ARTICLE IN PRESS

Journal of Biomechanics 41 (2008) 2821–2828

Contents lists available at ScienceDirect

journal homepage: www.elsevier.com/locate/jbiomech

Journal of Biomechanics

0021-92

doi:10.1

� Corr

E-m

www.JBiomech.com

Neuromuscular neutral zones response to cyclic lumbar flexion

Deborah Solomonow a,b, Bradley Davidson a, Bing He Zhou a, Yun Lu a,Vikas Patel a, Moshe Solomonow a,�

a Musculoskeletal Disorders Research Laboratory, Bioengineering Division, Department of Orthopaedic Surgery, Health Sciences Center, University of Colorado, Denver,

12800 E. 19th Avenue, P.O. Box 6511, Aurora, CO 80045, USAb Division of Engineering, Colorado School of Mines, Golden, CO 80401,USA

a r t i c l e i n f o

Article history:

Accepted 10 July 2008The in vivo lumbar spine of the anaesthetized feline was subjected to passive cyclic anterior

flexion–extension at 0.25 Hz and 40 N peak load for cumulative 60 min duration. Displacement (or

Keywords:

Spine

Lumbar

Stability

Muscles

EMG

90/$ - see front matter & 2008 Elsevier Ltd. A

016/j.jbiomech.2008.07.010

esponding author. Tel.: +1303 724 0383; fax:

ail address: moshe.solomonow@uchsc.edu (M

a b s t r a c t

displacement neuromuscular neutral zones—DNNZ) and tension (or tension neuromuscular neutral

zones—TNNZ) at which reflexive EMG activity from the multifidi muscles was initiated and terminated

were recorded, for single-test cycles, before and for 7 h after cyclic loading. Displacement and tension

NNZs increased significantly after loading. The displacement NNZs decreased exponentially to near

baseline by the 7th hour of rest. The tension NNZs, however, decreased to below the baseline by the 2nd

to 3rd hour after loading and continued decreasing into the 7th hour. Peak EMG significantly decreased

(49–57%) to below the baseline immediately after loading and then exponentially increased, exceeding

the baseline by the 2nd to 3rd hour and reaching 33–59% above baseline by the 7th hour. EMG median

frequency decreased after loading and then exceeded the baseline after the 3rd hour, indicating initial

de-recruitment, followed by recruitment of new motor units. These findings suggest that the lumbar

spine was exposed to instability for 2–3 h after cyclic loading, due to concurrent laxity of the viscoelastic

tissues and deficient muscular activity. A delayed neuromuscular compensation mechanism was found

to exist, triggering the musculature significantly earlier and at higher magnitude than baseline, while

the viscoelastic tissues were still lax. Thus, it is suggested that prolonged cyclic loading may

compromise lumbar stability during the immediate 2–3 h post-loading, increasing the risk of injury.

& 2008 Elsevier Ltd. All rights reserved.

1. Introduction

Spinal stability is a complex issue, the full extent of which isnot fully delineated (Adams, 2007). One component of thestructural stability of the intervertebral joints is composed ofthe effects of the viscoelastic properties of the ligaments, discsand facet capsules. When vertebrae are displaced relative to eachother, the various viscoelastic tissues deform and generate tensionexpressed as resistance to destabilizing motion. This passiveproperty of the viscoelastic tissues is minimal for small perturba-tions about the neutral position and sharply increases for largerdisplacements. Panjabi (1996, 1992) designated the range of smallperturbation where the viscoelastic tissues are minimally engagedas neutral zones (NZ) within which the spine is inherently stable.

Another important component of spine stability is thecontribution of the muscles and their motor control (Panjabi,1996; Reeves et al., 2007; Adams, 2007). Recently, (Eversull et al.,2001; Solomonow et al., 2001) described the neuromuscular

ll rights reserved.

+1303 724 0394.

. Solomonow).

neutral zones (NNZ), where passive intervertebral motion above acertain displacement or a load above a certain magnitude, triggersthe reflexive activation of the musculature to preserve thestability (Stubbs et al., 1998; Solomonow et al., 1998). Under suchdefinition, initiation of muscular activity at a higher displacement(Displacement NNZ) or higher tension (Tension NNZ) than normalis indicative of a decrease or deficiency in stability, as moremotion is performed without the protection of the musculature.Conversely, decrease in the displacement or tension level, or lowerdisplacement or tension NNZ, is indicative of increased stability.To date, it was shown that the NNZ during flexion are smaller thanduring extension and that the NNZ display a gradual decreaseconcurrent with EMG increase as the velocity of spinal flexio-n–extension increases.

