The Use of Sensors to Prevent, Predict Transition to Chronic and Personalize Treatment of Low Back Pain: A Systematic Review

Non-specific low back pain (NSLBP) is a highly prevalent condition that implies substantial expenses and affects quality of life in terms of occupational and recreational activities, physical and psychological health, and general well-being. The diagnosis and treatment are challenging processes due to the unknown underlying causes of the condition. Recently, sensors have been included in clinical practice to implement its management. In this review, we furthered knowledge about the potential benefits of sensors such as force platforms, video systems, electromyography, or inertial measure systems in the assessment process of NSLBP. We concluded that sensors could identify specific characteristics of this population like impaired range of movement, decreased stability, or disturbed back muscular activation. Sensors could provide sufferers with earlier diagnosis, prevention strategies to avoid chronic transition, and more efficient treatment approaches. Nevertheless, the review has limitations that need to be considered in the interpretation of results.


Introduction
Low back pain (LBP) is considered as one of the most common causes of disability in adults, involving high costs for society [1].It is estimated that between 5.0% and 10.0% of cases will suffer from chronic LBP at some point in their lives [2].Global disease research has shown that back pain has a major impact on indirect costs, which can account for between 50% and 89% of the total cost [3].It is the leading cause of disability, costing more than USD 100 million annually in the USA [3].
Non-specific LBP (NSLBP) is defined as LBP in which it is not possible to detect a specific cause.Patients diagnosed with NSLBP represent 90-95% of the cases of LBP [4,5].It is known that NSLBP may result from the interaction of different mechanical, anatomical, and psychosocial factors, amongst others, which makes NSLBP assessment challenging for clinicians [4,5].Chronic back pain diagnosis involves the understanding of the complex interaction amongst the different factors involved in LBP.People with NSLBP are not a homogeneous group, and therefore, the identification of sub-groups of patients with NSLBP who share some characteristics remains a challenge for clinicians in order to provide the best treatment and reduce the burden of the condition.Because of it, an appropriate assessment is essential for providing effective treatment of NSLBP and preventing the transition to a chronic condition, as well as monitoring the patient's response to treatment in a personalized way [6,7].
Although the most frequent assessments in clinical practice for patients with LBP are made through self-reported measures, sensors have been progressively incorporated in clinical practice to improve the assessment process [8].The inclusion of sensors in NSLBP diagnosis aims to complement the aforementioned self-reported measures with reliable, objective, and multi-dimensional parameters such as kinematic, kinetic, or electromyographic outcomes.Additionally, sensors could potentially identify common features inside the NSLBP population, which could benefit clinicians by grouping patients with similar patterns, which is a necessary challenge for clinical practice, as discussed above.Finally, sensors could be used by patients to monitor themselves and control their progress in a multi-dimensional way, identifying changes in parameters possibly induced by the pain experience itself or treatment approaches.Nevertheless, their applicability in obtaining valid information needs to be further explored before incorporating these data into the clinical decision-making process.In connection with the latter, the current literature has suggested numerous types of sensors and parameters that may serve to identify NSLBP and it is unknown which ones are more useful in clinical practice [9].
Recently, some reviews have been performed to analyze the use of sensors in patients with LBP from different approaches, such as (i) the analysis of the relationship between the sitting time and the immediate increase in perceived LBP [10], (ii) the applicability for assessing spine movement [9], (iii) the comparison of lumbo-pelvic movement differences [11], (iv) the use of sensors to monitor worker's activities and to assess biomechanical risk [12], or (v) assessing the contribution of the hip joint and the lumbar segments to assist in determining subgroups within the LBP population [13].However, there are no reviews that have analyzed the use of sensors in the specific context of personalized medicine for NSLBP.
Therefore, the objectives of this review are (1) to analyze how sensors are used to assess different outcomes related to NSLBP in clinical studies related to prevention, prediction of transition to a chronic condition, or personalization/effectiveness of treatment, (2) to determine if there are sub-groups inside the NSLBP population who have specific characteristics that can be identified with sensors and that require specific treatment strategies.

