Muscle Co‐Activation Across Activities of Daily Living in Individuals With Knee Osteoarthritis

Muscle co‐activation has been shown to be elevated in individuals with knee osteoarthritis (OA) during gait. Comparisons of muscle co‐activation across different activities of daily living such as stair negotiation have yet to be explored. The aim of this study was to explore muscle co‐activation across different activities of daily living in patients with knee OA.

Muscle co-activation (simultaneous coordinated agonist and antagonist muscle activity) is thought to be a major mechanism for joint stabilization, load distribution, and movement control during gait in knee OA (1)(2)(3)(5)(6)(7)(11)(12)(13)(14)(15)(16)(17). Baratta et al (9) suggested that muscle co-activation is necessary to aid the ligaments in maintaining joint stability, distributing joint surface pressure, and regulating joint mechanical impedance. In young healthy individuals and individuals with knee OA, 2 muscle co-activation strategies have been identified. Overall muscle co-activation is considered as high co-activation across all muscle combinations surrounding the joint (18). Selective muscle co-activation involves high co-activation in specific, but not all, muscle combinations (e.g., agonist/antagonist [2,3,18] or mediolateral [3,19] combinations, but not both). In individuals with knee OA, high levels of muscle co-activation are thought to stabilize the knee in the absence of sufficient stabilization from the passive-restraints system (20). This strategy has been associated with increased joint contact pressures and may be a risk factor for cartilage degeneration and knee OA disease progression (1)(2)(3)5,6,(11)(12)(13)(14)20,21).
For other activities of daily living (ADL), very little evidence of muscle co-activation in individuals with knee OA exists. Two studies looking at stair negotiation found conflicting results. Childs et al (2) found high tibialis anterior/gastrocnemius (G) co-activation in individuals with knee OA, while Hortobágyi et al (14) found that there was no difference between individuals with knee OA and controls. When activities were grouped, individuals with knee OA had higher biceps femoris (BF)/vastus lateralis (VL) co-activation. Patsika et al (25) found higher BF muscle activity and no difference in the VL between individuals with knee OA and controls during sit-to-stand activity. Bouchouras et al (4) also found significantly higher BF/VL co-activation in individuals with knee OA during sit-to-stand activity compared to controls. In healthy individuals, during more challenging activities (i.e., stair negotiation) requiring higher muscle activation, muscle co-activation would be expected to be higher. In individuals with neuromuscular deficits such as those with knee OA, this expectation may not be true, which may have implications for rehabilitation (i.e., may limit the tasks that can be undertaken), and understanding muscle co-activation strategies across different ADL and across different muscle combinations is important.
Agonist/antagonist co-activation, especially hamstrings (H)/quadriceps (Q), is suggested to increase joint stiffness, where its primary role is to influence anterior tibial shear force and internal rotation (1,2,(26)(27)(28). The vasti muscles, however, have been suggested to be general joint stabilizers (26,27), so that mediolateral co-activation is thought to respond to joint space narrowing and instability, increasing joint stiffness and joint load (2,3,26,27). This possibility raises questions about co-activation in knee OA. Specifically, do the same patients consistently demonstrate the highest muscle co-activation across different activities and muscle groups (e.g., with a high positive correlation between agonist/antagonist and mediolateral muscle co-activation across all activities)? Alternatively, do different individuals exhibit high muscle co-activation during different activities or muscle combinations (e.g., high mediolateral and low agonist/antagonist muscle co-activation during stair negotiation, and low mediolateral and high agonist/antagonist muscle co-activation during gait).
The purpose of this study was to explore muscle coactivation patterns across different ADL tasks and investigate specific areas of muscle co-activation during different phases of gait. We hypothesized that for a specific activity, patients will demonstrate high muscle co-activity across all muscle combinations, that muscle co-activation will be higher in the medial/ lateral than agonist/antagonist muscle combinations in patients with knee OA, and that muscle co-activation will be higher during more challenging activities (e.g., stair descent) compared to less challenging activities (e.g., gait).