Cyclic/repetitive sports and occupational activities were shownto trigger high rates of musculoskeletal disorders when performedover long periods (Marras, 2000; Silverstein et al., 1986; Punnettet al., 1991; Hoogendoorn et al., 2000). The epidemiology wasrecently confirmed biomechanically and physiologically in in vivo

models (Navar et al., 2006; Le et al., 2007; Hoops et al., 2007) andin humans (Granata et al., 2005, 1999; Olson et al., 2004, 2006, inpress; Shin and Mirka, 2007; Li et al., 2007; Dickey et al., 2003;

ARTICLE IN PRESS

D. Solomonow et al. / Journal of Biomechanics 41 (2008) 2821–28282822

Karajcarski and Wells, 2006). Prolonged periods of exposure tocyclic lumbar loading were shown to develop substantial laxity/creep in the viscoelastic tissues and in turn, significant changes inthe activation pattern of the spinal musculature. A disorderconsisting of spasms, temporary attenuation of muscle activityfollowed by hyperexcitability was observed. It is expected thatcyclic activity of the lumbar spine may also elicit pronouncedchanges in the NNZ and elicit pronounced changes in spinalstability.

We hypothesize that 60 min of cyclic lumbar loading at amoderate load will cause significant enlargement of the NNZ,immediately after loading and that several hours of rest will berequired to restore normal NNZ. We further predict thatpronounced changes may be observed in the EMG amplitudeand its motor control when comparing it before and after cyclicloading.

Such information can afford new insights into the changes inthe motor control responsible for the stability of the lumbar spineafter cyclic work, the potential for injury and development of ameans for its prevention, as well as baseline data for design of safework scheduling.

2. Methods

2.1. Preparation

Seven adult cats, with an average weight of 3.9570.37 kg, were used in this

study. They were anesthetized with 60 mg/kg chloralose, according to a protocol

approved by the Institutional Animal Care and Use Committee. A superficial skin

incision overlying the lumbar spine was made to expose the dorso-lumbar fascia,

and an S-shaped stainless steel hook made of 1.5-mm-diameter rod was applied

around the supraspinous ligament between L4 and L5. The preparation was then

positioned in a rigid stainless steel frame and the lumbar spine was isolated by

means of two external fixators, which were applied to the L1 and L7 posterior

process, respectively. The external fixation was intended to limit the elicited

flexion to the lumbar spine and to prevent interaction of thoracic and sacral and/or

pelvic structures, but not to prevent any motion.

2.2. Instrumentation

Three pairs of stainless steel fine wire EMG electrodes (interelectrode distance:

3–4 mm) were inserted into the right L3–4, L4–5 and L5–6 multifidus muscles

L1 L2 L3 L

R1 R2 RN-1

P1 P2 P3

rp1 rp2 rp3

Load / Rest PeriodPre-Load

Rest Periods During Recovery: r0 = 10 min, r1 = 20

ri = 60 min for i = 3

(Note: Time Base not to Scale)

Rest Periods During Pre-Load: rp1, rp2, rp3 = 10 min

Loading Periods: L1, L2, ..., LN = 40 N @f for 10 mi

Rest Periods During Loading: R1, R2, ... RN-1 = 10

Single-CycleTests During Recovery Period: T1, T2,

f = 0.25 HzPre-Load Cycles: P1, P2, P3 = 40 N @f

# of Loading Periods: N = 6

Fig. 1. Loading sequence schema

6–8 mm laterally from the posterior spinal processes. A ground electrode was

inserted into the gluteus muscle. Each electrode pair constituted the input to a

differential EMG amplifier with a 110 dB common mode rejection ratio, a gain of up

to 200,000 and a band-pass filter in the range of 6–500 Hz. The EMG was recorded

with a sampling rate of 1000 Hz, and it was continuously monitored on an

oscilloscope. The S-shaped stainless steel hook inserted around the L4�5

supraspinous ligament was connected to the crosshead of the Bionix 858 material

testing system (MTS, Minneapolis, MN), in which a load cell was located. The load

was applied through the MTS actuator with a computer-controlled loading system

and monitored continuously along with the vertical displacement of the actuator.

The load cell and displacement outputs of the Bionix 858 MTS were also sampled

into the computer at 1000 Hz, along with the EMG signals.

Under such loading condition, the lumbar spine underwent anterior flex-

ion–extension while straining the supraspinous, intraspinous and posterior

longitudinal ligaments as well as the ligamentum flavum, facet capsule and

dorso-lumbar fascia. The discs in the lumbar spine were also deformed, expanding

in their dorsal aspects while narrowing in their ventral aspects during the flexion

phase. Overall, the neuromuscular response represents the reflexive activation of

several lumbar viscoelastic components similar to those active during flexion–

extension.

2.3. Protocol

A pre-load of 1 N was applied just prior to each single period of cyclic loading

in order to produce a standard baseline across all preparations. Initially, single

cycles of 40 N peak load at 0.25 Hz were applied with a 10 min rest at no load

between each cycle. This phase was performed to establish the normal NZ before

cyclic loading. Following, a set of six 10 min cyclic loading periods at 40 N peak,

each followed by 10 min rest was applied for a cumulative cyclic loading period of

60 min. The following recovery phase consisted of 7 h of rest at no load, during

which single-test cycles of 40 N peak load at 0.25 Hz were applied. The single

cycles were applied at 10, 30 and 60 min after the cyclic loading phase terminated,

and then once at every hour. Overall, 9 test cycles were applied during the 7 h

recovery period. Fig. 1 shows the schematic representation of the cyclic loading

with the single-cycle tests before and after. EMG, load and displacement were

recorded throughout the protocol.

