Twenty-three subjects who reported no recent (prior year) history of back pain were compared to a group of twenty-one subjects who reported current LBP, (Table ). Subjects were studied after they signed the informed consent form that had been approved by the Institutional Committee on Human Research. Trunk muscle activation was recorded during steadily increasing isometric efforts to each subject’s maximum voluntary effort, and their muscle preactivation and muscle responses to a transient force perturbation were recorded with two different magnitudes of preload.
Details of the subjects studied
Subjects responded to announcements of the study posted in public places, in medical facilities and in a local newspaper. They were recruited into the study if they were between 18 years and 60 years of age. Those in the LBP group were determined in a structured interview to have a history of episodic LBP (one or more episode lasting more than 3 days and less than 4 weeks), and on the day of recruitment reported LBP greater than 3 on a 0–10 range visual-analog scale. Subjects were excluded from the LBP group if they had: (1) an objectively diagnosed cause of their pain (herniated nucleus pulposus, tumor, fracture, stenosis, or neuropathy evidenced by symptoms radiating below the knee); (2) prior spine surgery; (3) deformity: symptoms or treatment attributed to scoliosis or spondylolisthesis; or (4) pending legal action associated with their LBP.
For testing, the subjects first had surface EMG electrodes applied at ten locations on the trunk (described below). Then they stood in an apparatus [29
] with the pelvis effectively immobilized by a support structure with pads pressing on the regions over the anterior superior iliac spines and the sacrum (Fig. ). A harness around the subject’s thorax was connected via a cable and pulley to the system for applying a force perturbation of variable (and controlled) amplitude and duration (Fig. a). Prior to the perturbation, subjects pulled against the cable to generate the predetermined preload. The cable was aligned approximately horizontally and at the level of the T-12 vertebra. The pulley was attached to one of five anchorage points on a wall track surrounding the subject at angles of 0°, 45°, 90°, 135° and 180° to the anterior direction. The mechanical system for generating the force perturbation (Fig. b) consisted of an electric motor driving an eccentric-crank lever system via a single turn electromagnetic clutch, activated by the experimenter pushing a button. This produced a single full sine-wave displacement of the lever arm attached to the two springs in line with the cable connected to the harness around the subject. The amplitude of the sinusoidal displacement and the stiffness of the springs determined the amplitude of the force perturbation superimposed on the preload efforts. Here, ‘effort’ refers to the measured external force in the cable.
Diagram showing a the arrangement of the apparatus relative to the subject, and b the waveform of the force perturbation generated by the single turn of the motor
Initially, the cable was anchored to the wall at 0° (extension effort) and subjects generated a timed ramped effort test up to their maximum effort in 5 s with a further 5 s for gradual release of the load. A computer screen in front of the subject displayed a vertical bar whose height was proportional to the effort generated, and with a mark to indicate the prior maximum effort. Three trials were performed to help subjects to learn how to achieve a maximum effort. The maximum achieved was used as the basis for determining the preload effort and perturbation amplitude in the perturbation experiments.
At each of the five test angles (the sequence of angles was randomly selected) subjects first performed three ramped maximum effort tests. Then, they were instructed to generate a preload of nominal magnitudes 15% and 30% of the maximum effort recorded in the extension efforts, by pulling against the cable. The computer display with a target mark was used to help subjects maintain the desired steady-state preload effort. The subjects were instructed to maintain a normal erect posture, symmetrically oriented with the apparatus during all tests. A single full sine-wave force perturbation pulse (nominal amplitude 5% or 10% of maximum effort, nominal duration 80 ms) was triggered by the investigator without warning at a random time between 5 s and 10 s after the subject reached the desired steady-state preload effort. They were expecting the force perturbation and were instructed to maintain the target effort until after the perturbation (i.e., not to react actively to it); then they were allowed to relax. The subjects had experienced the force perturbations in a practice session prior to the recorded trials. At each angle there were four test conditions (two preload efforts, two pulse amplitudes) which were randomly presented. Three repeated trials of each test condition were made sequentially. The total time of the testing session was about 3 h, and the typical duration of a sustained effort was about 15 s. Subjects could rest between trials and while the load direction and perturbation parameters were altered, and were given a rest period of about 10 min after half of the testing protocol was completed.
Bipolar EMG electrodes (Delsys Inc., Type DE-02.3, Boston, MA, USA) recorded signals from five right and left pairs of muscles (rectus abdominis, internal and external obliques, longissimus, iliocostalis). The Delsys electrodes have 10×1 mm silver-bar electrodes with 10-mm spacing; their single differential amplifiers have a gain of 1,000, bandwidth 20–450 Hz, 92 dB (typical) common mode rejection ratio, and 1012 Ω input impedance. A ground electrode was placed over the lateral epicondyle of the elbow. EMG and load cell signals were recorded digitally at 2,048 Hz.