Materials and Methods
This systematic review was performed according to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) standard protocol [14] and has been registered in the International Prospective Register of Systematic Reviews (PROSPERO) (reference number: CRD42022301566).

Eligibility Criteria
Study eligibility and selection was based on the PRISMA checklist PICOS (P-Participants; I-Interventions; C-Comparators; O-Outcomes; S-Study design).
Studies were included in accordance with the following criteria: (1) patients over 18 years of age, including at least a group with chronic (>3 months/12 weeks) NSLBP, with no restrictions regarding race or gender; (2) studies using a sensor system to assess at least one of the following outcomes: lumbar kinematics, lumbar kinetics, and/or electromyography (EMG); (3) studies comparing the chronic NSLBP population with the healthy population, different groups of patients with NSLBP, or the same group pre-post-treatment, or studies exploring the association between outcomes and psychosocial measurements; (4) all types of clinical trials (randomized controlled trials (RCTs), matched controls, cohorts), only considering peer-reviewed journal papers to ensure the credibility of the sources; and (5) publications written in English or Spanish, from 2010 until the present day.
Studies were excluded if they met the following criteria: (1) nonhuman research or trials on animals; (2) publications including people aged under 18 years old; (3) patients who have a specific origin of LBP or have undergone a spinal surgery; (4) publications not specific for the NSLBP population or studying the reliability or validity of different kinds of sensors; or (5) other types of publications such as book sections, conference abstracts or generics, protocols, reviews, or meta-analysis.

Information Sources and Search Strategy
A search of PubMed, MEDLINE, Scopus, and Web of Science (WOS) databases was conducted to collect all available evidence from 2010 to 5 February 2022.Concerning search terms, 3 categories were defined.The first one was related to the population ("low back pain" and "LBP"), the second one was related to the sensor system used ("sensor", "sensors", "sensing", "inertial", "IMU", "IMUs", "accelerometer", "gyroscope", "goniom*", and "body sensor network"), and the third one was related with the outcome measures ("kinematic*", "kinetic*", "movement", "physical activity", "sedentary behaviour", "electromyography", and "EMG").The choice of these search terms was established after a preliminary literature search and keyword identification.The full search strategies were developed concerning the database in which they were used and the filters applied in Scheme S1 (Supplementary Materials).Furthermore, reference searching was performed to identify additional studies that the database search might have missed.

Study Selection
To decide if the studies met the inclusion criteria, two reviewers screened each report (D.L. and I.R.).They worked independently to avoid bias, following the same methodology after an agreement on the realization of search equations, and then compared their findings.In case of disagreement, a third researcher intervened to reach consensus (M.G.).
Firstly, all the registers were retrieved from the four databases and were introduced in the bibliographic gestor "Mendeley version 1.19.8" in order to remove duplicate publications.After that, a first screening of articles was performed and the articles that could comply with the inclusion criteria regarding the information available in the title and abstract were selected.The process progressed to a second screening phase, in which the studies that passed the previous phase were read in full, and those that fulfilled all the inclusion criteria were selected.

Data Extraction Process
To collect data from studies, the two reviewers worked independently (D.L. and I.R.), and afterwards the reviewers compared the extracted data for consistency.Both reviewers were required to achieve a consensus.All inconsistencies were resolved by discussion between the two data extractors.Any disagreement between the data extractors was solved by involving a third person (M.G.).

Data Items
The following data were extracted from each study: authors, study design, participants' characteristics (size sample, sex, age, diagnosis, pain duration), intervention (intervention or assessment), comparator, outcomes, and main results.

