics & self-reported measures Control population characteristics & self-reported measures Force steadiness task and contraction intensity Outcome measure Results
T A B L E 1 (Continued) 15322149, 0, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/ejp.4716by Manchester Metropolitan University, Wiley Online Library on [27/08/2024].See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions)on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
T A B L E 3 15322149, 0, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/ejp.4716by Manchester Metropolitan University, Wiley Online Library on [27/08/2024].See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions)on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License Risk of bias of studies of experimental pain studies.