FormalPara Key Points

Compared to controls, individuals with ACLR demonstrate greater hamstrings EMG amplitude (normalized to activity during MVIC) and hamstrings-to-quadriceps co-activation during gait and stair-related tasks.

Individuals with ACLR demonstrated longer electromechanical delay of the hamstrings during knee flexion and greater hamstrings-to-quadriceps co-activation during knee extension compared to controls.

Overall, these results align with known clinical impairments following ACLR with implications for re-injury risk and long-term joint health, thus warranting attention in rehabilitation.

1 Introduction

Anterior cruciate ligament reconstruction (ACLR) and rehabilitation is the gold standard treatment to restore knee joint stability and function following ACL injury in active individuals [1]. However, the high rates of incomplete return to competitive sport (55%) [2, 3], secondary ACL injury (23–35% within first two years) [4, 5], and long-term sequelae, such as post-traumatic osteoarthritis (33–51%) [6, 7], suggest multifactorial shortcomings of rehabilitation. As rehabilitation professionals, we now understand that long-term reductions in knee extensor moments (e.g. quadriceps avoidance) and neuromuscular activation during activities of daily living (e.g. gait, stair ambulation) and sport (e.g. jump landing, cutting) are contributing to increased compressive forces [8, 9] and joint degeneration over time [10,11,12]. Likewise, as anatomical ACL agonists and dynamic stabilizers against knee valgus and anterior tibial translation [13], impairments in hamstrings neuromuscular control may be predictive of graft rupture after ACLR [14].

Neuromuscular control is generally defined as unconscious muscular activity in the preservation of dynamic joint stability [15]. Surface electromyography (EMG) has been widely used to understand post-traumatic changes in neuromuscular function during functional tasks. Electromyographic signal directly reflects motor unit recruitment and firing characteristics from which we can infer the role of muscles in producing movement or maintaining joint stability [16, 17]. Using EMG recordings, neuromuscular function has often been operationalized in terms of amplitude, onset timing, electromechanical delay (EMD), hamstrings-to-quadriceps co-activation, and time to peak muscle activity (Table 1).

Table 1 Description of EMG Outcomes
Full size table

Following ACL injury, quadriceps neuromuscular dysfunction (e.g. muscle weakness, activation failure, a lower rate of torque development) is well described [18, 19], and contributes to self-reported disability [20], and long-term sequelae (e.g. post-traumatic osteoarthritis) [21]. Although not as widely reported, similar disruptions of the neuromuscular system may threaten the hamstrings’ capacity to attenuate and counteract anterior and rotational tibial shear forces during knee loading in activities of daily living and sport [13, 22]. Hamstrings neuromuscular function is a crucial component of functional movement (e.g. walking, jumping, cutting), and changes after ACL injury may contribute to risk factors for secondary injury and post-traumatic osteoarthritis [14, 22, 23]. In particular, hamstrings muscle facilitation (e.g. greater muscle activity and co-contraction) occurs despite relative hamstrings muscle weakness after ACL injury [24] and ACLR [25]. Although protective against subsequent injury (theoretically by reducing shear forces at the knee) [14], this neuromuscular behavior may be detrimental to joint health through reciprocal inhibition of quadriceps activation and increased compressive joint forces [8, 9]. Hamstrings EMG amplitude and onset timing are modifiable with strength [26] and neuromuscular training [27] in healthy individuals, making them a possible clinical target for interventions in such individuals following ACLR. Surprisingly, evidence-based treatment approaches to guide hamstrings recovery are sparse, as the scientific literature has historically focused on the quadriceps.

An accurate understanding of hamstrings neuromuscular function in individuals with ACLR is important to better target impairments to optimize muscular recovery in rehabilitation. Thus, we undertook a systematic review with meta-analysis to describe and quantify the neuromuscular function of the hamstring muscle complex after ACLR to elucidate global thigh neuromuscular consequence of injury. Specifically, this review sought to investigate EMG-derived metrics of hamstrings neuromuscular function compared to uninjured controls.

2 Methods

2.1 Registration

This systematic review was completed in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analysis (PRISMA) statement and was registered with PROSPERO prior to completion of the initial search (registration No: CRD42018110824, approval date: Oct 15th 2018). PROSPERO was searched to ensure no similar reviews were ongoing at time of registration.

