Which assessments are used to analyze neuromuscular control by electromyography after an anterior cruciate ligament injury to determine readiness to return to sports? A systematic review

Background Adequate neuromuscular control of the knee could be one element to prevent secondary injuries after an anterior cruciate ligament (ACL) injury. To assess neuromuscular control in terms of time, amplitude and activity, electromyography (EMG) is used. However, it is unclear which assessments using EMG could be used for a safe return to sports (RTS). Therefore, we aimed to summarize EMG-related assessments for neuromuscular control of the knee in adult patients after an ACL injury to decide upon readiness for RTS. Methods This systematic review followed guidelines of Preferred Reporting of Items for Systematic Reviews and Meta-Analyses (PRISMA) and Cochrane recommendations. MEDLINE/PubMed, EMBASE, CINAHL, Cochrane Library, Physiotherapy Evidence Database (PEDro), SPORTDiscus and the Web of Science were searched from inception to March 2019 and updated in November 2020. Studies identifying electromyographic assessments for neuromuscular control during dynamic tasks in adult, physically active patients with an anterior cruciate ligament injury were eligible and qualitatively synthesized. Two independent reviewers used a modified Downs and Black checklist to assess risk of bias of included studies. Results From initially 1388 hits, 38 mainly cross-sectional, case-controlled studies were included for qualitative analysis. Most studies provided EMG outcomes of thigh muscles during jumping, running or squatting. Outcomes measures described neuromuscular control of the knee in domains of time, amplitude or activity. Risk of bias was medium to high due to an unclear description of participants and prior interventions, confounding factors and incompletely reported results. Conclusions Despite a wide range of EMG outcome measures for neuromuscular control, none was used to decide upon return to sports in these patients. Additional studies are needed to define readiness towards RTS by assessing neuromuscular control in adult ACL patients with EMG. Further research should aim at finding reliable and valid, EMG-related variables to be used as diagnostic tool for neuromuscular control. Moreover, future studies should aim at more homogenous groups including adequately matched healthy subjects, evaluate gender separately and use sport-specific tasks. Registration The protocol for this systematic review was indexed beforehand in the International Prospective Register of Systematic Reviews (PROSPERO) and registered as CRD42019122188. Supplementary Information The online version contains supplementary material available at 10.1186/s13102-021-00370-5.


Background
Anterior cruciate ligament (ACL) injuries happen quite frequently and concern athletes (0.15 injuries per 1000 athletic exposures (AEs)) but also the active part of the general population [1,2]. Most ACL injuries are due to a non-contact, multiplane mechanism [3] and may lead to instability, secondary meniscal injury or even knee osteoarthritis in the long run [4]. Consequently, this injury means several months or even years of physical impairment with wide consequences for the patients concerning return to work, return to activity or return to sport (RTS). RTS rates between 63 and 97% are reported for patients after ACLR [5,6]. Most elite athletes return to sports return earlier than non-elite athletes [5], on average within 12 months [6]. However, it remains unclear whether this approach is safe [6], omitting further injury, respectively. Athletes after ACLR returning to highdemanding sports (including jumping, pivoting and hard cutting) show a more than fourfold increase in reinjury rates over two years [7]. More than 5% of athletes with an ACLR sustain a re-rupture of the graft [6,8] in the ipsilateral knee after RTS. The risk for an ACL tear in the contralateral knee is as double as high (11.8%) even five years or longer after an ACLR [8]. Overall, the recurrence rates even after successful ACLR and subsequent rehabilitation are high (29.5% or 1.82/1000 AEs), with a tear of the ACL graft (9.0%), an ACL injury of the opposite leg (20.5%), muscle injuries on the ipsilateral side or even bilateral consequences [9,10].
It is known that ipsilateral deficits in clinical knee function and knee laxity persist even years after ACLR [11,12]. ACL patients show altered kinematics and kinetics [13] and different neuromuscular strategies during walking [14], not only in the injured limb but also in the non-affected side [13,15]. These changes are referred to neuromuscular adaptations due to altered sensorimotor control [16] and are caused by altered afferent inputs to the central nervous system due to the loss of the mechanoreceptors of the native (original) ACL [17]. Current literature regarding in ACL patients emphasises the importance of understanding consequences of ACL injury regarding neuromuscular control and kinematics [18][19][20]. To describe neuromuscular control in terms of simultaneously activated agonist/antagonist muscle pairs, generalized knee muscle co-contraction parameters are used [21,22].
In daily clinical practice, physical performance tests batteries including jumps and tests of muscle function [23] are often used to assess neuromuscular control for RTS. However, there is only limited evidence that passing RTS test batteries-interpreted as having achieved adequate levels of mobility, stability, strength, balance, and neuromuscular control for RTS-reduces the risk for a second ACL injury [24]. Moreover, it remains unclear which measures should be used to bring athletes safely back to RTS with a low risk of re-injury [25]. In conclusion, the currently suggested RTS criteria do not seem to be adequate to assess neuromuscular control of the knee joint to judge upon a safe RTS or even competition. Therefore, meaningful, reliable, valid and accurate diagnostic tools for patients with an ACL injury (either treated surgically or conservatively) are needed and may aid clinical decision-making towards a safe RTS following ACLR. Objective measurements of neuromuscular control should include electromyography (EMG) of involved muscles to judge upon quantity, quality and timing of voluntary activation and reflex activity [13,20,26]. However, up to date it is unclear which EMG-related measurements for neuromuscular control are used in patients with an ACL injury to decide upon a safe RTS.

