Participants for this study were recruited from the Canadian Armed Forces. Asymptomatic participants volunteered by responding to on base posters and base-wide recruitment e-mails. Recovered LBI (rLBI) participants were identified if they reported a back related issue resulting in altered activities of daily living for at least 3 days (Ozguler et al., 2000) to the Canadian Forces Health Services Center. Potential participants were contacted to see if they were interested in participating in the study. For both groups a self-report questionnaire screened for the following exclusion criteria: previous abdominal or back surgery, cardiovascular, respiratory or neurological conditions that place them at risk for participating in the study. For participant in the rLBI group questions determined that their recent LBI occurred 4–12 weeks prior to their data collection, was not chronic lasting longer than 12 weeks, and not recurrent where a previous injury occurred within 12 weeks prior to their most recent episode (Delitto et al., 2012). Asymptomatic participants were screened to determine that they had not experienced an activity limiting LBP (Ozguler et al., 2000) in the last year. All rLBI were deemed recovered at the time of testing reporting minimal pain [Visual Analog Scale (VAS) <30/100 mm (Boonstra et al., 2014)], minimal disability [Roland Morris Disability (RMD) <9/24 (Roland and Morris, 1983; Stratford et al., 1998)], and resuming normal activities of daily living at the time of testing. Before testing, all participants gave written informed consent consistent with the Declaration of Helsinki, to the study protocol approved by the Institutes Research Ethics Board.

Participants attended two sessions. During session one all participants were screened by a registered physiotherapist (RDT) to determine whether they met the minimal pain and disability criteria and confirmed the absence of neurological conditions. During session two, pain, demographic and anthropometric data were measured. Self-reports of weekly engagement in physical activity (aerobic sessions >30 min, core/ abdominal and strength training) and if they frequently lifted objects weighing over 23 kg for their work, defined as a heavy job (Seidler et al., 2009) were recorded. Participants also filled out a Tampa Kinesiophobia (Kori et al., 1990) and Pain Catastrophizing Scale (Sullivan et al., 1995) to characterize their beliefs toward pain.

Participants were prepared for surface EMG (Section Surface Electromyography) and three dimensional motion capture (Section Kinematic and Kinetic Analysis) data collections, then they performed three trials of a highly controlled standardized right-to-left horizontal transfer task previously described (Butler et al., 2010) (Figure 1). Briefly, to a 5 s count, participants were instructed to on “1” lift a 3 kg mass orientated 60° to the right of the midline of their body with their right hand, on “3” transferring the mass to their left hand at the midline of their body, and then on “5” lower the mass 60° to the left of their midline. Mass lift and lower were recorded using a pressure sensor located at bottom of the mass, and hand transition was determined using an optoelectric switch. A height-adjustable table was used to ensure the mass was lifted from just below elbow height, and the optoelectric switch ensured the mass was lifted no higher than 5 cm. During the task, participants were instructed to “lift the mass with the arm in full extension and to minimize trunk and pelvis motion.” Consistent with our published protocol, tactile feedback was provided through light pressure on the upper thoracic spinous process using a wooden dowel adjusted while the participant was in their standing erect trunk posture (Butler et al., 2010). Participants were instructed to maintain contact to minimize trunk motion as trunk flexion would decrease the pressure and trunk extension would increase the pressure. A researcher monitored that task performance was consistent with the instructions and participants would repeat tasks if there was visual deviation from the instructions.

Participants then performed a series of maximum effort voluntary isometric contractions (MVIC) for EMG amplitude normalization including: (i) supine sit-up, (ii) side lying lateral flexion (left and right), (iii) prone back extension, and (iv) prone back extension coupled with axial rotation (right and left) (Butler et al., 2010) with each exercise repeated (two trials). Participants were provided verbal encouraged to maintain maximal effort for 3 s against non-elastic straps, with a 1 min rest between contractions to minimize fatigue. Following each contraction participants were asked to rate their effort and trials were repeated if perceived effort was below their maximum.