Assessment of Methodological Quality and Risk of Bias
Three different tools were used to assess the risk of bias of the articles included in the systematic review depending on the type of article.The two reviewers worked independently (D.L. and I.R.) and afterwards the results were compared.No intervention from a third reviewer was needed.
For RCTs and studies with random assignment, the "Cochrane Risk of Bias 2.0" (RoB2) tool was used [15].The assessment of risk of bias was based on 'assignment to intervention' for all five domains: (1) randomization process, (2) deviations from intended interventions, (3) missing outcome data, (4) outcome measurement, and (5) selection of the reported result.An overall risk of bias judgement was made for each outcome and each time point as either 'low risk', 'some concerns', or 'high risk' of bias.
On the other hand, for studies with non-randomized interventions, the ROBINS-I tool was used, that is, "Risk Of Bias In Non-randomised Studies-of Interventions" [16], which assesses the risk of bias through seven different domains: (1) confusion, (2) selection of study participants, (3) classification of interventions, (4) deviations from previously stipulated interventions, (5) lack of data or information, (6) measurement of the outcomes, and (7) selection of the exposed results.Each domain was scored based on responses to questions such as 'low risk', 'moderate risk', 'high risk', 'critical risk', or 'no information'.Once the score for each domain was evaluated, the total score for the study was obtained.
Finally, for cross-sectional, case-control, and case-series studies, the National Heart, Lung, and Blood Institute (NHLBI) tools were used: "Quality Assessment Tool for Observational Cohorts and Cross-Sectional Studies", "Assessment of the quality of case and control studies", and "Tool for assessing the quality of Case Series studies" [17], respectively.The quality of the studies was assessed by answering a series of questions about the study and the quality of the articles was scored as "good", "favourable", or "poor".

Study Selection
As a result of the search through the search strategies described and the filters mentioned, a total of 836 publications were found (PubMed: 85; Medline: 67; WOS: 419; Scopus: 265).After removing duplicate studies, 586 were screened by title and abstract.No additional records were added through other information sources.A total of 134 publications progressed to the second screening phase for a full-text reading, and finally, 24 articles were included in the systematic review for qualitative analysis.The flow diagram (Figure 1) shows in more detail the process of search and selection of studies, together with the different reasons for exclusion, following the PRISMA criteria.

Study Characteristics
The data extracted from the 24 articles are presented in Table 1, grouped according to the sensor system they used to measure the outcomes of interest.

Study Characteristics
The data extracted from the 24 articles are presented in Table 1, grouped according to the sensor system they used to measure the outcomes of interest.

Studies Comparing Population with Chronic NSLBP and Healthy Population
The findings of the different articles selected have made it possible to obtain different results in relation to the outcomes studied.In the case of kinematic measures, most articles found significant differences between the population with and without chronic NSLBP (decreased range of motion, speed, and acceleration, especially in the lumbar flexoextension movements) [21,22,25,27].However, a certain discrepancy has been observed in the angles of the spine in the sagittal plane, since two articles did not find significant differences [21,26], while two others did observe significant differences [20,23].One of these observed that the chronic NSLBP group and healthy group had visual differences in the kyphosis-lordosis axis and significant differences in excursion angle [20].The other one showed that the chronic NSLBP group, in comparison with the healthy group, had greater error magnitude and variability in the sagittal plane in repositioning the spine in neutral position [23].In spatiotemporal gait parameters, some differences were found between both groups (shorter step duration, reduced gait cadence and speed [19], higher base of support [30]), but most of them were not significant.
In the case of kinetic measures, the selected studies found that the population with chronic NSLBP presents, with respect to the healthy population, (1) a decrease in gait stability [18]; (2) an increased risk of falls and instability in functional tests [28]; (3) an increase in displacement of the center of pressure when the surface was unstable [31]; and (4) significant differences in postural sway while standing, being greater in the group with chronic NSLBP [31,32].Regarding the proprioceptive strategy, a single article states that patients with chronic NSLBP depend more on the ankle strategy [31].
The EMG measures indicate that the population with chronic NSLBP has greater muscle activation in the posterior muscles (lumbar spinal erector and lumbar multifidus) in different tasks: walking [30], prolonged standing [20,32,33,35], and sitting [20].However, no consensus has been reached for the anterior muscles, even though most of the articles show significant differences [20,30,32,34,35].Results also indicate that populations with chronic NSLBP have less isometric force production in the lower trunk [34] and take longer to reach peak muscle contraction [35].