2.4. Data analysis

The analysis considered the recorded EMG, vertical displacement and cyclic

load applied to the supraspinous ligament during the three pre-loading cycles and

the nine cycles during the recovery period. Each cycle was analyzed in a 5 s

window including 0.5 s before and 0.5 s after the 4 s cycle (0.25 Hz). EMG threshold

analysis was performed as follows: The first 500 ms of each EMG record, which

was before the initiation of movement by the actuator, was used as a benchmark

for baseline EMG signal level. The mean absolute value (MAV) of this baseline was

calculated: the signal recorded from each lumbar level was full wave rectified and

N

7 Hours Recovery Period

min, r2 = 30 min,

to 8

n

min

... Tn = 40 N @f

r0 r1 r2 r3 r4 r5 r8

T1 T2 T3 T4 T5 T6 T9

tic and protocol parameters.

ARTICLE IN PRESS

Fig. 2. Typical recording of raw EMG from L-3/4, L-4/5 and L-5/6, and corresponding tension and displacement vs. time during entire loading sequence.

D. Solomonow et al. / Journal of Biomechanics 41 (2008) 2821–2828 2823

smoothed with a 200 ms moving average filter (every 10 points) to yield the EMG

MAV. The filter was centered on the window under consideration in order to

prevent any time lags from affecting the data. After the initial 0.5 s period, any

point along the absolute value (full wave rectified) of the EMG during the given

cycle that exceeded three times the baseline MAV was denoted as the threshold of

activity in that channel. The corresponding tension and displacement values

during the stretch phase (increasing load) of the cycle were recorded as the tension

neuromuscular neutral zone (TNNZ) and displacement neuromuscular neutral

zone (DNNZ) DNNZ threshold values for that particular trial and lumbar level.

Similarly, as the EMG dropped below the three times baseline MAV threshold in

the release phase (decreasing load) of the cycle, the corresponding tension and

displacement were recorded as the TNNZ and DNNZ thresholds of cessation of

activity. This automatic procedure was visually supervised to ensure that any

unexpected signal artifacts were not detected as thresholds. Furthermore, in cases

where the EMG traces included spasms along the baseline, or the EMG discharge

during the flexion–extension was below three times the baseline (for example, see

Fig. 2, L-3/4 level, 110–170 min interval), the thresholds were determined visually.

The DNNZ and TNNZ from corresponding cycles during the recovery period for

each of the preparations were pooled across lumbar level, and the mean (7SD)

were calculated and plotted as a function of time. The DNNZ and TNNZ of the three

cycles applied before cyclic loading was initiated were averaged for each

preparation to yield a mean and SD, and then pooled with those of the other

preparations to use as the pre-loading NNZ baselines.

The peak MAV (PMAV) of the first three cycles recorded before cyclic loading

was used as a normalization value for the PMAV recorded in the test cycles of the

same preparation during the 7 h recovery period. A single-peak value was

determined for each cycle. The normalized values of all the preparations were

pooled, and the mean (7SD) was calculated and plotted as normalized PMAV vs.

time (Fig. 6) to assess whether the contraction level increased or decreased

because of cyclic loading.

EMG median frequency (MF) was found for each single-cycle test in order to

identify changes in motor unit recruitment (Solomonow et al., 1990). A 500 ms

window, centered at the peak load of each cycle, within which EMG was

approximated as a stationary signal, was zero-padded on both sides. A Tukey

window was applied to the zero-padded data, the power spectral density of the

signal in this window was found via the fast Fourier transform, and the MF, defined

as the frequency that divides the area under the power spectral density in half, was

calculated. The MF values of all corresponding time points and lumbar levels were

pooled and plotted as mean MF vs. recovery time with the mean of the pre-loading

MF as baseline.

2.5. Statistics

Since multifidus segments are not innervated from multiple roots (MacIntosh

et al., 1986), the data was divided into three lumbar levels (L-3/4, L-4/5, L-5/6) for

statistical analyses. The DNNZ, TNNZ and PMAV data were visually inspected for

normality. If a distribution did not appear normal, an appropriate data

transformation was applied. Two-way repeated measures ANOVAs were used to

test for differences in the stretch and relaxation phases and changes over time of

the DNNZ and TNNZ. The independent variables included time (pre-cyclic loading,

recovery times) and loading phase (stretch, relaxation). All higher order factorial

terms were included in the statistical models to test for interaction of the

independent variables. One-way repeated measures ANOVAs were used to test for

changes over time in the EMG-based variables. The independent variables included

time (pre-cyclic loading, recovery times) and the dependent variables were

normalized PMAV and median frequency. Upon determining a significant

interaction or main effect, pair-wise comparisons were performed using a

Student’s t -test. Level of significance for all tests was set as p ¼ 0.05.

2.6. Modeling

The mean7SD values of the DNNZ, TNNZ, PMAV and MF during recovery for

each lumbar level were fit with exponential-based models, as they represent the

classical response of viscoelastic tissues (Solomonow et al., 2000).