Following MVIC trunk strength (flexor and extensor moment production) was measured during two additional MVIC exercises. Participants were positioned in a prone or supine crook lying position with the HUMAC Norm Dynamometer (Computer Science Medicine Inc., Strongton, MA, USA) arm positioned anterior and inferior to the clavicle for trunk flexion, or superior and posterior to the spine of the scapula for trunk extension (Hasue et al., 1980), was secured around the torso using non-elasticized straps. The HUMAC centroid was positioned approximately 5 cm anterior to the posterior superior iliac spine, in line with the iliac crest, and non-elastic straps were secured to anchor the pelvis and shank. Following gravity correction, participants performed two trials with instructions and procedures consistent with the normalization tasks.

For strength testing, gravity corrected external moments sampled from the HUMAC, were digitized at 2,000 Hz using the same analog-to-digital converter for EMG acquisition (section Surface Electromyography). Custom Matlab code filtered these data using a 6 Hz second order zero-lag low pass Butterworth filter, and converted them to moment (Nm) according to equations provided by HUMAC. The maximum isometric moment for flexors and extensors was calculated as the highest average moment over a 500 ms window which was normalized to body mass to compensate for anthropometric differences between participants (Smith et al., 1985).

To test whether there were differences in muscle activation amplitudes and patterns between those with higher (STRONG) or lower strength (WEAK); a median split approach was applied to the mass normalized back extensor moments, within the rLBI and asymptomatic groups. Two-way analysis of variance models (ANOVA) (group and strength) tested for differences in demographics, anthropometrics, maximum flexor and extensor moment, motion, lifting moments, and timing data. Categorical data were compared using a Chi-square test for independence. These comparisons determined if confounders should be included as co-variates when testing for muscle activation (PC) differences. Normality of PC scores was assessed using a Kolmogorov-Smirnov test, with non-normal data transformed using a Johnson transformation. PC scores were analyzed using three-factor (group, strength, and muscle) mixed model ANOVAs for the abdominals and back extensors separately. Given previous work has shown PC1 is highly correlated to the overall muscle activation amplitude (Quirk and Hubley-Kozey, 2014), pearson correlation coefficients were calculated between average EMG amplitudes and PC1 scores for each individual back and abdominal muscle site to confirm this relationship and that the average amplitude provides a physiological references for otherwise unitless PC scores. Tukey simultaneous tests compared pairwise differences when significant. Statistical analyses were performed in Minitab (version 17, State Collage, PA). Alpha was set at 0.05 and Bonferroni corrected.

To compare the relative difference between the STRONG and WEAK groups for EMG and normalized strength, the percent ratios (STRONG:WEAK) were calculated for normalized muscle strength in Nm/Kg and for EMG activation amplitudes in percent MVIC within each subgroup (rLBI and ASYM).

Sixty-nine participants volunteered for the study and of those nine were women. Only two women were in the rLBI group. Given sex modifies muscle activation patterns (Hubley-Kozey et al., 2012) and only 2 women were in the rLBI group, the analyses were performed on the 60 men (30 rLBI reported LBI occurred 9 ± 2 weeks from the day of testing). For the STRONG and WEAK categories, the median back extensor moment threshold was 2.46 and 2.51 Nm/kg for the ASYM and rLBI, respectively. Characteristics of the separate participant groups are shown in Table 1. No group by strength interactions were significant. Significant group main effects found that the rLBI group had higher mass (p = 0.001), pain [VAS (p = 0.001)], disability [RMD (p = 0.003)], pain catastrophizing [PCS (p = 0.004)] and kinesiophobia [TSK (p < 0.001)] scores than ASYM. Differences between the STRONG and WEAK groups were significant for both the non-normalized (p < 0.008), and normalized maximum moment produced for both trunk flexors and extensors (p < 0.001) (Table 1).

Time to complete the task, trunk and pelvis motion, and peak external moments for the horizontal transfer task are shown in Table 2. There were no group or strength main effects or interactions (p > 0.05) supporting that the controlled task parameters of timing (approximately 4 s), minimizing trunk and pelvis motion and task demand of the external moment was consistent among subgroups.

