The effects of a three-week use of lumbosacral orthoses on trunk muscle activity and on the muscular response to trunk perturbations

Fourteen subjects (11 males and 3 females) volunteered for the study. Their age, weight, and height were on average (SD) 26 (8) yrs, 81(14) kg, and 180(13) cm, respectively. None of the subjects had any history of low back pain or neurophysiological disorders. At the beginning of the study, all subjects read and signed an informed consent form describing the experimental protocol that was approved by Yale University's Human Investigations Committee.

The study protocol required all 14 subjects to wear a LSO (Aspen Medical Products, Inc., Long Beach, CA, USA) (Figure 1) for a period of 3 weeks. This device restricted trunk range of motion and increased trunk stiffness by amounts similar to comparable orthoses on the market, but was rated significantly more comfortable to wear by subjects [4]. Per manufacturer recommendations, the back panel of each individual LSO was adjusted to fit the contour of each subject's lumbar lordosis. Subjects were instructed to wear the LSO for a minimum of 3 hours a day during periods of activity. Any period sitting or lying down while wearing the LSO did not count towards the 3-hour daily minimum. This 3-hour period did not need to be continuous. Our preliminary studies suggested that healthy subjects might not comply with wearing a LSO for a longer period. To further maximize compliance, subjects were required to maintain a daily log in which they recorded the times they started and stopped wearing the LSO. In addition, one member of the research staff contacted each subject daily by telephone, email, or in person to verify that the LSO was worn. All subjects were instructed on how to wear and tighten the LSO. Initially, the LSO tension was adjusted to reach a pressure of 35 mmHg (4.7 kPa, measured with the Therapoint pressure measurement system, Roho, Inc., Belleville, IL, USA) between the LSO and abdominal wall at the location just lateral (left or right) to the umbilicus. This pressure was empirically selected when a balance between maximum tension in the brace and comfort to the user was achieved. The subjects then positioned and tightened their LSOs each day during the study to approximate that tension. The pressure measurement was set to 35 mmHg at the beginning of each testing session to standardize the LSO tension for all subjects.

The measurement of dependent variables was performed in three testing sessions on days 0, 7, and 21, with and without the LSO in each session. Half of the subjects were randomly selected to begin the testing sessions wearing the LSO, and the other half began their sessions without the LSO. This sequence was reversed at each subsequent testing session to correct for the effect of testing order.

Each testing session consisted of a quick trunk force release, isometric lift, and unsupported sitting tasks. For the quick force release task, each subject was placed in a semi-seated position in a custom-built apparatus and exerted isometric exertions in trunk flexion, extension, left lateral bending, and right axial rotation (Figure 2). This apparatus restrained lower body motion, leaving the upper body free to move in any direction. The pelvis was fixed at 4 points: the acetabulum via fixed femurs, the ischium, and at the anterior and posterior superior iliac spines. Thus, any postural adjustments through hip, knee, or ankle joints were eliminated. A cable attached to a chest harness at approximately the T5 level was held with an electromagnet and served as a resisting force. The subject pulled against the cable until (s)he reached a target force, displayed on an oscilloscope. As in our previous studies, the target force was set at 115 N for men and 80 N for women. Upon release, these forces resulted in the largest possible trunk displacement without being physically uncomfortable to the subjects [31, 32]. An experimenter randomly released the electromagnet within 0 to 5 seconds after the target force was reached. The electromyographic (EMG) and trunk kinematics data were collected for 1 second prior to and 2 seconds after the release.

The isometric lifting task was performed by the subject while standing with slightly bent knees and the trunk flexed at the hips at 45 degrees with respect to the horizontal plane. A subject held 30% of his/her body mass in a crate with straight arms perpendicular to the floor. Trunk angle was measured with an inclinometer and the subject was instructed to maintain the natural lordotic curvature of the lumbar spine during the trial.

The sitting task was executed on a hard-surface seat with foot support but no back support. The knees formed a 90-degree angle and the arms were crossed on the chest. For both lifting and sitting tasks, EMG data were collected for 3 seconds after a correct, stable posture was achieved. All trials were repeated three times and the results were averaged across the three trials.