The time-course of the DNNZ thresholds during the stretch phase and

relaxation phase of the test cycles during the recovery period were described by

DISPðtÞ ¼ D0 þ ðt � trÞDLðe�ðt�tr=t1 ÞÞ þ DMðe

�ðt�tr=t2 ÞÞ ftj120ptp530g (1)

where D0 is the intercept of the displacement (mm), DL affects the rise amplitude

(mm/s), DM is the amplitude of the decay dominating the end of the recovery

period (mm), (t�tr)DL(exp(�(t�tr/t1))) allows for a transient rise at the beginning

of the recovery period; t1 affects the rates of rise and fall (s); t2 is the exponential

time constant of the decay that dominates the end of the recovery period (min).

As noted above, the model was evaluated from the time point of 120–530 min,

which constitutes the 7 h of recovery post-static loading.

The time-course of the TNNZ thresholds during the stretch phase and

relaxation phase of the test cycles during the recovery period were described by

TENðtÞ ¼ T0 þ ðt � trÞTLðe�ðt�tr=t3 ÞÞ þ TMðe

�ðt�tr=t4 ÞÞ ftj120ptp530g (2)

where T0 is the intercept of the tension (N); TL affects the rise amplitude (N/s); TM

is the amplitude of the decay dominating the end of the recovery period(N);

(t�tr)TL(exp(�(t�tr/t3))) allows for a transient rise at the beginning of the

recovery period; t3 affects the rates of rise and fall (s); t4 is the exponential time

constant of the decay that dominates the end of the recovery period (min).

The time course of the PMAV during the recovery period were described by

PMAVðtÞ ¼

P0 þ PLðe�ðt�tr=t5ÞÞ þ PMð1� e�ðt�tr=t6 ÞÞ; totd

P0 þ PLðe�ðt�tr=t5ÞÞ þ PMð1� e�ðt�tr=t6 ÞÞ

þðt � tdÞPHðe�ðt�td=t7 ÞÞ; tXtd

ftj120ptp530g

8><>:

(3)

where P0 is the intercept of the PMAV; PL is the amplitude of the exponential

decay; PM is the amplitude of the exponential increase; t is time measured since

the beginning of the experiment (min); t5, t6 are exponential time constants

(min); tr is the time of the first recovery measurement (120 min);

(t�td)PHexp(�(t�td/t7)) is the hyperexcitability term. This term has a delayed

onset during the recovery period and is equal to zero when totd.

The model for the time course of the MF data was described by the equations

below:

MFðtÞ ¼

F0 þ FLðe�ðt�tr=t8 ÞÞ þ FMð1� e�ðt�tr=t9 ÞÞ; totd

F0 þ FLðe�ðt�tr=t8 ÞÞ þ FMð1� e�ðt�tr=t9 ÞÞ

þðt � tdÞFHðe�ðt�td=t10 ÞÞ; tXtd

ftj120ptp530g

8><>:

(4)

ARTICLE IN PRESS

D. Solomonow et al. / Journal of Biomechanics 41 (2008) 2821–28282824

where F0 is the intercept of the peak MF (Hz); FL is the amplitude of exponential

decay dominating beginning of recovery period (Hz); FM is the amplitude of

exponential increase following decay in beginning of recovery period (Hz); tr is the

time of first recovery measurement (120 min); td is the time of onset of

hyperexcitability (min); t8 is the exponential time constant of exponential decay

dominating beginning of recovery period (min); t9 is the exponential time

constant of exponential increase following decay in beginning of recovery period

(min); t10 is the exponential time constant of hyperexcitability term dominating

end of recovery period (min); FL(exp(�(t�tr/t8))) Allows for exponential decay

dominating beginning of recovery period; FM(1�exp(�(t�tr/t9))) Allows for

exponential increase following decay in beginning of recovery period;

(t�td)FH(exp(�(t�td/t10))). Hyperexcitability term with delayed onset dominating

end of recovery period (equal to zero when toTd).

Levenberg–Marquardt nonlinear regression algorithms were used to generate

the best-fit models, optimizing for the regression coefficient.

Fig. 3. Typical recording of EMG from L-3/4, L-4/5 and L-5/6 with superimposed

MAV and arrows indicating initiation and cessation NNZ thresholds, and

corresponding tension and displacement vs. time during single loading cycle.

Fig. 4. Pooled DNNZ threshold data, separated by lumbar level, represented by mean7SD

empirical models. Data points at time zero represent baseline values from the average

Indicates a significant difference from baseline at the respective data point along the t

3. Results

A typical recording of the raw EMG from L-3/4, L-4/5 and L-5/6,and the associated tension and displacement before, during andafter the cyclic loading period is shown in Fig. 2. The sameparameters corresponding to a typical single cycle are given inFig. 3 with the arrows indicating the initiation and cessation ofEMG, and the corresponding displacement and tension.