2.2 Search Strategy

Studies were included if (a) the study population included adults following primary unilateral ACLR, (b) graft types included autograft (e.g. bone-patellar tendon-bone or hamstring tendon), and (c) any EMG derived outcomes of the hamstrings were published. Studies were excluded if (a) the study population included adolescent (under 18 years of age) or elderly (over 70 years or age) individuals, or (b) individuals who underwent ACL revision surgery. Studies that also included a comparison group of individuals who were ACL deficient were included, but those data were not considered.

We searched databases from inception to October 22nd, 2020. We searched the electronic bibliographical databases of PubMed, Web of Science, SPORTDiscus, CINAHL, and EBSCOhost (MEDLINE).

The search strategy included key terms relating to the population (e.g. anterior cruciate ligament injury OR anterior cruciate ligament reconstruction OR anterior cruciate ligament injuries) and describing variables of interest (e.g. electromechanical delay OR time to peak activity OR coefficient of variation OR muscle activity, etc.). Since there is no universal definition of neuromuscular control [15], this search strategy comprised a wide spectrum of EMG-derived outcomes to capture all potentially relevant studies. The search terms were adapted for database-specific filters and language as appropriate. An example search for PubMed is shown in Electronic Supplementary Material Appendix S1.

The search was restricted to English language studies. Unpublished research was not considered as it was deemed impractical to identify all unpublished work on EMG activity associated with hamstrings muscle activity in individuals with ACLR.

2.3 Study Selection

A single investigator (DAS) exported all studies identified by the search strategy to Endnote X9 (Clarivate Analytics, Jersey). Any duplicates were then deleted using the deduplication feature. The titles and abstracts of all publications were then screened by a single investigator (DAS) to remove irrelevant studies. Full text was acquired for all retained studies. Two independent reviewers (DAS and NRG) determined final eligibility and inclusion. Any discrepancies were resolved at a consensus meeting.

2.4 Quality Assessment

Two independent reviewers (DAS and NRG) assessed the quality of all the included studies using the modified Newcastle–Ottawa Scale (mNOS), which is provided in Electronic Supplementary Material Appendix S2. All criteria were operationally defined, discussed in detail, and approved by all investigators during a consensus meeting prior to qualitative assessment. Although not developed specifically for ACL research, the mNOS is increasingly recommended for the qualitative assessment of observational studies [28]. The mNOS was applied using pen and paper. Discrepancies were resolved at a consensus meeting. Lack of agreement was resolved by a third reviewer (GEN) by a majority vote.

The mNOS contains eight categories (total of 9 possible points) relating to methodological quality. A score of 0–3 points was considered a low quality (LQ) study, a score of 4–6 points was considered a moderate quality (MQ) study, and a score of 7–9 points was considered a high quality (HQ) study.

2.5 Outcome Measures

The outcome measures considered in this review are generally considered metrics of neuromuscular function, operationally defined here as EMG-derived measures of amplitude, onset timing, EMD, co-activation, and time to peak muscle activity. For a description and justification of each, see Table 1.

2.6 Data Extraction

The following information was extracted from each of the publications by the primary investigator (DAS):

  • • publication information.

  • • patient descriptors: sample size, sex, age, height, weight, source of graft, sport, level of participation.

  • • study methods: study design, muscles assessed with EMG, task(s) completed.

  • • filtering and processing of EMG data.

  • • outcome measures: EMG amplitude, onset timing, EMD, co-activation, and time to peak muscle activity.

In individuals with ACLR, only data from the involved ACLR limb were extracted. Likewise, in control individuals, only data from the matched control limb were extracted. Corresponding authors were contacted for original data where publications did not report these in the text. Engauge Digitizer software (Open Source, Version 11.2) [29] was used to extract data from figures when authors were unable to recover the original data files or did not respond to requests [9, 30,31,32,33]. In five cases [34,35,36,37,38], data were not able to be extracted or obtained for meta-analysis.