Objectives
The first objective of this systematic review was therefore to summarize the scientific literature regarding EMGrelated assessments for neuromuscular control in adult, physically active patients with an ACL injury (either treated surgically or conservatively) during functional tasks. The second aim was to analyze whether these assessments for neuromuscular control were used to decide upon readiness for RTS in these patients.
The protocol for this systematic review was indexed beforehand in the International Prospective Register of Systematic Reviews (PROSPERO) and got the registration number CRD42019122188.

Eligibility criteria
To define the relevant key words for the literature search, the Participants-Intervention-Control-Outcome-Study design (PICOS) scheme was used as follows (Table 1).
Studies were considered eligible for this systematic review if they met the following inclusion criteria: Study participants-either females, males or both-had to be 18 years or older, suffer from an ACL injury-either treated conservatively or surgically-with a time since injury/surgery of six months at least, be athletes or physically active people who participate in sports activities on a regular basis (as defined by each study, e.g. Tegner Activity Score (TAS) ≥ 3) to get data to decide upon RTS. Moreover, included studies had to have used active or functional tasks such as walking, stair climbing or jumps, applied assessments for neuromuscular control of lower limb muscles using EMG, be original articles published in peer-reviewed, scientific journals in English, German, French, Italian or Dutch, and available as full texts. Exclusion criteria were model-driven approaches, animals or cadavers, comparisons of surgical techniques, passive or non-functional tasks (such as isokinetic measurements for strength and isometric muscle activity), editorials, conference abstracts, book chapters, theses, systematic reviews and meta-analyses.

Data sources
The search was effectuated from inception until March 2019 and updated in November 27th, 2020 in the electronic databases MEDLINE/PubMed, EMBASE, CINAHL, Cochrane Library, Physiotherapy Evidence Database (PEDro), SPORTDiscus and in the Web of Science. To ensure new articles matching the search terms, e-mail alerts were established from each of the databases if possible [29]. Furthermore, a hand search was done using the reference lists of included articles to identify additional and potentially eligible articles that had been missed in the electronic database searches. The hits from these two additional sources were also screened for eligibility applying the same criteria as for the articles from the database search.