Studies Comparing Outcomes between Groups of NSLBP
The articles that compared different population groups with NSLBP by active extension/flexion patterns or by chronicity found statistically significant differences between groups that are not evident if they are not sub-classified.For example, in one of them, the velocity of the center of pressure and the total mean velocity were significantly lower in the sub-groups resulting from the "O'Sullivan Classification System" with respect to the healthy population [29].In another study, the sub-groups resulting from the same classification system had a worse estimate of the neutral position of the spine compared to the healthy population [20].Moreover, one article differentiated chronic and acute NSLBP, finding that the only variation was that the pain was more intense in those with chronic NSLBP in a weight-lifting test and that the decrease in pain was associated with kinematic changes only in the population with acute NSLBP [40].Finally, one article used the "STarT Back Assessment Tool" to form groups according to risk of NSLBP progression depending on the movement of the trunk in different directions and planes, in which they found differences in linear acceleration in extension movement, and it was suggested that an extension of 15º in right rotation could be a useful position to identify high-risk NSLBP patients [39].

Studies Comparing Outcomes before and after Treatment
The studies which performed an intervention obtained positive results for the population with chronic NSLBP, improving their disability [37,38], pain intensity [36][37][38]40], postural control [24,36,38], and kinematics [40].In addition, better results were obtained with specific exercise treatments than with general protocols in two of the studies [36,38], achieving improvements in pain, disability, and in the sense of spinal repositioning.

Assessment of Methodological Quality and Risk of Bias
A total of five RCTs were evaluated with the RoB2 tool.The five articles [24,34,[36][37][38] evaluated obtained a "low risk" of bias, in the five domains and in the total (Figure 2).extension of 15º in right rotation could be a useful position to identify high-risk NSLBP patients [39].

Studies Comparing Outcomes before and after Treatment
The studies which performed an intervention obtained positive results for the population with chronic NSLBP, improving their disability [37,38], pain intensity [36][37][38]40], postural control [24,36,38], and kinematics [40].In addition, better results were obtained with specific exercise treatments than with general protocols in two of the studies [36,38], achieving improvements in pain, disability, and in the sense of spinal repositioning.