2.7 Statistical Methods

Statistical analysis was completed using Review Manager 5 (The Cochrane Collaboration, Copenhagen, Denmark). Analyses were completed by one investigator (DAS). Standardized mean difference (SMD) and 95% confidence interval (CI) are reported. Individual and pooled SMD were calculated using Hedges’ g and categorized as small (≤ 0.50), medium (0.51–0.79), or large (≥ 0.80) [39]. To maximize clinical interpretability, raw mean differences (RMD, with 95% CIs) are reported for significant effects.

For studies reporting results for medial and lateral hamstrings musculature separately, data were analyzed separately. Semitendinosus and semimembranosus were reclassified as medial hamstring (MH), and biceps femoris was reclassified as lateral hamstring (LH). Studies not specifying or reporting MH and LH as pooled were reclassified as unspecified (US). For studies reporting results for males and females separately, or hamstrings tendon (HT) or patellar tendon (PT) graft types separately, data were pooled and have been presented as a heterogeneous cohort. The data for healthy control groups were also extracted in this way. Where methods, outcome measure, and task were comparable between studies, a random-effects meta-analysis was performed and the level of statistical heterogeneity for pooled data was established using the I2 statistics (p < 0.05). Heterogeneity (I2) was defined as low (0–40%), moderate (30–60%), substantial (50–90%), and considerable (75–100%) [40].

3 Results

3.1 Search Results

For detail of search results, including deduplication, exclusion, and full-text review, see Fig. 1. Following the screening of titles and abstracts, 163 publications were retained, and full text reviewed. Thirty-four studies were included for final review. Twenty-five studies evaluated the MH, 26 evaluated the LH, and 22 evaluated both MH and LH. One study [33] did not specify which hamstrings muscles were assessed, and three studies [31, 38, 41] reported bilateral hamstrings as pooled results. In these cases, they were classified as unspecified in the meta-analysis [31, 42].

Fig. 1
figure 1

Flow diagram summarizing study selection for inclusion. ACL anterior cruciate ligament, ACLR ACL reconstruction

Full size image

Study characteristics, such as population sources, sample size, and demographic descriptors, are shown in Table 2. In total, the data from 1299 individuals were considered (700 ACLR, 599 controls) with at least 37% of the total sample being female. Of the individuals with ACLR, 35.9% (n = 251) underwent autograft with PT, 48.6% (n = 340) autograft with HT, 0.4% (n = 3) autograft with quadriceps tendon, 1.7% (n = 12) allograft, and 13.4% (n = 94) were not defined. Study design, task, muscles, EMG variables, and normalization procedures are reported in Table 3.

Table 2 Study details included sample size, participant demographics, and population sources
Full size table
Table 3 Study design, tasks, muscles, and outcome measures evaluated
Full size table

3.2 Methodological Quality Assessment

Results from the modified Newcastle Ottawa Scale (mNOS) are shown in Table 4. The median score was 5.5 (range 2–8) out of 9. Of the 34 studies, 8 (23.5%) were rated as high quality [14, 30, 31, 42,43,44,45,46], 22 (64.7%) were rated as moderate quality [9, 32,33,34, 37, 38, 41, 47,48,49,50,51,52,53,54,55,56,57,58,59,60,61], and 4 (11.8%) were rated as low quality [35, 36, 62, 63].

Table 4 Modified Newcastle–Ottawa Scale scores
Full size table

3.3 EMG Amplitude

Twenty studies evaluated hamstrings muscle activation, 17 of which were meta-analyzed.

Fourteen studies normalized muscle activity as percent maximum voluntary isometric contraction (% MVIC) for 8 different tasks (Fig. 2). In this comparison, individuals with ACLR demonstrated no difference in strength-based activation (i.e. knee flexion), single-limb stance, in-step kicking, or lateral push on a slide board compared to controls. However, individuals with ACLR demonstrated higher EMG amplitude than controls in the more dynamic tasks of gait (1 study, LH: g = 1.13 [95% CI 0.73, 1.54], RMD = 7.17% [95% CI 0.24, 14.11]), stair ambulation (3 studies, MH: g = 0.53 [95% CI 0.31, 0.75]; RMD = 1.87% [95% CI 0.93, 2.80]; LH: g = 0.86 [95% CI 0.29, 1.43], RMD = 3.01% [95% CI 0.96, 5.07]), and double-limb (1 study, US: g = 1.96 [95% CI 0.30, 0.3.61], RMD = 7.63% [95% CI 1.15, 14.11]) and single-limb jumping and jump landings (2 studies, MH: g = 0.86 [95% CI 0.43, 1.30], RMD = 10.76% [− 0.05, 21.58]). No consistent pattern of difference was seen between MH and LH activation, as all 95% confidence intervals overlapped.