3.1. Displacement neuromuscular neutral zones (DNNZ)

Fig. 4 displays the mean7SD of the DNNZ before and after thecyclic loading. The DNNZ demonstrated changes with time(Po0.001 for all levels) and also loading phase (Po0.001 for alllevels), and no time� loading phase interaction (L-3/4: P ¼ 0.990,L-4/5: P ¼ 0.937, L-5/6: P ¼ 0.863). DNNZ for both phasesremained significantly elevated through 4 h after cyclic loadingfor levels L-3/4 and L-4/5, and through 3 h after cyclic loading forlevel L-5/6 (see Fig. 4). The mean of the baseline DNNZ during thestretch phase were 2.7, 2.1 and 2.7 mm for the L-3/4, L-4/5 andL-5/6 lumbar levels, respectively. These values increased overthreefold to 9.8, 9.2 and 9.3 mm immediately after the cyclicloading period. The final mean DNNZ at the end of the 7 hrecovery period were 2.6, 2.8 and 2.5 mm for the three lumbarlevels, respectively.

The mean of the baseline DNNZ during the relaxation phasewere 6.1, 5.6 and 7.1 for the three lumbar levels, respectively. TheDNNZ increase of roughly twofold to 12.7, 12.4 and 12.8 mmfollowing the cyclic loading. A gradual decrease to baseline overthe 7 h recovery was observed to final DNNZ of 6.3, 6.0 and6.0 mm for the three lumbar levels, respectively.

For each lumbar level, the DNNZ during stretch weresignificantly smaller than their counterpart during relaxation.Differences between the loading phases were discussed in detailin a previous paper and will not be duplicated here (Solomonowet al., 2001).

3.2. Tension neuromuscular neutral zones (TNNZ)

The mean TNNZ before and after the cyclic loading is shown inFig. 5. The TNNZ demonstrated changes with time (Po0.001 for all

of DNNZ thresholds vs. time for stretch and relaxation phases, with superimposed

of three pre-loading test cycles. Subsequent points are during recovery period. (*)

ime axis for both stretch and relaxation phases.

ARTICLE IN PRESS

D. Solomonow et al. / Journal of Biomechanics 41 (2008) 2821–2828 2825

levels) and also loading phase (Po0.001 for all levels), and notime� loading phase interaction (L-3/4: P ¼ 0.895, L-4/5:P ¼ 0.984, L-5/6: P ¼ 0.910). TNNZ for both phases remainedsignificantly elevated through 2 h after cyclic loading in levelL-3/4, and through 1 h after cyclic loading in levels L-4/5 andL-5/6. The TNNZ significantly decreased to below baselinevalues in the 6th and 7th hours in level L-3/4, and in the 5th–7thhours in levels L-4/5 and L-5/6 (see Fig. 5). The mean of thebaseline TNNZ during the stretch phase were 8.8, 7.0 and 9.1 N forthe three lumbar levels, respectively. The TNNZ increased to 19.0,15.6 and 16.9 N immediately after cyclic loading and graduallydecreased to reach 3.3, 3.9 and 3.2 N at the end of the 7th hour ofrecovery.

The mean of the baseline TNNZ during the relaxation phasewas 12.4, 11.1 and 18.3 N for the three lumbar levels, respectively.The TNNZ increased by nearly twofold to 25.7, 25.3 and 29.5 Nfollowing cyclic loading. The mean TNNZ gradually declined overthe recovery period, reaching 8.4, 7.3 and 9.1 N at the end of the7th hour of recovery.

For each lumbar level, the TNNZ during stretch weresignificantly smaller than their counterpart during relaxation.Differences between the loading phases were discussed in detail

Fig. 5. Pooled TNNZ threshold data, separated by lumbar level, represented by mean7SD

empirical models. Data points at time zero represent baseline values from averaged three

significant difference from baseline at the respective data point along the time axis for

Fig. 6. Pooled PMAV data, separated by lumbar level, represented by mean7SD of PMAV

baseline values from averaged three pre-loading test cycles. Subsequent points are d

respective data point along the time axis.

in a previous paper and will not be duplicated here (Solomonowet al., 2001).

3.3. EMG peak mean absolute value (PMAV)

Fig. 6 shows the normalized PMAV before and after the cyclicloading. The PMAV demonstrated a significant change over timefor each lumbar level (Po0.0001). A significant decrease of 49%,57% and 51% was evident immediately after cyclic loading thatremained below the baseline values for one hour during recovery(see Fig. 6). The PMAV exceeded baseline levels after the thirdhour for the L-3/4 level, about the second hour for the L-4/5 level,and just before the third hour for the L-5/6 level. The L-5/6 lumbarlevel was significantly higher in the 5th–7th hours duringrecovery. The PMAV reached 1.14, 1.45 and 1.59, or 14–59% abovebaseline, at the 7th hour.

3.4. EMG median frequency (MF)

Fig. 7 shows the MF before and after the cyclic loading. The MFdemonstrated a significant change over time (Po0.0001) for each

of TNNZ thresholds vs. time for stretch and relaxation phases, with superimposed

pre-loading test cycles. Subsequent points are during recovery period. (*) Indicates

both stretch and relaxation phases.

vs. time, with superimposed empirical models. Data points at time zero represent

uring recovery period. (*) Indicates a significant difference from baseline at the

ARTICLE IN PRESS

Fig. 7. The pooled median frequency (MF) mean7SD for the three lumbar levels before the cyclic loading and during the recovery period. (*) Indicates significant difference

from baseline at the respective data point along the time axis.

