The acute effects of targeted abdominal muscle activation training on spine stability and neuromuscular control

Participants

Thirteen healthy students (7 M, 6 F) with no history of LBP or current musculoskeletal/neurological
injuries were recruited through an online university study recruitment database. Before
participating, each eligible volunteer was required to sign the personal information
and informed consent statement that was approved by the Nipissing University Research
Ethics Board (14-07-04).

Instrumentation

Surface EMG data were recorded at 2000Hz from four muscles bilaterally (Fig. 1a): thoracic and lumbar erector spinae (TES and LES), and internal and external oblique
(IO and EO) (Trigno, Delsys Inc., USA). Prior to application of surface EMG sensors,
all locations were shaved and cleaned with alcohol to ensure low impedance. Fine wire
EMG was recorded synchronously from three muscles unilaterally (right side): IO, TrA,
and the deep fibers of multifidus (MF). All indwelling electrodes were inserted under
the guidance of ultrasound imaging (USI) (Voluson i, GE Health Care, UK) to ensure
correct positioning (Fig. 1b). Kinematic data were collected concurrently at 50 Hz from 12.7 mm reflective markers
(BL Engineering, Santa Ana, CA, USA) placed over key body landmarks to track 3D whole-body
motion using 13 motion capture cameras (Oqus 400+, Qualisys, Sweden) (see Additional
file 1, which lists all tracking and calibration markers).

Fig. 1. a Experimental setup for surface EMG: external oblique (EO), internal oblique (IO),
thoracic erector spinae (TES) and lumbar erector spinae (LES). b Ultrasound images of the muscles of interest for indwelling EMG: internal oblique
(IO), transversus abdominis (TrA), and deep multifidus (MF)

Protocol

Following instrumentation, participants performed three maximum voluntary isometric
contractions (MVICs) against manual resistance for the abdominal muscles (upper trunk
flexion, upper trunk flexion combined with left and right twisting, and upper trunk
flexion combined with left and right lateral bending while lying supine) and three
MVICs against manual resistance for the back muscles (upper trunk extension, upper
trunk extension with left and right twisting, and upper trunk flexion with left and
right bending while lying prone) 20].

Each participant then performed two trials of 35 cycles of repetitive unloaded spine
flexion with a constrained pelvis to the beat of a metronome at a rate of 15 cycles/min
16] (Fig. 2a). This number of cycles was chosen as it provides sufficient data to obtain accurate
dynamic stability estimates 21], and the rate was chosen as it has been found to be the preferred movement rate in
several studies 16], 21]. Within each cycle, participants were required to touch two targets with their hands
extended in front of them: the top target was located in front of them at shoulder
height in the mid-sagittal plane so that it could be reached when standing upright
with the arms extended, while the second target was located in the mid-sagittal plane,
50 cm anterior to the knee when participants were standing upright with their hips
and knees extended 16], 22], 23] (Fig. 2a).

Fig. 2. a Experimental setup and task requirements. b Visual-3D model while performing experimental task

Between trials, participants were instructed by a Registered Physiotherapist with
post-graduate training and experience in USI on how to perform the ADIM in a standing
position, and USI was used as biofeedback to ensure successful contraction of the
TrA. To improve the external validity of this study, the training of the ADIM closely
mimicked how the ADIM would be taught in clinical practice. Participants were verbally
cued to slowly draw their lower abdominal wall towards their spine. By visual inspection
and palpation, the physiotherapist ensured that the participant had no posterior rotation
of their pelvis and did not hold their breath during the maneuver 24]. A successful contraction (ADIM) was defined as a visible increase in thickness of
the TrA on USI prior to any observed thickness changes to the IO or EO 25]. Participants were then instructed how to perform the ADIM at the onset of each repetitive
movement cycle and to hold the contraction throughout the entire range of motion.
Participants released their ADIM when they returned to the initial stance position
and were instructed to contract their TrA at the start of every cycle. The ADIM was
practiced with the movement cycle while the physiotherapist manually palpated the
contraction to ensure the participant was able to hold the contraction throughout
the entire movement. Once the participant could perform the activity correctly and
consistently, testing began.

Data processing analyses

Three-dimensional whole-body motion data were processed in Visual-3D (C-Motion Inc.,
USA), but only 3D lumbar spine kinematics were analyzed here (Fig. 2b). EMG signals were processed by first removing the DC offset by subtracting the mean
of the entire signal from each data point. Next, EMG data were bandpass filtered between
20 and 450Hz, full-wave rectified, and then linear enveloped using a second order,
dual-pass Butterworth filter with a low-pass frequency cutoff of 2.5Hz 16]. EMG signals obtained during testing were then normalized to the highest smoothed
amplitude obtained during any of the MVICs performed for each muscle group, after
ensuring there were no non-physiological spikes in the data. Mean and peak normalized
EMG amplitudes were then calculated for each muscle during each cycle and data from
all cycles were used to calculate an average peak and average mean EMG signal for
each participant and trial (pre/post training). Using the Euclidean norm of the 3D
lumbar spine angles, local dynamic spine stability was calculated using the maximum
finite-time Lyapunov exponent (?
_max
) method, which is described in more detail in a previous publication 16].

Local dynamic stability values under both baseline and trained trials were normally
distributed according to Shapiro-Wilk testing and, therefore, values were compared
across trials using repeated-measures ANOVA in SPSS 22 (IBM Corporation, Armonk, NY,
USA). Conversely, the average peak and average mean EMG amplitudes were not normally
distributed for many muscles and, therefore, were compared between baseline and trained
trials using Friedman’s rank test. In all cases alpha was set to 0.05. To explore
the relationship between changes in muscle activation and changes in spine stability,
the absolute changes in both peak and mean muscle activation between baseline and
trained trials (trained %MVIC—baseline %MVIC) were determined for all muscles (these
data were normally distributed), and then added as covariates into the repeated-measures
ANOVAs. A stepwise removal approach was used whereby interactions between the main
effect (training) and covariates were removed if the p-value was greater than 0.2,
and the covariates themselves were removed if their p-values were greater than 0.2
and they did not interact significantly with the main effect of training. This was
done sequentially until all remaining interactions between the covariates and the
main effects, and all remaining covariates had significance levels (p-values) less
than 0.2. Moreover, to cross-validate the ANCOVA model results, Pearson’s correlations
were computed between the changes in both peak and mean muscle activation for all
muscles and the changes in local dynamic spine stability.