Fig. 2
figure 2

EMG amplitude normalized to maximum voluntary isometric contraction by medial hamstrings, lateral hamstrings, and task. (1) Nordic hamstring exercise, (2) hamstring curl exercise, (3) concentric phase of squat, (4) eccentric phase of squat, (5) propulsion, (6) heel-strike, (7) stair ascent, (8) stair descent, (9) single-limb anterior reach, (10) step down, (11) step down with unstable platform perturbation, (12) step down with dual-task challenge, (13) step down with dual-task challenge and unstable platform perturbation, (14) concentric phase, (15) landing phase, (16) lift-off, (17) reactive, following contact, and (18) preparatory, prior to contact. ACLR anterior cruciate ligament reconstruction, EMG electromyography, SD standard deviation, Std standard, CI confidence interval

Full size image

Two studies in the comparison were of high methodological quality. One study [30] reported higher MH (g = 1.18 [95% CI 0.46, 1.89]) but not LH EMG amplitude (g = 0.05 [95% CI − 0.60, 0.71]) in individuals with ACLR compared to controls during a single limb crossover hop test. Using pooled hamstring activity, another study [42] reported higher EMG amplitude in individuals with ACLR compared to controls during double limb jumping at ascent (g = 2.99 [95% CI 1.70, 4.27]), take-off (g = 0.48 [95% CI − 0.37, 1.33]), and landing (g = 2.56 [95% CI 1.38, 3.74]).

Three studies normalized activity as percent peak activity during the dynamic trial (% peak) for double limb or single limb jump landing tasks (Fig. 3). All were of high methodological quality. In the double limb comparison, there was no difference between hamstrings EMG amplitude between groups during double-limb jump landing (1 study, US: g = − 0.34 [95% CI − 1.06, 0.38], RMD = − 3.19% [95% CI − 9.53, 3.15]). Similarly, the LH and US single-limb jump landing comparisons showed no effect between groups (3 studies, LH: g = − 0.72 [95% CI − 1.51, 0.07], RMD = − 10.48% [95% CI − 18.38, − 2.58]; US: g = 0.10 [95% CI − 0.62, 0.81], RMD = 0.81% [95% CI − 5.03, 6.65]), as 95% confidence intervals crossed zero. However, a single study reporting MH activity during single limb jump landing found individuals with ACLR to be lower compared to control (MH: g = − 0.96 [95% CI − 1.83, − 0.07], RMD = − 14.00% [95% CI − 25.73, − 2.27]).

Fig. 3
figure 3

EMG amplitude normalized to maximum muscle activity in a dynamic trial by medial hamstring, lateral hamstring, and task. (1) reactive, following contact, (2) preparatory, prior to contact. ACLR anterior cruciate ligament reconstruction, EMG electromyography, SD standard deviation, Std standard, CI confidence interval

Full size image

3.4 Co-activation

Eleven studies evaluated hamstrings-to-quadriceps muscle co-activation indices (peak or average EMG amplitude) (Fig. 4) [9, 31, 32, 38, 41, 44, 52, 55, 58, 60, 61]. Individuals with ACLR demonstrated no difference in co-activation during in-step kicking, or double-limb and single-limb jump landings compared to controls. However, individuals with ACLR demonstrated higher hamstrings co-activation than controls during strength-based activation (i.e. knee extension, 3 studies, MH: g = 0.85 [95% CI 0.53, 1.18], RMD = 1.03% [95% CI 0.29, 1.76]; LH: g = 0.93 [95% CI 0.22, 1.64], RMD = 1.35% [95% CI 0.38, 2.31]; US: g = 1.34 [95% CI 0.65, 2.03], RMD = 12.90% [95% CI 7.06, 18.74]), gait (1 study, MH: g = 3.57 [95% CI 3.04, 4.09], RMD = 10.72% [95% CI 9.45, 11.98]; LH: g = 1.45 [95% CI 0.90, 2.01], RMD = 3.84% [95% CI 2.60, 5.08]), and stair ambulation (2 studies, MH: g = 0.43 [95% CI 0.16, 0.69], RMD = 2.75% [95% CI 0.97, 4.53]; LH: g = 1.04 [95% CI 0.63, 1.44], RMD = 5.90% [95% CI 2.69, 9.11]). During gait, MH co-activation was characterized by a larger effect than LH in all phases (terminal swing, load-acceptance, and heel-strike). However, there were no further patterns of difference between MH and LH co-activation during strength-based or stair tasks, as all 95% confidence intervals overlapped.