D. Solomonow et al. / Journal of Biomechanics 41 (2008) 2821–28282826

lumbar level. The MF decreased as the site of measurement moveddown the lumbar levels (from L-3/4 to L-5/6). The baselinemean7SD MF values for the L-3/4, L-4/5 and L-5/6 lumbarlevels were 218.10717.81, 200.71712.63 and 190.10714.89 Hz,respectively. Following cyclic loading, the MF values decreasedfrom baseline values to 215.40724.07, 194.75713.14 and183.04714.16 Hz. The MF for L-3/4 and L-5/6 continued todecrease exponentially until 1 h into the recovery period andbecame significantly lower than baseline 208.43714.94 and177.46716.24 Hz, respectively. The L-4/5 lumbar level MF didnot clearly show the exponential decrease in the beginning of therecovery period, as it might have been masked by the largestandard deviations. The average percent difference from baselineat the first hour of recovery was 4.4%. Thereafter, the MF increasedto baseline values by the 4th hour for L-3/4, the 3rd hour for L-4/5and after the 2nd hour for L-5/6, after which it continued toexponentially increase above the baseline. In the 5th–7th hours ofrecovery, MF for the L-4/5 and L-5/6 lumbar levels weresignificantly higher than baseline being 213.8775.69, and209.26719.64 Hz, respectively (see Fig. 7).

3.5. Models

The models derived for the DNNZ, TNNZ and PMAV are shownsuperimposed on their corresponding data in Figs. 4–7. Theregression coefficient (r2) for the DNNZ models ranged between0.9583 and 0.9993, for the TNNZ between 0.9737 and 0.9991, forthe PMAV between 0.9337 and 0.9738 and for the MF between9629 and 9985.

3.6. Statistics

A square root transformation was applied to the TNNZ andPMAV data for all lumbar levels to obtain a normal distribution.All figures presented in the text were composed using theuntransformed data.

4. Discussion

Recalling that increasing NNZ is indicative of decreasedstability, the major findings of this investigation consist of theobservations of profound changes in the neuromuscular neutralzones (NNZ) and the motor control of muscular activity as a direct

response to cyclic loading. In specific, 3 major issues are ofimportance: cyclic loading elicited a period of 2–3 h post-loadingduring which there was a significant increase in the NNZ andexposure to instability. During the same 2–3 h post-loading, asignificant decrease in the magnitude of muscular activity furthercompromised the stability of the spine. A new compensatoryneural mechanism emerged. This mechanism is capable ofallowing muscular activity to trigger significantly earlier andprovide more force while the DNNZ are still compromised and notfully recovered together with the laxity in the viscoelastic tissue.

Overall, the first 2–3 h post-cyclic loading are associated withsubstantial increase in exposure to spinal instability due to laxityin the ligaments together with significant impairment in muscularactivity.

The original activation of the multifidus muscles was shown tobe a direct reflexive response to stretch of the ligaments and otherviscoelastic tissues of the lumbar spine (Stubbs et al., 1998;Solomonow et al., 1998). In essence, elongation or loads above acertain threshold triggered reflexive muscular activity thatstiffened the spine. It was further demonstrated that as tension–relaxation or creep develop in the spine over a period of flexion orloading, respectively, the trigger threshold for the reflex shiftssubstantially (Solomonow et al., 1999). The observed increase inthe TNNZ or DNNZ after the cyclic loading period was, therefore,the manifestation of the creep that developed in the viscoelasictissue. Similarly, the reduction in the PMAV was observed beforeand is a typical response to development of creep in the tissuesduring cyclic loading (Solomonow et al., 1998; Hoops et al., 2007;Le et al., 2007).

While the DNNZ approximates the creep recovery over the 7 hrest period, significant differences between these variables exist.The models for the DNNZ recovery predict a time constant in therange of 108.5–276 min while the time constants of the creeprecovery were about 50 min (Le et al., 2007). The two- to fivefoldlarger time-constants for the DNNZ emerge from the fact that thetrue creep is masked by intervertebral stiffening as a result of themuscular contractions. The measured creep reflects the stiffeningwhereas the DNNZ reflects the actual displacement wheremuscular activity was triggered. While one can suggest that thecreep is related to the DNNZ, it cannot be relied upon toapproximate the displacement NZ, as an artificial fast rate ofrecovery may be obtained.

It was expected that as the creep in the tissues recovers overthe 7 h of rest, the TNNZ, DNNZ, PMAV and the MF will return tonormal. The DNNZ indeed returned to near normal level at the 7th

ARTICLE IN PRESS

D. Solomonow et al. / Journal of Biomechanics 41 (2008) 2821–2828 2827

hour of recovery, yet the behavior of the TNNZ, PMAV and the MFwas completely different. Muscular activity was triggered at lowerthan baseline tension and at a higher level of activation after 2–3 hof rest. The TNNZ continued to decrease below baseline and thePMAV continued to increase above baseline with rest, reachingsaturation near the 6th–7th hour. The MF pattern followed that ofthe PMAV, increasing after the first 2–3 h of recovery andindicating that recruitment of larger motor units is taking place.