SUBJECTS AND METHODS
Subjects. Data analysis presented here is part of the Neuromuscular Control in Knee Osteoarthritis study. A convenience sample of adults (age ≥40 years), with physician-diagnosed unilateral/bilateral knee OA, with self-reported knee pain, and stiffness lasting <30 minutes, confirmed by ultrasound and/or magnetic resonance imaging (data not shown), were recruited through rheumatology clinics, general practitioner practices, and a local newspaper advertisement. Participants were excluded if they had any current neuromuscular skeletal injury or disease, knee replacement, knee surgery in the past year, steroid injections in the past 3 months, or severe comorbidity that would limit participation in the study. All participants gave written informed consent. The assessment protocol was approved by the West of Scotland Research Ethics Committee (13/WS/0146) and by Glasgow Caledonian University (HLS12/86) and carried out in compliance with the Declaration of Helsinki.

SIGNIFICANCE & INNOVATIONS
• The same patients demonstrated consistently high or low muscle co-activity across all muscle combinations. • Muscle co-activation was significantly different across activities, so that muscle co-activation was higher during more challenging activities, e.g., stair negotiation, than during less challenging activities, e.g., gait. • Neither overall nor selective muscle co-activation strategies were predominant; thus, both muscle co-activation strategies appear to modulate in unison to promote joint stability.

| 653
Electromyography (EMG) and muscle co-activation. Wireless surface electrodes (99% silver, four 5×1-mm bar Trigno sensors, fixed inter-electrode distance 10 mm, Delsys) were placed over the belly of the vastus medialis (VM), rectus femoris (RF), VL, semitendinous (ST), BF, medial G (MG), and lateral G (LG) muscles of the test leg (6,12,29). The test leg was defined as the most symptomatic knee based on self-report. The electrode placement was in accordance with Surface Electromyography for the Non-invasive Assessment of Muscles recommendations (30,31). The area was shaved, lightly abraded, and cleaned with alcohol. Isolated contractions were used to assess the quality of the EMG recordings. The raw signal was passed through a Trigno differential amplifier, input impedance 10,000 MΩ, common mode rejection ratio >80 dB, and gain of 1,000 with a bandwidth of 20-450 Hz. The EMG signal was recorded with a 16-bit analog-to-digital converter (PCI-DAS6402/16, Measurement Computing Corporation), at a sampling rate of 2,400 Hz. All EMG and force data were collected in Qualysis Track Manager, version 2.7-2.9 (Qualysis Motion Capture Systems) and processed in Spike2, version 2.7.10 (Cambridge Electronic Design).

Measures of ADL.
Participants performed a series of ADL tasks in the following order: stair ascent and stair descent, walking, and sit-to-walk transitions, during a single visit to the human performance laboratory at Glasgow Caledonian University. The number of trials performed for each activity as stated in the protocol was a pragmatic decision to enable high-quality data to be collected while safeguarding patients against high levels of fatigue.
Participants performed 3-stair ascent and descent trials using a 4-step instrumented staircase with a force plate (Kistler, 9286BA) embedded in the second step, aligned with a second Kistler force plate in the walkway. Participants ascended the stairs, turned, and descended, ensuring the test leg landed on both force plates (walkway and second step). A successful trial was defined as the entire foot landing within the boundaries of the force plate with no obvious signs of targeting the plate. The use of handrails was permitted if required, and step-over-step (alternate leg on each step) was preferred; however, when this method was not possible, then step-by-step (both legs on the same step with the test leg as lead leg) was permitted. Participants performed 7 successful walking trials at a self-selected walking speed. A successful trial was defined as above and within ± 10% of movement time (Brower timing system). A standard armchair (height 48 cm) was placed on the walkway next to the force plate. Participants sat with their back against the chair and the test leg on the force plate, and they were instructed to stand up, walk 3.6 meters before turning, and return to a seated position. The use of the chair arms was permitted if required. For the purpose of this analysis, the stance phase (onset of force to toe-off), from 3 sit-to-walk trials was used.