Search strategy
In all sources, the advanced search mode was used if available. A search matrix combining relevant keywords (if possible MeSH-terms) with the Boolean operators AND and OR was used and customized for searches in all databases if necessary (see Additional file "Search string for MEDLINE/PubMed" 1): "anterior cruciate ligament/anterior cranial cruciate ligament/ACL"; "anterior cruciate ligament injuries/strains and sprains/rupture/ tear/injury/deficiency"; "anterior ligament reconstruction/anterior cruciate ligament/surgery/reconstructive surgical procedures/orthopedic procedure/ orthopedic procedure/tendon graft/tendon transfer/conservative treatment/non-surgical/rehabilitation/physical therapy modalities/physiotherapy/kinesiotherapy/ exercise/instruction/resistance training/neuromuscular training/postoperative care"; "neuromuscular control/ neuromuscular activity/sensorimotor control/muscle activity/active stability"; "electromyography/EMG/electromyogram/amplitude/timing/mean activity/peak activity/duration of activity/onset/offset/on-off-pattern/ pre-activity/latency/reflex response". In the updated search, articles were filtered by date of publication, with the aim of including only those published between March 2019 and November 2020.

Study selection
All hits obtained by the database searches were downloaded to the Rayyan reference management platform [30] and inserted into EndNote (Clarivate Analytics, Philadelphia, USA). Prior to screening, duplicates were  [14,20] Study design (S) Any laboratory or interventional study, cross-sectional or longitudinal such as randomized controlled trials, clinically controlled trials without randomization, laboratory/experimental controlled trials etc removed. Two authors (AB and IK) independently screened title and abstract of the records, one by using the software EndNote (Clarivate Analytics, Philadelphia, USA) and the other with the help of the free software "rayyan" [30]. After screening, full texts of relevant hits were read by the two authors (IK, AB) to decide upon inor exclusion. If their decisions did not match, discussion took place until consensus was achieved. If consensus would not have been reached, a third author (IB or HB) would have finally decided upon in-or exclusion of the record in question; however, this was not necessary.

Data collection process and data extraction
After final decision of all studies, data extraction for each eligible study was performed by the first author (AB) with a predefined Microsoft ® Excel (Microsoft Corporation, Redmond WA, USA) spreadsheet as piloted form. The first author (AB) extracted necessary information from each article describing the study design, groups measured and their characteristics, the tasks to be fulfilled by all participants, and all EMG-related assessments or methods used to evaluate neuromuscular control. Furthermore, the chosen assessment for neuromuscular control were judged whether they were used to clear the participants for RTS. The second author (IK) checked the extracted data at random. As all included studies provided enough information to be qualitatively analyzed, it was not necessary to contact corresponding authors for obtaining or confirming data.

Assessment of risk of bias in included studies
The risk of bias of all the included articles was independently assessed by two raters (AB, IK) by using the Downs and Black checklist [31] in a modified form [29,32]. The following categories were evaluated: (1) reporting bias: objectives/hypothesis, main outcomes, patients' characteristics, interventions, principal confounders, main findings, estimates of random variability, actual probability values; (2) external validity bias: study subjects/staff/places/facilities representative; (3) internal validity bias: blinding subjects/assessors, data dredging present, different lengths of follow-up/same time period between intervention and outcome for cases and controls, statistical tests/main outcome measures appropriate; (4) selection bias: patients and controls from same population and over same period of time, randomization, allocation concealed, adjustments for confounding, loss to follow-up; and (5) power analysis (see Additional file "Methodological quality assessment" 2). Each question of the categories was scored with 1 or 2 points if the criterium was fulfilled (answer "yes"), zero points if the answer was "no", "not fulfilled" and an "X" if the criterium was not applicable, e.g. randomization for a case-control or cross-sectional study, "IC" for intrasubject comparison, respectively. For this systematic review, studies with a total score of 17 or above out of 25 (more than 2/3 of the maximum total score) were considered as being of high methodological quality, showing a "low" risk of bias respectively [29]. Studies which reached 13 to 16 points (more than 50% of the maximum total score) were rated as being of "medium" quality, and total scores below 13 were rated as being of low methodological quality, "high" risk of bias respectively. As the aim of this systematic review was to summarize the applied measures for neuromuscular control, the methodological quality of the included studies was of secondary interest. Therefore, no study was excluded due to a low total score in the risk of bias assessment.