For the purpose of EMG normalization in each testing session, a series of maximum voluntary exertions (MVE) was executed, which were designed to maximally activate the latissimus dorsi and other trunk muscles while the subject attempted to perform sit-up, trunk extension, and lateral bending exercises [33]. These tasks were performed on the examination table with the experimenter providing manual resistance. Each MVE was sustained for 3 seconds.

EMG data from twelve major trunk muscles were recorded using bipolar, Ag-AgCl, disposable, surface electrodes (Graphic Controls, Buffalo, NY, USA). The electrodes were placed with center-to-center spacing of 3.5 cm over the following muscles on each side of the body: rectus abdominis (3 cm lateral to the umbilicus), external oblique (15 cm lateral to the umbilicus), internal oblique (approximately midway between the anterior superior iliac spine and symphysis pubis, above the inguinal ligament), latissimus dorsi (lateral to T9 over the muscle belly), thoracic erector spinae (5 cm lateral to T9 spinous process), and lumbar erector spinae (3 cm lateral to L3 spinous process). The EMG data recorded from these sites have previously been shown to be unaffected by the additional tension from wearing a LSO [34]. All EMG signals were band-pass limited between 20 and 420 Hz, differentially amplified (input impedance = 100 GΩ, CMRR > 140 dB), and sampled at 1600 Hz. The cardiac QRS waves were removed using the modified turning point and adaptive sampling algorithm [35].

The trunk kinematic response to the quick force release was measured with a three-dimensional electromagnetic motion-measurement device (Flock of Birds, Ascension Technology, Corp., Burlington, VT, USA). The source of a magnetic field was mounted on the testing apparatus and the sensor was attached to the subject at the T5 level using elastic straps. Angular displacements of the trunk were recorded at 120 Hz. All recorded data were processed to yield the following outcome measures: muscle response to the quick trunk force release, effective trunk stiffness, and spine compression force as a measure of total muscle activity.

As shown in our previous work, following the quick force release, agonistic muscles that were active before the release were expected to shut-off reflexively. The antagonistic muscles that were inactive before the force release were expected to switch-on reflexively [36, 37]. In flexion, trunk flexors acted as agonists and extensors as antagonists. In extension, trunk extensors acted as agonists and flexors as antagonists. In lateral bending to the left, ipsilateral muscles (left side) acted as agonists and contralateral muscles (right side) acted as antagonists. In right axial rotation, the right internal oblique, left external oblique, and left lumbar erector spinae muscles were considered agonists and the left internal oblique, right external oblique, and right lumbar erector spinae muscles were considered antagonists. The activation of the rectus abdominus, latissimus dorsi, and thoracic erector spinae were not considered in axial rotation because of the difficulty in classifying them functionally as agonists or antagonists.

The detection of onset and offset of muscle activities was automated using a model-based algorithm developed by Staude and Wolf [38] and implemented in the Matlab software environment (The MathWorks, Inc., Natick, MA, USA). This algorithm showed superior performance when compared to the traditional threshold based methods [38, 39]. The EMG signal was first pre-processed with an adaptive whitening filter. The potential onsets and offsets were detected when significant changes occurred in the EMG signal modeled as a Gaussian random process. All such detected events were then ranked based on their maximum likelihood statistics. An appropriate event with the largest generalized likelihood ratio was selected as the onset or offset from the 20 to 150 ms time interval following the force release. The assumption was made that reflex responses could not occur any earlier than 20 ms after the stimulus. Any responses occurring later than 150 ms after the stimulus may represent voluntary and not reflexive muscle activity.

To quantify the overall trunk muscles' response to the quick force release, the average latency and the number of muscles that responded to the quick force release were computed as in our previous studies [36, 37]. There were six muscles expected to switch-on and six to shut-off in the trunk flexion, extension, and lateral bending trials. Thus, the dependent measures in the quick force release trials were (i) average latency and (ii) the number of the muscles switching-on and shutting-off in each direction of the quick force release. A high background muscle activity in axial rotation precluded the reliable identification of muscle onsets and offsets in that quick force release direction.