The electrodes placements were: rectus abdominis—30 mm lateral to the midline at the level of the umbilicus, aligned vertically; external oblique—halfway between the iliac crest and the 12th rib along the mid-axillary line, aligned at an 80° angle to the horizontal; internal oblique 20-mm medial and superior to the anterior superior iliac spine, aligned vertically; longissimus—30 mm lateral to the midpoint of the spinous process of L-3, aligned vertically; iliocostalis—60 mm lateral to the midpoint of the spinous process of L-3, aligned vertically. Suspect data resulting from technical problems such as loose electrodes, or EKG artifacts were excluded from further statistical analyses by a dual process of visual inspection, and identification of outliers. Overall, 1.3% of recordings were excluded, with no evident predominant rate of exclusion by electrode location.
For the ramped-effort tests, the EMG signals were passed through an RMS filter with a moving window having a width of 250 ms. Then a linear regression analysis was performed for the increasing-effort part of the recording between the RMS-EMG signal and the force (effort) generated. Muscle activation was obtained from the gradient of the EMG–effort relationship, normalized in each of two ways. In the first, the gradient was divided by the maximum EMG value recorded for that muscle and subject in all the ramped-effort tests, and multiplied by the value of the maximum effort for that testing angle, hence, it was nondimensional. In the second normalization, the gradient was divided by the maximum EMG value for that muscle, without normalization by effort (hence, it had units of kN−1). Thus, in both normalizations there was division by the maximum EMG to account for differences in electrode and amplifier gain, etc. Only the first normalization accounted for differences in effort. The source of differences in maximum effort was unknown. If it was due to anatomical differences (e.g., in muscle size) then normalization by the maximum effort would provide better comparison between different individuals’ muscle recruitment patterns. If the different effort was due to differences in subject motivation, then normalization by maximum effort would not be appropriate for measuring relative muscle activation.
The muscle preactivation was recorded while subjects were maintaining the preload just prior to the perturbations. This preactivation was quantified by averaging the EMG signal magnitude in the window 25–150 ms prior to the onset of the perturbation and it was expressed (nondimensionally) as a proportion of the maximum EMG value for that muscle obtained from all the ramped-effort tests of the corresponding subject.
For the perturbation tests, EMG signals were first bandpass filtered by a 10–100 Hz Chebyshev type II filter with no lag, and rectified. The filtration was intended to reduce any EKG or motion artifact and high frequency noise contamination of the signals. A 25 ms moving average of the rectified EMG signal was then calculated. The onset of the force perturbation was first identified from the load cell recording by detecting the time at which there was a significant increase in the effort–time slope. A 25–150 ms time window after the force perturbation was examined to detect any short (reflexive) and medium (automatic) latency muscle responses to the perturbation. Two different methods of response identification were employed, which are as follows:
A muscle response to force perturbation was considered to occur if the processed EMG signal exceeded a threshold of 3 standard deviations (SDs) above the baseline (preactivation) EMG signal. If a response was detected, a value of one was assigned, otherwise zero was assigned [29
]. The latency was measured as the time from the start of the force perturbation to onset of the EMG response, and then was used to select responses between 25 ms and 150 ms after the beginning of the perturbation for inclusion in the analyses (Fig. ).
Fig. 2 Sample recording of an EMG signal from a perturbation experiment. The upper panel shows the force recording, that defined the force perturbation onset time (that defined zero time). Relative to this time, pre-perturbation and post-perturbation ‘windows’ (more ...)
Mean EMG difference (MEMGD) method [29
The difference between the mean EMG signal in a 25–150 ms window after the perturbation and a 25–150 ms window prior to perturbation was computed for each EMG signal. A value of zero was assigned if the difference was negative, and unity if positive. The observed numbers of responses were examined relative to the 50% rate expected by chance if there was no true increase is muscle activity. This detection method is not influenced by the differences in the SD of the baseline EMG signal between experimental conditions that could potentially produce detection bias with the Shewhart method.
Antagonistic/agonistic activation ratios
The coactivation pattern for individual muscles was evaluated by two methods: (1) Flexion-extension measure: the mean activation (or preactivation) was calculated at the angle where the muscle was considered to be antagonist (activation averaged over the three repeat trials) and was expressed as a proportion of the activation (or preactivation) when it was considered to be an agonist. For dorsal muscles, this was the ratio of activation at 180° (antagonist) and 0° (agonist) effort angles, and for abdominal muscles, the ratio at the 0° (antagonist) and 180° (agonist) effort angles. (2) Lateral bend measure: the right muscle activation (or preactivation) was expressed as a proportion of the left muscle activation (or preactivation) at the 90° effort angle.
To analyze the likelihood of muscle responses, a dichotomous response was recorded for the Shewhart and MEMGD methods for each of three trials at each combination of experimental conditions. The estimated muscle response frequency was the average over the three trials. An arcsine square root transformation was applied to these frequencies to improve compliance with assumptions made in the subsequent analysis of variance (ANOVA). ANOVA was also used to evaluate the significance of group-wise differences between muscle activation and preactivation. In all analyses, a probability less than 0.05 was considered statistically significant.