For all activities, the stance phase was analyzed, defined as initial contact (ground reaction force exceeded 20 N) to toe-off (ground reaction force fell below 20 N). During walking, the stance phase was also split into 4 sub-phases: loading (0-14.9% of stance), early-stance (15-39.9%), mid-stance (40-59.9%), and late-stance (60-100%), with an additional pre-stance phase (-150 msec to initial contact) (17). Stair ascent and descent were each split into 2 subphases, including walk-to-stair transition (stance on the floor force plate) and continuous (stance on the force plate embedded in the stairs).
Participants performed a series of maximal voluntary isometric contractions (MVICs), using an isometric dynamometer (Biodex 4 Pro). Participants were seated, with their knee and hip flexed at 50° and 90°, respectively. Following a series of warm-up contractions, participants performed 3 flexion/extension MVICs lasting 3 seconds with 30 seconds rest for the H and Q muscles. For the G assessment, participants were seated with their knee at full extension and foot in an anatomically neutral position. Following a series of warm-up contractions, participants performed a series of 3 plantar flexion MVICs lasting 3 seconds, with 30 seconds of rest. Data were analyzed over a 500 msec window: 250 msec on either side of the peak force for H and Q, and 250 msec on either side of the peak EMG amplitude for G. For measurements of symptom severity, participants completed the Knee Injury and Osteoarthritis Outcome Score (32) and self-reported the duration of their symptoms.
Data management. EMG data were filtered at 20-450 Hz using a Butterworth fourth-order zero-lag bandpass. The average root mean square amplitude (RMS amp ) was calculated for the stance phase, with subsequent subphases, defined above, and normalized to MVIC RMS amp (33)(34)(35). RMS amp was chosen because it is assumed to be more robust and directly linked to electrical power, having more physiologic significance over the linear envelope (33,36). MVICs were used for normalization over peak dynamic amplitude because MVICs provide an estimate of neuromuscular control and information about muscle activation, enabling individual variation that precludes direct comparison to be taken into account (33,34,36). In individuals with knee OA, normalization to MVIC has been used to understand neuromuscular control alterations (3,35,(37)(38)(39) and serves to provide a physiologic reference (40).
Muscle co-activation was calculated using RMS amp normalized to MVIC, and normalized RMS amp data were used to calculate muscle co-activation using the equation below, where lowerEMG i and higherEMG i are respectively the lowest and highest RMS amp at sample i. Division by 100 takes the average across the normalized interval (41). Muscle co-activation strategies were explored using the muscle groups Q (VL, RF, VM)/G (MG, LG), G/H (ST, BF), H/Q, and medial (VM, ST, MG)/lateral (VL, BF, LG). Strategies were also explored using muscle pairs (VL/VM, ST/BF, and MG/LG). For muscle groups involving multiple muscles, the mean RMS for the muscles involved was used. To explore agonist:antagonist versus mediolateral muscle co-activation, the muscle combinations H/Q and VL/VM were used.
Statistical analysis. Descriptive statistics including mean ± SDs and frequencies of the demographics were determined. Skewness, kurtosis, and box plots were obtained to examine the distribution and to identify outliers for all variables. Hierarchical sensitivity analysis was performed with all data, with extreme outliers (>3 × interquartile range [IQR]) removed, with all outliers (>1.5 × IQR) removed, with all outliers and device users removed (valid data), and with valid data with 1.5 × IQR outliers associated with low MVIC or pain during MVIC included. Device users were defined as individuals who used the stair handrails and/or a walking-aid while performing the ADL tasks. After extreme outliers were removed, some variables remained insignificant, while others became significantly different between individuals with knee OA and controls (data not shown). This finding did not change when further outliers were removed (42). The main analysis was run with only extreme (3 × IQR) outliers removed. Sensitivity analysis was performed with and without device users; there was no difference between device users and non-device users.