Discussion
The main aim of this review was to analyze the use of sensor systems in the assessment of the population with chronic NSLBP in the context of personalized medicine.The results suggest that sensor systems could be effective in identifying some characteristics that could be indicative of chronic NSLBP, playing an important role in the diagnosis, prevention, and progression of this condition.However, there are few outcomes in which the different studies reach a consensus, and the articles related to each other are also very heterogeneous.
With reference to the range of motion, angular velocity, and linear acceleration, statistically significant differences in at least one of these measures have been found between the groups in five publications [21,22,25,27,39], especially in the lumbar flexo-extension movement, with greater differences in extension according to two of them [25,39].On the other hand, Papi et al. [42] in their systematic review did not reach firm conclusions as to which kinematic and kinetic outcomes could be used to assess people with chronic NSLBP and differentiate them from the rest of the population.In their review, they found some discrepancies regarding differences in range of motion, suggesting that asymmetry of motion, angular velocity, and acceleration might be more effective in assessing people with chronic NSLBP.Regarding spinal angles, the studies reviewed did not agree on the main differences between people with and without chronic NSLBP, finding some studies that showed differences between populations [20,23], while others did not [21,26].According to Koch and Hänsel [43], there was moderate evidence that spinal angles do not differ between subjects with and without NSLBP, although they did not find high-quality studies.One possible reason for this discrepancy could be, as pointed out by Laird et al. [21], the variability of the anatomy of the hip, pelvis, and spine between individuals.
Differences were observed between the population with and without chronic NSLBP in spatiotemporal parameters in the three articles that analyzed them [18,19,30].However, only some outcomes were statistically significant, such as length of stride, step, and gait cycle duration.Smith et al. [44] carried out a systematic review on the characteristics of gait in the population with LBP, including all types of temporality and causality, finding that individuals with chronic LBP used reduced gait speed and stride length compared to healthy individuals and that the reduction in kinematic and kinetic parameters could be part of a strategy to reduce the demands of walking in patients with LBP.The coordination of gait phases has also been altered in several studies included in the current review, suggesting that people with chronic NSLBP have a less predictable and more variable gait [18,19].These findings agreed with the results of Smith et al., who suggest that the population with NSLBP has greater variability of movement patterns in the same phase and relate it to a dysfunction in the dissociation of movement between the trunk and pelvis [44].
A single article stated that patients with chronic NSLBP depend more on the ankle strategy than the healthy population [31].These results are consistent with those of Claeys et al. [45], who suggested that individuals with NSLBP depended more on the proprioceptive strategy of the ankle and less on that of the hip and spine.It suggested a reduced capacity to switch to a more multi-segmental postural control strategy and less ability to rely on back muscle proprioceptive inputs during complex postural conditions, which leads to decreased postural robustness [45].However, another study included in the review found that anticipatory postural adjustments did not present significant differences between groups with and without chronic NSLBP, with muscle activation time being similar in both groups [36].
Four of the studies analyzed it either directly or indirectly [18,28,31,32], and their results suggest that the population with NSLBP could have less stability in walking and standing and greater risk of falling than the healthy population.
Regarding EMG measures, most of the articles showed significant differences between the population with and without chronic NSLBP, suggesting that EMG could be an effective method in the early detection of NSLBP and could play an important role in the choice of treatment.Several articles stated that the chronic NSLBP group had higher muscle activation in the posterior muscles (lumbar multifidus and erector spinae) than the healthy population [20,30,32,33,35].Other articles also found significant differences in the activity of the anterior muscles, but the results were contradictory, with some studies showing more muscle activation in the rectus abdominis [32], external oblique [20,32], and transverse fibers of the internal oblique [20], and others finding less activation in these muscles [30,34].On the other hand, Ghamkhar and Kahlaee [46] assert that greater muscular activity in the hip and spine in standing may be a strategy to compensate for instability in the spine, which is consistent with the results of this review.
In connection with the latter, a previous review of gait parameters in the LBP population by Smith et al. showed an increased activation of the lumbar paraspinal muscles, which could be related to different phenomena: (1) the reduction in movement and the protection of sensitive and painful tissues; (2) compensation of muscle weakness due to atrophy and fatty infiltration of the multifidus in response to back pain; or (3) proprioceptive alteration [44].Regarding the abdominal muscles in walking, these same authors concluded that it is highly variable between individuals and that it is more dependent on the speed of locomotion [44].With regard to muscle strength, most articles measured it in order to determine the % of force used by the different groups in the different functional tasks.One of them compared the maximal voluntary contraction between the groups with and without chronic NSLBP, and it turned out to be lower in the participants with chronic NSLBP in trunk flexo-extension [32].In addition, one of the articles included in this review showed significant differences in muscle thickness of the transversus abdominis and lumbar multifidus [35], and another ensured that the time to the highest peak of contraction was longer in the population with chronic NSLBP [34].