Fig. 4
figure 4

Hamstrings to quadriceps co-activation by medial hamstrings, lateral hamstrings, and task. (1) load-acceptance, (2) heel-strike, (3) preparatory, prior to contact, (4) descent, (5) ascent, (6) step down, (7) step down with dual-task challenge, (8) step down with unstable platform perturbation, (9) step down with dual-task challenge and unstable platform perturbation, (10) reactive, following contact, (11) single-limb landing, and (12) double-limb landing. ACLR anterior cruciate ligament reconstruction, SD standard deviation, Std standard, CI confidence interval

Full size image

3.5 Onset Timing

Four studies [35, 43, 56, 59] evaluated hamstrings EMG onset timing, 3 of which were meta-analyzed (Fig. 5) [43, 56, 59]. In this comparison, there were no differences between groups in either preparatory (i.e. feedforward, 1 study, MH: g = 0.16 [95% CI − 0.41, 0.72], LH: g = 0.11[95% CI − 0.46, 0.67]) or reactive (i.e. destabilizing platform, 2 studies, MH: g = 0.41[95% CI − 1.29, 2.10], LH: g = − 0.40 [95% CI − 0.86, 0.06]) hamstrings activation onset timing.

Fig. 5
figure 5

EMG onset timing by medial hamstrings and lateral hamstrings. ACLR anterior cruciate ligament reconstruction, EMG electromyography, SD standard deviation, Std standard, CI confidence interval

Full size image

3.6 Electromechanical Delay

Two studies evaluated hamstrings EMD (Fig. 6) [47, 51]. Overall, individuals with ACLR demonstrated longer EMD than controls (MH: g = 1.78 [95% CI 0.81, 2.75], RMD = 27.31 ms [95% CI 19.85, 34.77]; LH: g = 0.85 [95% CI 0.29, 1.41], RMD = 20.50 ms [95% CI − 10.47, 51.48]).

Fig. 6
figure 6

Electromechanical delay by medial hamstrings and lateral hamstrings. (1) isokinetic and (2) isometric. ACLR anterior cruciate ligament reconstruction, SD standard deviation, Std standard, CI confidence interval

Full size image

3.7 Time to Peak Muscle Activity

Three studies evaluated hamstrings EMG time to peak muscle activity during a jump landing task (Fig. 7) [31, 43, 49]. Overall, there was no difference in hamstrings time to peak muscle activity between groups (MH: g = 0.06 [95% CI − 0.76, 0.88], LH: g = 0.41 [95% CI − 0.21, 1.03], US: g = − 0.06 [95% CI − 0.82, 0.70]).

Fig. 7
figure 7

Time to peak muscle activity by medial hamstrings and lateral hamstrings. (1) single-limb landing and (2) double-limb landing. ACLR anterior cruciate ligament reconstruction, SD standard deviation, Std standard, CI confidence interval

Full size image

3.8 Heterogeneity of Studies

Overall comparisons for each outcome of interest demonstrated moderate (amplitude [% peak], EMD, time to peak activity) or substantial (amplitude [% MVIC], onset timing) heterogeneity upon meta-analysis. Additionally, there was high variability in tasks within the EMG amplitude, onset timing, co-activation, and time to peak muscle activity studies. As a result, emphasis on interpretation is placed on task comparisons (sub-comparisons) in each outcome.