Repeated-measures analysis of variance (ANOVA) followed up with the Bonferroni post hoc test was performed to compare muscle co-activity within each activity. Pearson's correlations between muscle co-activation combinations within the same activity, and partial correlations controlling for muscle strength and age, assessed hypothesis 1 (muscle coactivation would be high across all muscle combinations within a given activity). Correlation strength was defined as r <0.1 (no association), r = 0.1-0.29 (weak), r = 0.3-0.49 (moderate), and r >0.49 (strong) (43). Hypothesis 2 (muscle co-activation will be higher in the mediolateral than agonist/antagonist pairs) was assessed with paired sample t-tests using VL/VM and H/Q combinations. The VL/VM co-activation provides a clear metric for mediolateral co-activation to provide neuromuscular control of the knee joint, because the vasti muscles were general joint stabilizers (26). Repeated-measures ANOVA (muscle co-activation by activity) followed up with the Bonferroni post hoc test addressed hypothesis 3 (muscle co-activation will be higher during more challenging activities). All statistical analyses were conducted using SPSS software, version 22.0, with alpha set at 0.05.

Gait.
A total of 77 individuals with knee OA were recruited from rheumatology clinics (n = 15), general practitioner practices (n = 4), and a local newspaper advertisement (n = 58) ( Table 1). Thirteen subjects (17%) had missing data for the stairs. During gait, VL/VM demonstrated higher muscle co-activation than ST/BF during pre-stance, loading, and early-stance, and MG/LG during loading. During mid-stance, late-stance, and overall-stance, MG/LG was higher than ST/BF and VL/VM. Mediolateral co-activation was higher than Q/G and G/H during pre-stance and loading; H/Q and G/H during early-stance, midstance, and overall-stance; H/Q, Q/G, and G/H during latestance (waveform data are shown in Supplementary  Within the same phase of walking, correlations between muscle co-activation combinations ranged from no association to strong positive associations (Figure 1    Muscle co-activation was significantly higher for VL/VM than H/Q for loading (P = 0.008), early-stance (P < 0.001), mid-stance (P < 0.001), late-stance (P < 0.001), and overall-stance (P < 0.001), with no difference for pre-stance (P = 0.319) (Figure 2). Stair negotiation. MG/LG co-activation was higher than VL/VM during stair ascent transition (SUT) and continuous stair descent (SDC), while MG/LG and VL/VM were similar and higher than ST/BF during continuous stair ascent (SUC) and descent transition (SDT). Mediolateral co-activation was higher than H/Q and H/G during SUT, SUC, and SDC; Q/G was higher during SUT and SDT. During SDC, Q/G was similar to H/G and mediolateral, and higher than H/Q.
Within the same phase of stair negotiation, correlations across muscle co-activation ranged from no association to strong positive associations (Figure 1  LG) of activity combinations. Pre-stance was significantly different from loading, early-stance, overall-stance, sit-to-walk, and stair negotiation across all muscle combinations except ST/BF. Pre-stance was significantly different from loading, mid-stance, and late-stance for ST/BF. Mid-stance and late-stance were different from loading, overall-stance, and sit-to-walk for all muscle combinations. Overall-stance was different from sit-to-walk (H/G) and SUC (all combinations except H/G and ST/BF), sit-to-walk was different from SUC (all combinations except ST/BF), and stair ascent and descent phases were also different from each other for all combinations except ST/BF.

DISCUSSION
The results indicate that muscle co-activation was positively correlated across different muscle combinations within the same activity. Mediolateral co-activation within Q was higher than anteroposterior co-activation across all activities in knee OA. Muscle co-activation was higher during more challenging activities (stair negotiation) than less challenging activities (gait).