Study selection
Hits from the first and the updated database search including e-mail alerts and hand search were screened for duplicates. After applying in-and exclusion criteria according to PRISMA flowchart [27], a total of 38 articles involving 1236 subjects-809 participants with ACLR or ACL deficiency and 427 healthy controls-could be used for qualitative analysis. Reasons for exclusion were participants younger than 18 years, not able to achieve RTS, time since injury or surgery less than six months, static or non-functional task, study design (e.g. systematic review, study protocol), unclear or inadequate outcome, healthy participants or without ACL injury. Included studies had mainly a cross-sectional, case-controlled study design. Details about every step of the search are illustrated in the following flowchart ( Fig. 1).

Characteristics of included studies Study design
All included studies were case-control studies, except two which where case series [56] or a single-case study [69]. Two reported a retrospective or secondary data analysis [52,55] or provided a subgroup analysis from a larger trial [45,[47][48][49][65][66][67] (Table 2). Thirty-five studies compared the ACL participants with at least one control group (other ACL treatment, e.g. surgical versus conservative, or healthy controls), the remaining three studies made a comparison between the injured and the non-injured leg of the participants [42,44] or compared the pre-injury status with follow-up data from pre-and post-surgery [69].
Details regarding methodological aspects of all included studies are presented in Table 4 below.
The tasks used were very diverse: there were activities of daily life such as walking on even ground and downhill [33,47,53,57,58,62,63,68], and stair climbing [13,49]. Other activities went more towards sports such as running [44,45,60,67,68] and jumping [36, 37, 39-42, 48, 50-52, 54-56, 59, 61, 65, 66, 68] where mainly the singleleg hop for distance, drop jumps and countermovement jumps were used. Some authors chose typical rehabilitation exercises such as forward lunges [34], Nordic  Table 2 Risk of bias assessment with modified Downs and Black checklist [29,31,32] Authors                    Rudolph and Snyder-Mackler (2004) [49] Step up and over a 26 cm high step; 10 trials, 5 each with the right and left leg ascending a 26 cm step (higher than a typical step, provide a more challenging condition), EMG collected from landing limb LH, VL, SO, medial head of the gastrocnemius muscles of both limbs Peak EMG activity (%max); onset and termination of muscular activation; duration of muscular activity; cocontraction     hamstrings or hamstrings curls [35] and squats [64]. At the other end of the scale, more complex, highly demanding, sport-specific tasks such as an instep soccer kick [38] or a sidecutting maneuver [69] were reported. Only few research groups used perturbation platforms to simulate injury mechanisms during walking [53] or squatting [46,64], or applied devices to stress the ACL in the posterioranterior direction [50]. In addition, two studies even investigated the influence of fatigue on neuromuscular control [41,52].

Outcomes
All included studies used surface EMG as method to assess neuromuscular control and provided EMG-related variables such as peak and mean amplitudes, timing and peak of muscle activity, preparatory and reactive muscle activity, on-and offset of muscular activation, coactivation/co-contraction ratios, or asymmetry index. The outcome variables were expressed as percentage of maximum voluntary (isometric) contraction (%MVIC or %MVC) or reported in microvolts or milliseconds according to the variable chosen in amplitude or time domain.

Decision for Return to Sports (RTS)
None of the included studies used the surface EMG measurements to decide upon readiness for RTS (Table 4). However, the results from about a third of the studies (31.6%, 12 studies) could provide useful information by the choice of the assessed groups such as copers versus non-copers [33,34,[47][48][49]57], intervention and control group from the same team or level/league [35,38,40], data from pre-injury/pre-surgery including postsurgical follow up [62,69] or participants with full RTS versus limited RTS [53]. In addition, two studies even investigated the influence of fatigue on neuromuscular control [41,52]. More detailed information regarding EMG methods and procedures such as EMG type, detection, normalization, data processing and electrode placement can be found in Additional file "EMG methods and procedures of included studies" 3.

Discussion
The aim of this systematic review was to summarize the scientific literature regarding EMG-related assessments for neuromuscular control in patients with an ACL injury (either treated surgically or conservatively). The second aim was to analyze whether these assessments for neuromuscular control were used to decide upon readiness for RTS in these patients.
There were many factors present which could have an influence on neuromuscular control.