To address the primary objective, strength main effects and interactions are first presented along with any other main effects and interactions to fully describe the data. Representative ensemble average profiles for the abdominals, and back extensors are shown in Figures 3A–D, respectively to depict group by strength interactions.

Two PC's explained 96.6% of the total abdominal muscle and 93.9% of the total back extensor muscle activation waveform variance. For both the abdominals and back extensors PC1 captured the overall waveform shape and amplitude (Figures 4A,E) where high scores captured higher activation amplitudes (Figures 4B,F). PC1 scores were highly correlated (r = 1) with the average amplitude of muscle sites across the entire task so the average activation amplitudes in %MVIC are also provided to give context to PCs unitless score to facilitate data interpretation. For both the abdominals and back extensors, PC2 captured a differential in periods of higher vs. lower muscle activation in response to the changing lateral flexion moment (Figures 4C,G). High scores corresponded with muscle sites having higher initial (0–25%) activation vs. lower terminal (75–100%) activation, whereas low (negative) scores corresponded to the opposite with lower initial activation (Figures 4D,H). Beyond directionality, captured by positive and negative scores, the magnitude of a PC2 score captures relative “responsiveness.” High magnitude scores convey a greater relative change (differential) in muscle activation patterns to the changing external moment whereas a low magnitude score captures that the muscle activation pattern had consistent activation occurring at the beginning and end of the task.

Three of the four EMG comparisons examined had a significant strength main effect or group by strength interaction. Consistent with the EMG to force relationship (Brown and McGill, 2007), WEAK recruited more motor units, reflected by higher EMG amplitudes of the primary agonist back extensor muscles (Butler et al., 2010). This finding is consistent with studies of lower limb muscles that showed for a variety of tasks (calf raise, walking, sit-to-stand, stair ascent, and descent) the maximum muscle specific strength was negatively correlated with agonist EMG amplitudes of the knee extensors (r = −0.3–0.7) and ankle plantar flexors (r = −0.4–0.5) (Takai et al., 2008).

The current study also found that participants with weaker back extensors recruited higher antagonist abdominal activation. If muscle recruitment is designed to minimize antagonist activation (Brown and Potvin, 2005) one might question why higher abdominal activation was observed. There are four plausible factors that could explain this finding. First, the line of action of the abdominal oblique sites, specifically the lateral sites have been shown to contribute to generating lateral flexor moments (Brown and Potvin, 2007). These lateral obliques would contribute as an agonist to the lateral flexion moments generated at the beginning and end of the lifting task. Second, the EO1 and IO sites (Arjmand et al., 2008a), balance axial rotation moments produced by unilateral activation of specific back extensor and abdominal muscles (Brown and Potvin, 2007; Arjmand et al., 2008b). These two factors provide a plausible explanation why the oblique sites had higher activation amplitudes than rectus abdominus sites which are designed to produce flexor moments (Table 4) (Brown and Potvin, 2007; Arjmand et al., 2008b). A third factor could be related to the stability demands of the lifting task. These three explanations are related to reduced active system function. The fourth explanation is that those in the WEAK group might have systematic changes in neuromuscular control that result in heightened antagonist co-activation during task performance. This antagonist co-activation may be explained by changes in other spinal systems, and would reduce the net moment produced by a participant. Computational modeling suggests the spine requires antagonist force production to maintain stability (Brown and Potvin, 2005). Whether there is a standard moment that must be produced by the antagonist to ensure joint stability is unknown. If a threshold did exist, weaker individuals would require higher activation amplitudes to meet this threshold. Our data showed that the WEAK group also had significantly lower trunk flexor strength than STRONG (Table 1), supporting that the abdominal muscles compensated for this weakness to fulfill the above-mentioned roles. This finding is consistent with previous work showing individuals with lower knee extensor strength had higher knee flexor EMG activation amplitudes for fundamental tasks such as sit-to-stand and stair ascent/descent (Mizelle et al., 2003).