Investigations into muscle co-activation in knee OA typically focus on walking. This study aimed to explore muscle co-activation across different ADL, during which different muscle co-activation strategies were observed. Overall muscle co-activation is deployed when the limb is preparing to accept and then accepts weight and starts to transition toward single limb support. Overall muscle coactivation appears to be a strategy adopted when the limb is least stable, in more vulnerable positions requiring all muscles to acti- Figure 2. Muscle co-activation for vastus lateralis:medalis (black bars) and hamstrings/quadriceps (dotted bars) across different activities for individuals with knee osteoarthritis. Significant differences between vastus lateralis:medalis and hamstrings/quadriceps: * = P < 0.05; ** = P < 0.01; † = P < 0.001. vate simultaneously to stabilize the joint. During transitions from single-to-double limb support and when increased muscle force is required to propel the body from a flexed position into extension (mid-stance and late-stance, sit-to-walk, and stair ascent), selective muscle co-activation was used. Specifically, high muscle coactivation in MG/LG and VL/VM, which are thought to act as joint stabilizers, contributes toward rotational moments or increases compressive loads to facilitate moment generation needed to direct ground reaction forces, and potentially increases medial joint stability (11,26,27,44,45). Our results demonstrated that neither overall nor selective muscle co-activation was predominant, with a combination of both strategies used. Mills et al (11), in a systematic review of 14 articles, highlighted the fact that during walking, specific muscle co-activation is believed to play a role in distributing loads, while Lloyd and Buchanan (18) found in their modeling study that specific muscle co-activation (H/Q) contrib- uted to muscular support in response to static valgus-varus loads. These results suggest that both muscle co-activation strategies are modulated throughout different phases of walking or other activities to increase joint stability, distribute joint loads, and support joint moments at the potential cost of increased compressive loads.
Within the same activity, the same patients demonstrated high or low muscle co-activity across all muscle combinations. With increasing age and the addition of joint space narrowing associated with knee OA, the passive restraints (e.g., ligaments) become increasingly lax (39,44). To prevent lateral joint opening and the transfer of load, medially higher antagonist muscle force is required (46). Higher antagonist muscle activation is thought to increase joint stiffness (46), but the ability to adopt movement strategies that remain normal is lost with muscle weakness (39). Alterations in muscle co-activation strategies may, therefore, try to accommodate this lack of joint stability. Individuals with selective high muscle co-activation may be at an increased risk of disease progression as a result of high joint loads combined with high joint pressures associated with high muscle co-activation.
VL/VM co-activation was higher than H/Q in individuals with knee OA across all activities except pre-stance. H/Q co-activation increases joint stiffness to counteract joint instability (2). H activation is thought to increase joint stiffness and reduce loads on the anterior cruciate ligament by reversing the shear force on the tibia, counterbalancing the main knee flexion moment, at the expense of increased patellofemoral and tibiofemoral load (28). VL/VM co-activation has been suggested to be a response to joint space narrowing, increased joint stiffness, and joint surface loading (2,3,19,37,47). Flaxman et al also identified the vasti muscles as general joint stabilizers bracing the knee (26,27). Increased joint contact pressures associated with high muscle co-activation may increase the risk for cartilage degeneration (1)(2)(3)6,(12)(13)(14)18,19,21). Hodges et al (48) found that increased duration of medial (vastus medalis/semimembranosus) co-activation was associated with medial cartilage loss in medial knee OA, while Zeni et al (12) found high medial co-activation controlled medial laxity and instability in medial knee OA. Lateral (VL/BF) co-activation was inversely related with medial cartilage loss in knee OA (48) and is thought to unload the medial compartment (3,6,15,17).