Influence by type of comparison (intraversus inter-subject)
The use of the contralateral, non-injured leg in intrasubject comparison, without a "real" control group [42,44] may lead to an overestimation of the physical performance in the ACL reconstructed or -injured leg. After ACLR, functional performance is often expressed with the LSI [70]. As the non-affected limb may also have deteriorated, the LSI may overestimate the right time for a safe RTS, and therefore, the risk for secondary injury may be higher [23]. In acutely injured ACL patients, intra-individual comparison showed bilateral consequences during stair ascent and indicates an alteration in the motor program (''pre-programmed activity'') [71]. In addition, in case of a case-controlled study design, the subjects in the control group should be matched to the ACL participants regarding sex, age, body mass, height, activity level and leg dominance.

Influence by level of activity and fatigue
Some of the included studies used very challenging, sports-specific task to assess neuromuscular control, some even assessed neuromuscular control after fatiguing tasks. It is known that most of ACL tears are noncontact injuries happening at the end of a training session or a play [72]. Therefore, the closer the task to the sports and injury-risky situation, the safer the decision towards full RTS or even return to competition will be. However, assessing performance-based tests or movement quality may be more difficult to standardize, require more complex equipment and large amounts of space. But if only impairments will be tested, there will be a lack of information regarding an "athlete's capacity to cope with the physical and mental demands of playing sport" [73]. It is therefore recommended to search for a standardized assessment close to the injury mechanism.

Influence by gender
Not all included studies reported findings of mixed groups separately by gender. Some did not even state whether study participants were male or female. This could partly be explained by the date of publication as gender difference in ACL patients has not been in the focus of former ACL research. It is known that female athletes are more likely to sustain an ACL injury than men [74,75]; the increased risk is probably multifactorial [76]. Several studies indicate that hormonal factors play a role [3,77] contributing to an increased laxity of ligaments in the first half of the menstrual cycle. However, biomechanical and neuromuscular aspects as indicators are discussed controversially in literature: Gender-specific neuromuscular adaptations and biomechanical landing techniques are considered being the most important ones to explain the increased risk of injury in women [78,79]. The higher risk for females to suffer from an ACL injury can be explained by motion and loading of the knee joint during performance [74]. Female athletes typically perform movements in sports with a greater knee valgus angle than men. Therefore, the amount of stress on the ACL in these situations is higher caused by a high activation of the quadriceps despite limited knee and hip flexion, greater hip adduction and a large knee adduction moment [80,81]. The dominance of the quadriceps muscle in women could contribute to increased anterior tibial translation [82,83] and was found in various activities such as jumps and cutting maneuvers [84][85][86]. Moreover, females typically land with an internally or externally rotated tibia [87], leading to an increased knee valgus stress due to greater and more laterally orientated ground reaction forces [83]. In contrast, other researchers did not find any gender-specific differences in the quadriceps-hamstrings ratio [88], not even in landing and cutting maneuvers [89]. A systematic review summarized biomechanical gender differences and stated that these were based on questionable clinical relevance [89]. In addition, strength-paired women and men showed no significant differences in neuromuscular activity [90].

Influence by treatment
The included studies reported different treatment options (ACLR with different graft types, conservative treatment). Depending on the classification of the participants in copers and non-copers, the results in neuromuscular control may differ from a population of ACLR participants. Therefore, all researchers who worked with copers and non-copers made intra-and inter-group comparisons without an ACLR group. A Cochrane review revealed low evidence for no difference in young, active adults after two and five years after the injury, assessed with patient-reported outcomes. However, many participants described as "non-copers" with unstable knee with conservative treatments remain symptomatic, and therefore, later opt for ACL surgery [91]. It has been described that persistent co-contraction and joint stiffening in these "non-copers" is likely to be due to an abnormal neuromuscular strategy failing to restore joint stability in these ACL deficient group [92]. Furthermore, the choice of graft would influence the neuromuscular control of measured muscles due to the morbidity of the harvesting site of the graft (e.g. hamstrings).