The results of this study suggest increased EMG activation amplitudes potentially act as a form of adaptation toward reduced force production. While these changes may be necessary to prevent instability events associated with insufficient force production, increases in muscle activation amplitudes represent a unique risk for low back pain. In participants who are WEAK, increased activation amplitudes indicate that muscles are at higher risk to fatigue (van Dieën et al., 2008) resulting in an impaired ability to restore stability through active force (van Dieën et al., 2011). Furthermore, data showing higher agonist and antagonist activation amplitudes suggest that despite the peak net moment being similar between groups (Table 2) altered internal force balance could result in increased spinal compression forces (Granata and Marras, 2000) that can exacerbated joint degeneration (Wang et al., 2007). Going forward more sophisticated models will be required to estimate the moments produced by antagonist muscle groups to support whether this co-activation could lead to higher compressive forces.

Novel to this study was the change in the relative responsiveness of trunk muscles to the dynamic changing moments. For the back extensor sites, WEAK participants were more responsive to changes in the lateral flexion moment produced at the beginning and end of the task (PC2) (Table 3) than STRONG. While not directly measured, sagittal plane trunk strength has been correlated to frontal (r = 0.81) and transverse plane (r = 0.91) strength (Kocjan and Sarabon, 2014). Thus, weaker individuals should require higher activation of lateral flexor sites to meet the frontal plane moment demands produced by the horizontal transfer task consistent with previous studies showing greater responsiveness (PC2 scores) of trunk muscles sites with increasing external loads (Quirk and Hubley-Kozey, 2014).

The secondary purpose of this study was to determine whether the effects of strength on muscle activation patterns were modified by the recent experience of low back pain despite recovery of symptoms. The rLBI reported below clinically meaningful VAS and RMD scores consistent with our inclusion thresholds aimed at assessing recovery. However albeit small, they did have, significantly higher levels of pain (Jensen et al., 2003; Boonstra et al., 2014) and disability (Turner et al., 2004) than the ASYM group on the day of testing (Table 1). Self-reported psychosocial measures were also at the lower end of the pain catastrophizing and fear avoidance scales (Sullivan et al., 1995), however, the statistical analysis captured small but significant differences between the rLBI and ASYM groups (Table 1). Despite mass differences between the rLBI and ASYM group, mass normalized strength (Table 1) and the external moments produced by the horizontal transfer task (Table 2) were similar between groups. As the primary objective of this study was to compare between WEAK and STRONG participants mass deviation between the rLBI and ASYM group would not modify these differences. This was confirmed by including mass as a co-variate in the ANOVA model where there was no change in any group, strength or group by strength interactions captured on PC scores.

The mean strength of participants from this study while within the range of supine trunk flexor strength (136 ± 30 vs. 139.8 ± 35.2 Nm) was above average for prone back extensor strength (212 ± 36 vs. 179.2 ± 31.2 Nm) when compared (vs.) to a cohort of similar 20–60 year old men (Hasue et al., 1980). Higher back extensor strength can partially be explained by the military participants in the current study who have higher levels of fitness (including grip strength) than civilian controls (Deuster et al., 1987), where grip strength has been associated with higher back extensor strength (Wang et al., 2005). In contrast to reports of lower trunk strength in individuals with chronic LBP compared to healthy controls (Hasue et al., 1980; Newton et al., 1993; Steele et al., 2014a), our study found there was no difference between the rLBI and ASYM groups. A result consistent with previous work showing that participants with a history of LBP of varying durations do not have different trunk strength compared to controls (Hultman et al., 1993; McGill et al., 2003a).

Key findings of this study were the two significant strength by group interactions for overall muscle activation amplitudes (PC1) of both the back extensors and abdominals. For the abdominals, the WEAK group had relatively greater differences in overall abdominal activation than STRONG within the rLBI group compared to the ASYM group (Table 3). Comparing ratios for EMG and strength (Table 6), showed that the relative difference in activation amplitudes are likely explained by a greater flexor strength ratio between the STRONG and WEAK rLBI participants. Similarly, for the back extensors, the WEAK rLBI subgroup had higher activation amplitudes than all other subgroups (Table 3). However, unlike in the abdominals contrasting qualitative strength and EMG ratios (Table 6) would suggest this relative difference could not be explained by back extensor strength ratios which were comparable between the STRONG and WEAK ASYM and rLBI subgroups (Table 6).