Two articles related kinematic outcomes with psychosocial parameters.One of them found statistically significant associations between some ranges of movement and "Oswestry Disability Index" scores [41], while another found no significant associations [25].Another study related psychosocial parameters with impaired trunk control in people with chronic NSLBP [18].The rest of the studies used different questionnaires and scales but did not analyze their relationship with outcomes measured by the sensors.
Regarding the analysis of the samples by subgroups, two articles [20,29] used the "O'Sullivan classification system" for LBP, which identifies flexor patterns and active extension patterns.In both articles, greater differences were found with the control group when a sub-classification of the population with NSLBP was made.These findings suggest that patients could benefit from more specific, cheaper, and more effective treatments using this rating scale.Papi et al. [42] state that this sub-classification highlights the different pain mechanisms and how they affect function and suggest taking it into consideration in future studies.In another study included in this review [39], the "STarT Back Assessment Tool" was used to classify participants with NSLBP according to the risk of worsening the pathology, observing significant differences in some kinetic outcomes depending on the subgroup.However, none of the articles used sensor systems to classify patients with NSLBP.The absence of studies with sensors on this topic indicates that this is still a topic that requires research in order to determine whether the use of sensors in the diagnosis and assessment of patients could be beneficial as a complement when carrying out assessments, treatments, and/or more specific or personalized follow-up of patients with NSLBP.
In terms of treatment, supervised therapeutic exercise is recommended as the first line of treatment for chronic LBP, despite strong evidence that therapeutic exercise alone is no more effective than conventional physiotherapy methods on chronic LBP [47].In the current review, two articles [36,38] obtained better results with a specific treatment (therapeutic exercise and postural intervention, respectively) than with the standard care, serving the use of sensors to objectify these improvements.In the European Guideline for the approach of chronic NSLBP, there is limited evidence (Level C) that a home program with specific exercises (individualized, based on a previous evaluation) is more effective than a home program using general exercises, such as those usually included in "back school" activities [47].In addition, of the six articles that compare outcomes before and after treatment [24,29,34,[36][37][38], five of them used physiotherapy treatment (kinesiotaping, postural interventions, therapeutic exercise programs, or feedback with different strategies) [24,34,[36][37][38].Of these, four obtained positive results [24,[36][37][38], suggesting that physical therapy could be an effective tool for improving the condition of patients with chronic NSLBP.Likewise, the improvements in the "Oswestry Disability Index" score, pain intensity, postural control, and kinematics suggest that patients with chronic NSLBP are capable of improvement, so each case must be approached individually and personalized to find the most appropriate treatment for the improvement of the patient.
Among the limitations of this review, the heterogeneity of the articles should be highlighted.Publications in which the participants had to perform different functional tasks (functional tests, standing, walking), used different sensor systems, and measured the outcomes with different parameters were included.In addition, to maximize the number of articles, studies that not only compared the population with chronic NSLBP with the healthy population, but also compared the outcomes between groups of NSLBP and assessed the changes before an intervention were included.For this reason, it has been difficult to reach a firm consensus that can be extrapolated and to analyze data quantitatively with a metaanalysis.If there were an adequate number of articles of each type, a more specific review could have been carried out, obtaining more useful and applicable results for clinical practice.
On the other hand, the percentage of articles that obtained little risk of bias in relation to the total is low.According to the "Oxford Table of Evidence Levels" [48], the RCTs constitute level 2 evidence ("1" being the highest level of evidence and "5" the lowest); non-randomized and cross-sectional intervention studies represent level 3; and case-control studies and case-series are included in level 4. This, together with the lack of RCTs that met the inclusion criteria in the review, meant that the risk of bias of the evidence found was high.Likewise, due to the lack of translation resources, the search was limited to articles published in Spanish and English, which could generate a cultural and language bias.
For future research, we suggest performing the following projects: (1) RCTs that use sensors to test their diagnostic capacity; (2) reviews that determine what type of sensors are more effective when assessing a given task in the population with NSLBP; (3) RCTs that assess the efficacy of sensors to classify patients with NSLBP based on one outcome and perform treatments supported by scientific evidence to see if they are more effective than without classifying the population; and (4) studies in which the relationship is investigated between chronic NSLBP and psychosocial parameters such as kinesiophobia, anxiety, or depression, among others.In the same way, it is necessary to establish a protocol to assess these patients which specifies the task or the measurement outcomes that provide the maximum possible information in order to make the best medical decisions.Authors should discuss the results and how they can be interpreted from the perspective of previous studies and of the working hypotheses.The findings and their implications should be discussed in the broadest context possible.Future research directions may also be highlighted.