The models developed expose the link between the TNNZ,PMAV and MF: activation by the same source. The dominant timeconstant from mid to end of recovery of the three variables werenearly overlapping; TNNZ, T4, ranged from 100 to 280 min, T7 forthe PMAV ranged from 170 to 400 min and T10 of the MF rangedfrom 171 to 321 min. Essentially, the motor control activation ofthe compensatory mechanism recruited additional, larger motorunits after the 3rd hour of recovery, increasing the averageconduction velocity and therefore the MF of the EMG. This wasmanifested by the corresponding increase of the EMG amplitude,expressed by the PMAV, as well as by the corresponding earlyincrease in the tension, reflected by the TNNZ. Further confirma-tion that a single compensatory motor control triggered the TNNZ,PMAV and MF could be obtained from the Td associated with thecorresponding models. The Td values ranged from 224 to 294 min,indicating that the compensatory motor control was initiated andaffected all the above parameter simultaneously. Furthermore, thebehavior of the TNNZ, PMAV and MF trends could not beexplained by a simple ligamento-muscular reflex, as the viscoe-lastic tissues were still substantially lax by the 2nd or 3rd and upto the 7th hour of recovery. Increasing PMAV and MF withdecreasing TNNZ would require tightening of the viscoelastictissues above their baseline to elicit such an increase as a reflexiveresponse. Based on the observations above, it is becomingapparent that a different, compensatory neural control mode isactivated 2–3 h after cyclic loading.

The different neural control modes could be associated withthe clinical finding that tissue damage and the associated painresults in spasms, elevated muscular activity and joint stiffness(Pedersen et al., 1956; van Dieen et al., 2003). The work of Wooet al. (1981, 1982) demonstrated that cyclic creep of viscoelastictissues within the physiologic range is associated with micro-damage in the collagen fibrils. The viscoelastic tissue could beconsidered damaged in this experiment as it was exposed to60 min of cumulative cyclic loading. This neural control mode,therefore, may be triggered by the tissue damage and theassociated pain mechanism.

The micro-damage in the viscoelastic tissues could beclassified as sub-clinical for such physiological loads and dis-placements, yet was shown in humans to result in stiffening of thespine for several hours after the work was completed (Granataand Marras, 2000). Similarly, Olson et al. (2004, 2006), Li et al.(2007), Shin and Mirka (2007), Granata et al. (2005), Dickey et al.(2003) found that such significant changes in muscular activityoccur after moderate and mild cyclic loading in humans. Inessence, these complementary findings validate that the observa-tions made in this project, using an in vivo cat model, are also seenin humans subjected to similar loading conditions, and that mildloading dose duration can trigger such muscular responses.

In conclusion, a period of cyclic loading significantly increasedthe tension and displacement NZ while decreasing peak muscularactivity in the 2–3 h immediately after the work. This suggeststhat the lumbar spine is exposed to significant reduction instability control in that period. A compensatory neural controlmechanism is triggered by the 3rd hour post-cycling loading, andsignificantly enhanced the magnitude and timing of the muscularcontributions while allowing the viscoelastic tissues to recoverfrom the imposed creep for several hours longer.

Conflict of interest statement

There is no conflict of interest regarding the publication of thispaper.

Acknowledgment

This work was supported by Grant R01-OH-007622 from theNational Institute of Occupational Safety and Health.

References

Adams, M., 2007. Spine stability: the six blind men and the elephant. ClinicalBiomechanics 22, 486.

Dickey, J., McNorton, S., Potvin, J., 2003. Repeated spinal flexion modulates theflexion-relaxation phenomenon. Clinical Biomechanics 18, 783–789.

Eversull, E., Solomonow, M., Bing He Zhou, E.E., et al., 2001. Neuromuscular neutralzones sensitivity to lumbar displacement rate. Clinical Biomechanics 16,102–113.

Granata, K.P., Marras, W.S., 2000. Cost-benefit of muscle co-contraction inprotecting against spinal instability. Spine 25, 1398–1404.

Granata, K., Marras, W., Davis, K., 1999. Variation in spinal load and trunk dynamicsduring repeated lifting exertions. Clinical Biomechanics 14, 367–375.

Granata, K.P., Rogers, E., Moorhouse, K., 2005. Effects of static flexion-relaxation onparaspinal reflex behavior. Clinical Biomechanics 20, 16–24.

Hoogendoorn, W.E., Bongers, P.M., de Vet, H.C., et al., 2000. Flexion and rotation ofthe trunk and lifting at work are risk factors for low back pain: results of aprospective cohort study. Spine 25, 3087–3092.

Hoops, H., Zhou, B., Lu, Y., Solomonow, M., Patel, V., 2007. Short rest between cyclicflexion periods is a risk factor for lumbar disorder. Clinical Biomechanics 22,745–757.

Karajcarski, S., Wells, R., 2006. The time variation pattern of mechanical exposureand the reporting of low back pain. Theoretical Issues in Ergonomic Science,1–27.