To account for these EMG differences two theories were investigated. First, previous work in those with chronic LBP suggests that those with higher pain and disability have lower trunk strength (Newton et al., 1993). This link between symptoms and strength may be associated with a reduced ability to produce a true maximum voluntary isometric contraction in those with chronic LBP (Chiou et al., 2013). These differences would inflate the amplitude of MVIC normalized EMG. However, participants in our study were deemed recovered, reporting minimal pain which, has not been shown to limit the ability to produce a true MVIC relative to healthy asymptomatic controls (Chiou et al., 2015). To explore whether pain could explain these differences in EMG amplitudes subjective questionnaires were implemented; however, VAS, RMD, PCS, and TSK were not different between the WEAK vs. STRONG rLBI subgroups (Table 1). Furthermore, to explore whether pain could confound our results we found in the 4 participants with VAS scores characterizing mild pain (ranging from 18 to 28/100 mm) two were in the STRONG and WEAK group, respectively, suggesting the current experience of pain was not a confounder.

A second theory for the higher activation is that the influence of recent pain modified how trunk muscle activation patterns adapt to a stability challenge associated with lower back extensor strength. Experimental work showed, that when an individual is in pain they utilize muscle activation patterns that generally increase the overall activation for all muscle sites to increase joint stiffness and thus stability (Hodges et al., 2013) and according to the motor adaptation to pain theory, once pain resolves some individuals retain these adjusted muscle activation patterns (Hodges et al., 2013). This study expands on this model to hypothesize that individuals with lower active spinal system function (WEAK) might experience greater adaptations to ensure spinal stability that is beyond a margin of safety, whereas those with higher active system function (STRONG) have fewer adjustments. This hypothesis is consistent with recent work showing that changes in trunk muscle activation was greatest when the spine has lower stiffness presenting a higher stability demand (Shojaei et al., 2018). While the results of this study are not definitive, they do support and expand upon the motor adaptation to pain theory. Future research should explore whether those with lower function in a spinal system experience more instability events (painful events). If this were the case it suggests that muscle activation pattern adaptations would remain insufficient and necessitate further adaptations to mitigate the risk of pain. Continued painful events could lead to the appearance of exaggerated adaptation and possibly retention of these pain developed muscle activation patterns. However, more empirical evidence is needed to support the motor adaptation to pain theory and retention of motor patterns post injury.

A limitation of the current study was that only one measure of active spinal system function was assessed; the maximum voluntary isometric moment of the two main muscle groups with no measure of axial or frontal plane muscle strength. Other work has shown isometric trunk strength is correlated between all three fundamental planes (Kocjan and Sarabon, 2014) and hence we chose the back extensor sagittal plane moment as a representative measure of strength that would directly oppose the principal flexion moment produced during this fundamental task. Second, while this study explored whether those with deficits in the active spinal system would have different muscle activation patterns our definition of a deficit was based on participants having below median back extensor strength. While our measure of lower function (WEAK) has been reported as is a potential risk for experiencing future low back pain (Cho et al., 2014), the threshold used in this study was not based on a clinical threshold of deficient function. Encouraging however was that given the potential for overlap in the back extensor strength values between the two groups using the median split approach, there were significant differences in muscle activation patterns with strength and groups. Only including men was a limitation and this was not by study design and not within control. A secondary analysis of only strength main effects (rLBI could not be compared) including the 9 women did not change the key results of this study; however, generalizing the findings from this study to women or non-military populations should be done with caution. A strength of this study was our relatively large sample size with equal numbers between those who have and have not experienced a recent low back injury within the last year. Finally, the study sample included a rLBI group whereas much of our in vivo knowledge on low back conditions to date have come from studies focused on those with chronic LBP. Understanding muscle function and adaptations early in the injury process might help in developing clinical targets, where the chronic group findings are often confounded with psychosocial issues and habitual movement patterns.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.