Conclusions
In this review, it has not been possible to determine which sensors and outcomes are useful for determining specific characteristics of the population with chronic NSLBP and preventing its appearance and progression or determining the best possible treatment strategy.The heterogeneity of the bibliographic material and the lack of RCTs have made it difficult to obtain consistent results.Even so, findings of interest for clinical practice have been achieved.(1) The population with chronic NSLBP differs from the healthy population in kinematic and kinetic parameters (angular velocity, acceleration, and stability) measured with sensor systems; (2) activation of spinal muscles is greater in the population with chronic NSLBP than in the healthy population, in different functional tasks assessed with EMG; (3) LBP classification systems can form groups that have shown differences in kinetic and kinematic characteristics when they are assessed with sensors, and therefore can benefit from early diagnosis and specific treatments; (4) patients with chronic NSLBP are able to improve with personalized physical therapy treatments; and (5) sensors have the potential to be useful tools in the assessment of NSLBP, but more research is required in this field.

Figure 1 .
Figure 1.PRISMA flow diagram from search strategy.

Figure 2 .
Figure 2. Risk of bias in RTCs with RoB2.Abbreviations: D1, bias due to randomization process; D2, bias due to deviations from intended interventions; D3, bias due to missing data; D4, bias in measurement of outcomes; D5, bias in selection of the reported result.The studies shown in the figure correspond to the following references in the reference list: [24,34,36-38].

Figure 2 .
Figure 2. Risk of bias in RTCs with RoB2.Abbreviations: D1, bias due to randomization process; D2, bias due to deviations from intended interventions; D3, bias due to missing data; D4, bias in measurement of outcomes; D5, bias in selection of the reported result.The studies shown in the figure correspond to the following references in the reference list: [24,34,36-38].

Sensors 2023 , 20 Figure 3 .
Figure 3. Risk of Bias in Studies with non-Randomized Intervention with ROBINS-I.Abbreviations: D1, bias due to confounding; D2, bias in selection of participants into the study; D3, bias in classification of interventions; D4, bias due to deviations from intended interventions; D5, bias due to missing data; D6, bias in measurement of outcomes; D7, bias in selection of the reported result.The studies shown in the figure correspond to the following references in the reference list: [31-33].

Figure 3 .
Figure 3. Risk of Bias in Studies with non-Randomized Intervention with ROBINS-I.Abbreviations: D1, bias due to confounding; D2, bias in selection of participants into the study; D3, bias in classification of interventions; D4, bias due to deviations from intended interventions; D5, bias due to missing data; D6, bias in measurement of outcomes; D7, bias in selection of the reported result.The studies shown in the figure correspond to the following references in the reference list: [31-33].

Figure 4 .
Figure 4. Risk of bias in cross-sectional studies, case-control and case-series with NHLBI.Abbreviations: Q, question; NA, not applicable; NR, not reported.The studies shown in the figure correspond to the following references in the reference list: [18-23,25-30,35,39-41].

Figure 4 .
Figure 4. Risk of bias in cross-sectional studies, case-control and case-series with NHLBI.Abbreviations: Q, question; NA, not applicable; NR, not reported.The studies shown in the figure correspond to the following references in the reference list: [18-23,25-30,35,39-41].

Table 1 .
Characteristics of included studies.
• Maximal LyE ↑ in LBP affected by environment in A-P (* LBP/CG), not in M-L (stability ↓ in LBP) • Changes in trunk motor control in LBP related to NPRS, TSK, and deficit in ADLs (Activities of Daily Living) • Step and gait cycle duration, cadence and velocity ↓ in LBP (* LBP/CG) • Step length of both foot ↑ in LBP (* LBP/CG) • • AE and VE in sitting and standing in thoracic and lumbar spine ↑ in LBP (* LBP/CG) • No associations in thoracic and lumbar CE in sitting and standing (* when LBP is classified) • TrIO and EO activity in sitting and standing ↑ in LBP (* LBP/CG)/LM and ILPT activity ↑ in LBP (LBP/CG) • LM activity in standing ↑ in FP (* FP/CG) • % in flexion ↓ in LBP (more pelvic contribution)
• Similar patterns LBP/CG, except for a slight displacement ↑ in kyphosis-lordosis angle (visually) and lateral bending at maximal rotation and lateral bending excursion angle ↑ in LBP (* LBP/CG) • Sensor group ↑ in LBP and CG, for both tasks and in lumbar spine and hip (* control/mirror/sensor) • No group effect (control/mirror/sensor) in post-intervention questionnaire • LBP equally capable of ↑ lumbopelvic movement control