Le, P., Solomonow, M., Zhou, B., Lu, Y., Patel, V., 2007. Cyclic load magnitude is a riskfactor for accumulative low back disorder. Journal of Occupational andEnvironmental Medicine 49, 375–387.

Li, L., Patel, N., Solomonow, D., Le, P., Hoops, H., Gerhardt, D., Johnson, K., Zhou, B.,Lu, Y., Solomonow, M., 2007. Neuromuscular response to cyclic lumbartwisting. Human Factors 49, 820–829.

MacIntosh, J., Valencia, F., Bogduk, N., Munro, R., 1986. The morphology of thehuman lumbar multifidus. Clinical Biomechanics 1, 196–204.

Marras, W.S., 2000. Occupational low back disorder causation and control.Ergonomics 43, 880–902.

Navar, D., Zhou, B., Lu, Y., Solomonow, M., 2006. High repetition of cyclic loading isa risk factor for lumbar disorders. Muscle and Nerve 34, 614–622.

Olson, M.W., Li, L., Solomonow, M., Interaction of viscoelastic tissue compliancewith lumbar muscles during passive cyclic flexion–extension. Journal ofElectromyography and Kinesiology, in press.

Olson, M.W., Li, L., Solomonow, M., 2004. Flexion-relaxation response to cycliclumbar flexion. Clinical Biomechanics 19, 769–776.

Olson, M., Solomonow, M., Li, L., 2006. Flexion-relaxation response to gravity.Journal of Biomechanics 39, 2545–2554.

Panjabi, M.M., 1992. The stabilizing system of the spine. 2. Neutral zone andinstability hypothesis. Journal of Spinal Disorders 5, 390–397.

Panjabi, M.M., 1996. Low back pain and spinal stability. In: Weinstein, J.N., Gordon,S.L. (Eds.), Low Back Pain: A Scientific and Clinical Overview. AmericanAcademy of Orthopedic Surgeons, Rosemont, IL, pp. 367–384.

Pedersen, H., Blunk, C., Gardner, E., 1956. The anatomy of lumbosacral posteriorrami and meningeal branches of spinal nerves. Journal of Bone and JointSurgery(A) 38, 377–391.

Punnett, L., Fine, L.J., Keyserling, W.M., et al., 1991. Back disorders and nonneutraltrunk postures of automobile assembly workers. Scandinavian Journal of Work,Environment and Health 17, 337–346.

Reeves, N.P., Narendra, K.S., Cholewicki, J., 2007. Spine stability: the six blind menand the elephant. Clinical Biomechanics 22, 266–274.

Shin, G., Mirka, G., 2007. An in-vivo assessment of the low back response toprolonged flexion: interplay between active and passive tissues. ClinicalBiomechanics 22, 965–971.

Silverstein, B.A., Fine, L.J., Armstrong, T.J., 1986. Hand wrist cumulative traumadisorders in industry. British Journal of Industrial Medicine 43,779–784.

Solomonow, M., Baten, C., Smit, J., Baratta, R.V., Hermens, H., D’Ambrosia, R., Shoji,H., 1990. EMG power spectra frequencies associated with motor unitrecruitment strategies. J Appl Physiology 68, 1177–1185.

Solomonow, M., Zhou, B., Harris, M., Lu, Y., Baratta, R., 1998. The ligament–mus-cular stabilizing system of the spine. Spine 23, 2552–2562.

Solomonow, M., Zhou, B.H., Baratta, R.V., et al., 1999. Biomechanics of increasedexposure to lumbar injury caused by cyclic loading: Part 1. Loss of reflexivemuscular stabilization. Spine 24, 2426–2434.

ARTICLE IN PRESS

D. Solomonow et al. / Journal of Biomechanics 41 (2008) 2821–28282828

Solomonow, M., Zhou, B., Baratta, R., Lu, Y., Zhu, M., Harris, M., 2000. Bi-exponentialrecovery model of lumbar viscoelastic laxity and reflexive muscular activity afterprolonged cyclic loading. Clinical Biomechanics 15, 167–175.

Solomonow, M., Eversull, E., He Zhou, B., et al., 2001. Neuromuscular neutral zonesassociated with viscoelastic hysteresis during cyclic lumbar flexion. Spine 26,E314–E324.

Stubbs, M., Harris, M., Solomonow, M., Zhou, B., Lu, Y., Baratta, R., 1998. Theligamento-muscular protective reflex in the lumbar Spine. Journal of Electro-myography and Kinesiology 8, 197–204.

van Dieen, J.H., Selen, L.P., Cholewicki, J., 2003. Trunk muscle activation in low-backpain patients, an analysis of the literature. Journal of Electromyography andKinesiology 13, 333–351.

Woo, S., Gomez, M., Amiel, D., Akeson, W., 1981. The effect of exercise on thebiomechanical and biochemical properties of swine digital flexor tendon.Journal of Biomechanical Engineering 103, 51–56.

Woo, S., Gomez, M., Woo, Y., Akeson, W., 1982. Mechanical properties of tendonsand ligaments; the relationships of exercise in tissue remodeling. Biorheology19, 379–408.