4 Discussion

This systematic review and meta-analysis presents a wide spectrum of EMG-derived hamstrings neuromuscular impairments in individuals with ACLR compared to uninjured controls. Those with ACLR demonstrated higher (moderate to large effect) hamstrings EMG amplitude (% MVIC) and co-activation during gait and stair ambulation compared to controls. Interestingly, MH EMG amplitude during single limb jump landing activities differs based on the normalization procedure (higher in ACLR group normalized to % MVIC, and lower in studies normalized to % peak activity). Additionally, there is a large prolongation of EMD during knee flexion and a large increase in hamstrings co-activation during knee extension in individuals with ACLR. Although MH co-activation demonstrated a larger magnitude effect than LH during gait, there were no other significant differences between muscles for any outcome or task. Most meta-analyzed studies comparing EMG amplitude between individuals with ACLR and controls varied in the task assessed, owing to the heterogeneity of results. Overall, the included studies varied in methodological quality, muscles evaluated, and task, resulting in some outcomes of interest having fewer included studies. The results and the clinical implications of the findings, including percent activation differences, are discussed by category.

4.1 EMG Amplitude

EMG amplitude results varied by normalization technique, as well as task subgrouping (Table 3). Most notably, with double-limb and single-limb jumping tasks, the differences in normalization technique (whether to % MVIC or % peak) resulted in contrasting effects. In these comparisons, it appears hamstrings activity is higher (as % MVIC) during double-limb landing in an individual with ACLR yet is no different when normalized to the peak of the dynamic trial. Further, MH activity is higher when normalized to MVIC but lower during single-limb landing in individuals with ACLR compared to the rest of the dynamic trial (i.e. other phases of the jump). Although speculative, this dichotomy conveys that EMG amplitude must be interpreted with respect to the normalization procedure. For example, a weaker individual (with lower MVIC or volitional activation) may use a greater proportion of their maximal muscle activation than a stronger person to complete a standardized task. Thus, the weaker individual would exhibit higher percent activation during the task when normalized to MVIC compared to relative activation. Impairments in hamstrings strength are prevalent in individuals following ACLR [64] so differences in normalization may help explain the difference between groups. As normalization to MVIC was most common in these results and is reported to be more reliable than the alternative [65], differences should be considered while interpreting the remaining amplitude results.

The LH of individuals with ACLR demonstrated higher EMG amplitude during gait (large effect, g = 1.13, RMD = 7.17%) and stair ambulation (large effect, g = 0.86, RMD = 3.01%). The MH demonstrated only a moderate effect during stair-related tasks (moderate effect, g = 0.53, RMD = 1.87%) although the confidence intervals did overlap, indicating the difference between hamstring muscles is inconclusive. There is a pattern of greater hamstrings activity after ACLR as dynamic joint stability demands increase (e.g. gait, stairs, and jump landing), but not during static tasks (e.g. isolated knee flexion, single limb stance, kicking). Greater amplitude during tasks that demand greater dynamic stability may indicate demands for a greater proportion of hamstrings neural drive (as % of theoretical maximum during MVIC) or an adaptive upregulation of hamstrings activity in an effort to preserve dynamic joint stability following ACLR [16, 17, 22]. Considering the nuances of the normalization technique (e.g., whether differences in raw amplitude [1.87–10.76%] between groups represent the same activity), intra-limb normalization via co-activation may be more clinically interpretable.

4.2 Co-activation

Individuals with ACLR demonstrated greater co-activation during isometric knee extension, gait, and stair ambulation. Specifically, voluntary quadriceps activation during knee extension was characterized by higher hamstrings co-activation in MH, LH, and US comparisons (large effects, g = 0.85–1.34, RMD = 1.03–12.90%), indicating that hamstrings co-contraction may dampen the mechanical efficiency of the quadriceps in those with ACLR. The effect was also presented during gait, with the MH (g = 3.57, RMD = 10.72%) demonstrating higher co-activation than the LH (g = 1.45, RMD = 3.84%), and stair ambulation (moderate to large effects, g = 0.43–1.04, RMD = 2.75–5.90%). Hamstring-to-quadriceps co-activation is important to provide stability to the knee and reduce the amount of tensile force placed on the ACL or graft tissue. Through simulated work in cadaveric knees [66], we can appreciate that greater hamstrings co-activation is associated with decreased strain in the ACL. Therefore, these results may indicate that greater levels of co-activation occur as an adaptive strategy to better resist anterior tibial shear and rotation during functional tasks to maintain dynamic knee stability. Recent work [14] provides evidence to this effect, reporting that individuals with greater co-activation were less likely to suffer graft rupture. Other authors [67] found similar associations between intralimb muscle strength (hamstring-to-quadriceps ratio) and risk of graft rupture, which collectively implicate intralimb muscle function and imbalance in secondary injury prevention.