The effective trunk stiffness was estimated from the trunk kinematic response to the quick force release [6]. The trunk was represented with a second order, inverted pendulum model. Its angular displacement response (θ) to the force release is determined by the trunk stiffness (K), inertia (I), and damping (B) (Equation 1):

where C is a constant related to force, mg is trunk weight, and L is the height measured from the L4-L5 joint to the center of trunk mass, assumed to be at the T9 level. Muscle reflex response modifies the trunk kinematics after the release [40]. Therefore, the stiffness (K) and damping (B) are called the "effective trunk stiffness" and "effective trunk damping", respectively, because they combine the effects of the trunk muscles' stiffness and damping established prior to the release and the muscle reflex response following the release [40]. The effective trunk stiffness and damping was obtained from fitting the parameters B, K, and C in the Equation 2 to the experimental data (θ) up to the maximum trunk displacement, according to the previously established method [6, 40]. The inertia (I) was computed as 53.6% of body weight times L² [41].

The spine compression force reflected the overall trunk muscle activation level established immediately prior to the quick force release. The estimate of the spine compression force was obtained from a detailed biomechanical model of the lumbar spine, which we have described earlier [29]. The model consisted of five lumbar vertebrae between the rigid pelvis and ribcage, and 90 muscle fascicles. Each intervertebral joint was represented by a nonlinear, lumped parameter disc and ligament equivalent for stiffness about the three axes of rotation. Thus, the system consisted of 18 degrees of freedom (6 joints × 3 degrees of freedom each).

Muscle forces were estimated based on 200 ms of EMG data recorded immediately before the force release from 12 trunk muscles. The EMG data were rectified, averaged, and expressed in %MVE. After accounting for the contributions of passive tissues, the moments and forces necessary to balance external loads and upper body weight were distributed among all 90 muscle fascicles using an EMG-assisted optimization method [42]. The muscle forces and muscle stiffness were first estimated from EMG data using a cross-bridge bond distribution moment model reflecting muscle contraction dynamics [43]. A quadratic optimization algorithm was subsequently applied to minimize the adjustment of individual muscle forces (muscle gains), while requiring that the moment equations about the three rotational axes of every intervertebral joint were balanced [44]. All muscle forces and external loads were then summed along the axis perpendicular to the L4-L5 intervertebral disc to obtain a joint compression force. The direct effects of intra-abdominal pressure or the 35 mmHg LSO pressure on spine compression force were not considered.

The only significant main effect in the trunk muscle response latencies to the quick force release was due to the direction of the force release (Table 1). On average, the muscle offset latencies were shorter in trunk extension or flexion than in lateral bending (67(35) and 67(31) vs. 78(35) ms, respectively) (df = 2; F = 3.78; P = 0.024). The average muscle onset latencies were shortest in trunk extension, then in flexion, and the longest in lateral bending (55(7), 60(10), and 64(10) ms, respectively) (df = 2; F = 25.85; P < 0.001). There were no other main effects or their interactions present in the muscle reflex latencies involving the LSO or Session (P > 0.5).

The average number of agonist muscles switching off following the quick force release was significantly greater when the LSO was worn compared to the No LSO condition (1.8(1.0) LSO vs. 1.4(1.0) No LSO) (df = 1, H = 8.58, P = 0.003) (Table 2). On the other hand, the number of antagonists responding with an onset of muscle activity was significantly greater on the 2^nd test (day 7) (df = 2; H = 8.64; P = 0.013) (Table 2). No other main effects were present in the number of muscles responding to the quick force release.

The LSO significantly (df = 1; F = 17.21; P < 0.001) increased the effective trunk stiffness by 160 Nm/rad or 27% when averaged across all directions and testing sessions (Figure 3). The testing direction also had a significant effect on trunk stiffness (df = 3; F = 20.85; P < 0.001), but no statistical interactions with the other factors of interest (i.e. LSO, Session) were found. There was no significant change over time in the effective trunk stiffness estimated from the trunk kinematic response to the quick force release (P > 0.05).

The spine compression force reflected the aggregate trunk muscle activation during the isometric exertions just prior to the quick force release, and during the sitting and lifting tasks. No statistically significant effects of the LSO or Session were present (P > 0.05) (Figure 4). The only significant differences in spine compression force were due to the different tasks and the direction of the quick force release (Figure 4).