EMG variables
If the researchers mentioned the procedures for collecting EMG data, they referred to standardized applications and guidelines such as SENIAM (Surface Electromyography for the Non-Invasive Assessment of Muscles) [93]. The provided EMG-related variables were in accordance to the ones mentioned in a systematic review searching for knee muscle activity in ACL deficient patients and healthy controls during gait [14]. Current literature suggests greater co-contraction indices, increased joint stiffness and earlier muscle activation onset times as measures of neuromuscular function reflecting the incomplete restoration of normal joint stability [19,92]. Some of the included studies reported values of muscle onset activity in milliseconds and percentage of gait cycle as a systematic review did by summarizing and quantitatively analyzing muscle onset activity prior to landing in patients after ACL injury [20]. However, no cut-off values out of EMG-related variables were provided to determine an adequate level of neuromuscular control. Moreover, some of the researchers only provided integrated EMG values which would make it difficult to be compared to other studies using the respective units (milliseconds, millivolts) or widely used percentage values (%MVIC, %MVC).

Return to sports (RTS)
Regarding the determination of RTS after ACLR, there is some evidence for the use of functional performance tests, which had also been widely used in the included studies. Multiple functional performance measures-a battery including strength and hop tests, quality of movement and psychological tests [25]-might be more useful for the determination of RTS than a single performance measure. However, it is still unclear, which measures should be used to bring athletes safely back to RTS with a low risk of a second ACL injury [25]. Currently used RTS criteria or assessments, such as time, strength tests, hop tests, patient-reports, clinical examination, thigh circumference, ligamentous stability, range of motion, effusion and performance-based criteria, may be suboptimal at reducing the risk of a second ACL injury [73,94]. Recovery of neuromuscular function was mentioned to be important because of the existing connection between the variables time since surgery and the risk for re-injury of the knee joint; but adequate assessment procedures to assess neuromuscular function are still a matter of debate [7]. In contrast, authors of an included study stated that "studies like ours that focus on the objective measurement of the change of the muscle latency time over time may allow patients to return to full activity and to sports earlier than the standard time of 6-12 months" [42].
However, this statement only based on one outcome measure and contrasts with current criterion-and timebased recommendations for RTS. Therefore, this recommendation seems to be rather dangerous.

Limitations
The sample size of all the studies was quite low, however, providing reasonable sample size calculations and depending on the variable investigated, the results were acceptable. Furthermore, the more restrictive the inclusion criteria for the participants, the more homogeneous the intervention and the control groups were, but the more challenging the recruitment process was, leading to smaller groups to be investigated.
The used assessment for the risk of bias, the Downs and Black checklist [31] in a modified form [29,32] is designed for randomized and non-randomized controlled studies, however, the latter score lower in some items, get lower total scores and therefore a worse overall rating of the methodological quality. Despite this disadvantage, we decided to use the modified checklist as we could assess all studies with different designs included in this systematic review. However, the use of total scores and choice of cut-off values for low, medium and high risk of bias, respectively, were arbitrary and not based on literature.

Implications for clinical practice
This systematic review summarized assessments using EMG variables for neuromuscular control of the knee in patients suffering from an ACL injury (either treated surgically or conservatively). Despite 38 articles providing a wide range of EMG-related assessments, none was used to decide upon readiness towards a safe and successful RTS in patients after an ACL injury. So far, there is no diagnostic measure to assess neuromuscular control and therefore, clinicians should use a multimodal approach including assessments for active and passive knee stability under different sports-related conditions but be aware of not being able to evaluate neuromuscular control in depth without EMG-related assessments. Moreover, the widely used LSI may overestimate the physical performance of an ACL patient as the non-affected limb is likely to have deteriorated, too.

Implications for further research
Additional studies are needed to define readiness towards RTS by assessing neuromuscular control in adult ACL patients with EMG. Further research should aim at finding reliable and valid, EMG-related variables to be used as diagnostic tool for neuromuscular control.
Due to the heterogeneity in participants, interventions and outcomes used, future studies should aim at more homogenous patient groups, evaluate females and males separately, provide adequately matched healthy subjects (gender, height, weight, activity level etc.), control for confounding factors such as type of treatment, and use tasks close to the injury mechanism, as sport specific as possible, respectively. Moreover, it would be interesting to assess not only lower leg but pelvic and core muscles in addition. This would help to give insight in the complex field of ACL injuries and subsequent rehabilitation strategies, and therefore improve knowledge towards a safe RTS in these patients.