Interestingly, co-activation was 2.75–10.72% higher during activities of daily living (e.g. walking and stairs), but not more dynamic athletic tasks (e.g. double-limb or single-limb jumping), suggesting those with ACLR may fail to effectively carry-over this compensation to a sport where it may be desirable for improved knee stability. Conversely, greater co-activation results in increased compressive forces about the knee and lesser knee flexion–extension excursion during gait [9, 13], representing a negative consequence of this strategy during highly repetitive activities of daily living as these impairments have been linked to cartilage degeneration and incidence of post-traumatic knee osteoarthritis [9, 68]. Furthermore, greater MH co-activation has been associated with greater medial tibiofemoral joint loading, which may contribute to reports of higher incidence of medial compartment post-traumatic osteoarthritis [22, 69].

4.3 Onset Timing

Hamstrings muscle onset timing was not different between groups regardless of reactive (e.g. reflexive) or preparatory (e.g. planned) paradigm [43, 56]. As a dynamic stabilizer, the time from perturbation to the onset of hamstrings muscle activity is important for active stabilization against anterior tibial shear and rotation [22, 59]. It is theorized that altered joint afference following ACL injury contributes to the delayed reaction of the lower extremity musculature [70, 71]. Here, large delays in reactivity were seen with a larger degree of platform perturbation (30° inversion + 10° plantarflexion [56] vs. 20° degree inversion [59]) despite both samples being similarly removed from ACLR surgery (approximately 9 months). The absence of an overall effect may suggest that reactive muscle activation is a modifiable neuromuscular impairment in individuals with ACLR. However, specific interventions to achieve this remain unclear. Considering the complexity of both reactive and preparatory aspects of the sport, the functional implications of this nuanced relationship are hard to define and should be the subject of future research.

4.4 Electromechanical Delay

Both included studies demonstrated longer EMD in individuals with ACLR compared to controls, with larger delay in the MH (g = 1.78, RMD = 27.31 ms) compared to LH (g = 0.85, RMD = 20.50 ms). In vivo evidence suggests that peak ACL strain occurs in the first 100 ms of joint loading during dynamic tasks (e.g. cutting, landing) [72], suggesting the need to rapidly develop muscle activity and subsequent force production to protect the ACL. Increased time between the onset of EMG activity and force production is associated with a lower rate of force production and a delay in muscular stabilization about a joint (e.g. dynamic joint stabilization). Considering both studies utilized isolated knee flexion exercise to assess this outcome, it is not clear whether these deficits translate to more dynamic activities relevant to ACL re-injury (e.g. jump landing). Despite this, longer EMD might be implicated in high graft failure rates seen in individuals with ACLR. This is especially true considering the largest effect in the MH, which helps resist external tibial rotation and knee valgus associated with ACL strain [22]. There is a need to identify clinical interventions which may target this impairment. Additionally, future studies utilizing finite element modeling with EMG may help to derive this outcome during functional tasks.

4.5 Time to Peak Muscle Activity

These findings indicated no difference in hamstrings time to peak muscle activity between groups. Clinically, this may represent a normal time to peak muscle activity during jump landing maneuvers in individuals with ACLR. However, time to peak activity is nuanced by potential differences in EMG amplitude between groups. For example, Ortiz et al. [31] reported similar hamstrings amplitude between groups during single limb jump landing, but those with ACLR demonstrated significantly faster time to peak hamstrings activity, which suggests more rapid progression of activation (i.e. higher relative activation in same elapsed time). As discussed previously, with peak ACL strain occurring in the first 100 ms of knee joint loading [72], this neural facilitation strategy may be indicated to protect the joint. Future research should normalize time to peak muscle activity to rate of activation to better appreciate the effectiveness of muscle activation strategies that preserve time to peak muscle activity.