According to findings from Bae et al (49), tibiofemoral OA is either confined to the medial compartment or generalized over the medial and lateral compartments. Several studies in medial and generalized knee OA are in support of selective lateral activation (3,6,15,17), but others are not (1,44,45). These results appear to be consistent with medial and generalized knee OA across the literature. Three studies investigated muscle co-activation and included medial knee OA patients only, with mixed results. Rudolph et al (39) and Lewek et al (45) found higher medial activation, while Lewek et al (37) demonstrated high lateral muscle coactivation. Including both medial and generalized knee OA in this study may dilute any compartmental differences, if they exist, but further research is required to understand muscle co-activation differences between medial tibiofemoral and generalized disease.
Muscle co-activation across activities was significantly different. We hypothesized that muscle co-activation would be higher during more challenging activities such as stair negotiation compared to less challenging activities such as gait. Muscle coactivation was higher during stair negotiation than during overallstance and sit-to-walk, and overall-stance was higher than sit-towalk. This finding is potentially due to a combination of greater joint instability and muscle force required to perform more challenging activities, whereby knee joint stability is required to propel the body up each step or control the lowering of the body down each step. During pre-stance, the results demonstrated higher Q/G, and similar Q/H activity to that found by Schmitt and Rudolph (1), where Q/G, G/H, and MG/LG are low, while Q/G, mediolateral, VL/VM, and ST/BF appear to be increasing in preparation to accept load (1,3) and slow the acceleration of the joint. During loading, our results were higher compared to findings in the literature, and higher than pre-stance except for MG/LG, which is in keeping with the literature, showing a peak in Q activity (3,6). Additionally, high mediolateral co-activation during loading was found, which is similar to results found by Heiden et al (17).
During early-stance, all combinations were lower than loading, in line with the findings of Schmitt and Rudolph (1), while mediolateral remained higher than other combinations (17). During mid-and late-stance, there were no studies using the same equation MG/LG, which increased, peaking during late-stance. Muscle co-activation was higher during sit-to-walk across all combinations compared to gait, except for loading and overall-stance; stair ascent was higher than sit-to-walk and gait except for loading and overall stance. During SUC, muscle co-activation was higher than ascent transition for ST/BF and MG/LG. Muscle co-activation during stair descent was generally higher than gait and lower than continuous ascent and descent. During more biomechanically challenging activities requiring greater muscle activation, elevated co-activation is expected. This study showed that in patients with knee OA, muscle coactivation was higher during more challenging activities.
This study has a number of strengths and limitations. First, it is a relatively large convenience sample (n = 77) with a substantial sensitivity analysis performed prior to and during the statistical analysis. We did not screen or grade participants for radiographic disease severity, making comparisons with previous literature difficult. MVICs were performed for H and Q muscles, but reference contractions were performed for the G to prevent discomfort to the patient. During stair negotiation and sit-to-walk transition, participants were permitted to use the handrails, step-by-step stair negotiation style, and chair arm. While this approach showed muscle co-activation during normal daily living, it meant that movement was not standardized across the entire sample. Sensitivity analysis indicated that this approach did not affect the results presented here. Other studies that looked at muscle co-activation during stair | 659 negotiation did not allow the use of handrails. Muscle co-activation values were higher in the study participants compared to the values reported for individuals with knee OA in the literature (2,15,37,38). Why muscle co-activation values were so high compared to those in the literature is unclear, but possible explanations include varying disease severity and participant demographics. Another possibility is the way in which the EMG signal was processed, because the studies that used the same equation and normalization methods used linear envelope to process their data rather than RMS, while other studies used different co-activation equations, normalization methods, and different time epochs over which the data was analyzed. Alternatively, low muscle activation during MVIC as a result of not fully activating the musculature or particularly low muscle activation may elevate the normalized EMG.
In conclusion, muscle co-activation patterns appear to be high across all muscle combinations within the same activity. Higher muscle co-activation was observed during more challenging activities that require greater stability. While neither overall nor selective muscle co-activation was predominant, they appear to modulate in unison to maintain joint stability and respond to the demands on the joint. While high muscle co-activation appears to be a mechanism to maintain joint stability, it may also increase the susceptibility to cartilage damage and the risk of incidence and progression of knee OA.