4.6 Methodological Quality, Limitations, and Directions for Future Research

There were many methodological limitations identified using the mNOS, with only eight of the included studies having adequately described the selection of controls (mNOS item 3). Most notably, this could result in inappropriate matching of participants across groups, which may have had an influence on the results. This limitation should be addressed in future research. Further, five studies could not be included in the meta-analysis; however, their reported results did not differ from the noted findings.

All six overall comparisons within this meta-analysis were characterized by moderate to substantial heterogeneity, which indicates a large degree of variability in the results. This is likely a representation of the wide range of functional tasks and inclusion of both MH and LH results represented in the collective sample. We have attempted to address this limitation by reporting the more conservative random effects for all comparisons. We have included subgrouping by hamstring musculature and functional task in an effort to demonstrate the drivers of this heterogeneity for each outcome of interest. However, this reduces the number of studies in each sub comparison. Van Tulder et al. [73] recommends incorporating methodological quality, I2 statistic, and number of studies to establish the level of evidence. In so doing, sub comparisons of EMG amplitude and co-activation during stair ambulation and double and single limb jumping, as well as EMG onset timing during preparatory reactions, can be classified as moderate evidence (i.e., pooled results including at least one of high quality or multiple homogenous studies of moderate or low quality). EMD and time to peak activity results can also be classified as moderate evidence. Conversely, all other sub comparisons are classified as limited (i.e., results from multiple heterogeneous moderate or low-quality studies) or very limited evidence (i.e., results from one moderate or low-quality study), and as such, caution should be used in interpreting these comparisons.

Demographic factors such as sex, age, and graft-type are reported to influence recovery with ACLR [74]. However, the influence of these factors on hamstrings neuromuscular function is not fully understood. To this end, the included sample is largely male (approximately 63%), which is not representative of the reported prevalence between sexes and threatens the generalizability of these findings. Regarding graft type, we set out to primarily include individuals who underwent ACLR with autograft in this review. However, five studies [9, 38, 41, 42, 53] included several individuals who received an allograft. Due to the small number (n = 12, 1.7% of total ACLR sample) with reference to the included sample, we felt this was an acceptable exception; exclusion of these data would have also resulted in the removal of a large number of individuals with autograft (n = 111, 15.8% of total ACLR sample). Additionally, individuals with HT autograft are known to have larger deficits in hamstrings muscle strength after ACLR compared to those with PT autograft [75]. Here, neuromuscular impairments were observed despite a heterogeneous sample (35.9% PT, 48.6% HT). Sex and graft-type subgroup analyses were not undertaken as part of the review and remain an area of further research.

Although outside the scope of this report, neuromuscular abnormalities may be present prior to and thus contribute to initial ACL injury. As all but two studies [14, 35] identified in this search were cross-sectional, longitudinal studies describing the neuromuscular impact of initial ACL injury and evaluating the role of clinical interventions which address these hamstrings muscle dysfunctions are still needed to further our understanding in this regard. On this note, the primary findings of Palmieri-Smith et al. [14] suggest that higher hamstrings activity during single-limb jump landing may be protective against subsequent injury in those with ACLR. Secondly, due to a lack of data in the literature, comparison to the contralateral limb was not included in this review. Inter-limb comparisons within individuals would be useful to determine if hamstrings neuromuscular function is impaired bilaterally with ACLR. Lastly, future work should determine the interaction between co-activation (intralimb ratio of EMG amplitude) and hamstrings-to-quadriceps ratio (intralimb ratio of muscle strength) to clarify the clinical implications of intralimb function.

5 Conclusion

Moderate quality evidence suggests that individuals with ACLR demonstrate higher EMG amplitude (1.87–10.76%) and hamstrings-to-quadriceps co-activation (2.75–10.72%) during gait and stair-related tasks compared to controls. Additionally, hamstrings co-activation was higher during knee extension (1.03–12.90%), suggesting less efficient isolated quadriceps activation. Moderate quality evidence also suggests that individuals with ACLR demonstrate longer EMD (20.50–27.31 ms) of the hamstrings musculature with a greater negative impact in the MH than the LH. Collectively, these neuromuscular adaptations align with known clinical impairments (e.g. decreased quadriceps and hamstrings muscle strength) and are theorized to contribute to poor outcomes seen in this population, such as re-injury and post-traumatic osteoarthritis. These impairments warrant attention in rehabilitation from ACLR.