Professional Documents
Culture Documents
The Temperature-Dependent Viscoelasticity of Porcine Lumbar Spine Ligaments
The Temperature-Dependent Viscoelasticity of Porcine Lumbar Spine Ligaments
Study Design. A uniaxial tensile loading study of 13 developing material models for in vivo force response,
lumbar porcine ligaments under varying environmental and further suggests that previously published material
temperature conditions. property values derived from room temperature testing
Objectives. To investigate a possible temperature de- may not adequately represent in vivo response. These
pendence of the material behavior of porcine lumbar an- findings have clinical relevance in the increased suscep-
terior longitudinal ligaments. tibility of ligamentous injury in the cold and in assessing
Summary of Background Data. Temperature depen- the mechanical behavior of cold extremities and extrem-
dence of the mechanical material properties of ligament ities with limited vascular perfusion such as those of the
has not been conclusively established. elderly.
Methods. The anterior longitudinal ligaments (ALLs) Key words: temperature, viscoelastic, lumbar, liga-
from domestic pigs (n ⫽ 5) were loaded in tension to 20% ments, porcine. Spine 2007;32:E436 –E442
strain using a protocol that included fast ramp/hold and
sinusoidal tests. These ligaments were tested at temper-
atures of 37.8°C, 29.4°C, 21.1°C, 12.8°C, and 4.4°C. The Mechanical testing of human and representative nonhu-
temperatures were controlled to within 0.6°C, and liga- man biologic materials, including lumbar spinal liga-
ment hydration was maintained with a humidifier inside ments, has historically been performed at room temper-
the test chamber and by spraying 0.9% saline onto the
ligament. A viscoelastic model was used to characterize
ature. However, human ligamentous tissues operate at
the force response of the ligaments. body temperature. This study was performed to investi-
Results. The testing indicated that the ALL has strong gate the primary hypothesis that the mechanical material
temperature dependence. As temperature decreased, the response of ligamentous biologic material from the lum-
peak forces increased for similar input peak strains and bar spine is temperature dependent and that this behav-
strain rates. The relaxation of the ligaments was similar at
each temperature and showed only weak temperature
ior may be characterized using a viscoelastic material
dependence. Predicted behavior using the viscoelastic model.1 The implications of a strong temperature depen-
model compared well with the actual data (R2 values dence of lumbar spinal material properties are signifi-
ranging from 0.89 to 0.99). A regression analysis per- cant. In a number of previous investigations, material
formed on the viscoelastic model coefficients confirmed models of human and animal tissue have been incorpo-
that relaxation coefficients were only weakly temperature
dependent while the instantaneous elastic function coef-
rated into computer models for human response in au-
ficients were strongly temperature dependent. tomobile crashes and other potentially lethal events.2 In
Conclusions. The experiment demonstrated that the such models, the global human response is dependent on
viscoelastic mechanical response of the porcine ligament local mechanical material properties of tissues, including
is dependent on the temperature at which it is tested; the ligamentous tissue. If ligamentous response is strongly
force response of the ligament increased as the temper-
ature decreased. This conclusion also applies to human
temperature dependent, the resulting computational
ligaments owing to material and structural similarity. This models may be inaccurate and result in an increased risk
result settles a controversy on the temperature depen- of injury. Automobile crashes and similarly high acceler-
dence of ligament in the available literature. The ligament ation/deceleration events typically have high strain rates;
viscoelastic model shows a significant temperature de- therefore, the ligaments in this study were strained to
pendence on the material properties; instantaneous elas-
tic force was clearly temperature dependent while the
20% of their initial length at an average rate of 78 ⫾ 10
relaxation response was only weakly temperature depen- seconds⫺1.
dent. This result suggests that temperature dependence The temperature dependence of ligaments has not
should be considered when testing ligaments and been established. Dorlot et al tested the temperature de-
pendence of the dog anterior cruciate ligament by per-
forming tensile tests at temperatures ranging from 23°C
From the *Center for Applied Biomechanics, University of Virginia, to 45°C over an extended period.3 Dorlot et al concluded
Charlottesville, VA; and †NAVAIR, Patuxent River, MD.
Acknowledgment date: August 24, 2006. First revision date: February that temperature did not have a significant effect on the
22, 2007. Acceptance date: February 23, 2007. load-elongation curves of the ligaments, although there
The manuscript submitted does not contain information about medical is little information on the testing performed. Later, Has-
device(s)/drug(s).
Federal funds were received in support of this work. No benefits in any berry and Pearcy studied the lumbar interspinous liga-
form have been or will be received from a commercial party related ments of sheep by performing extension tests at temper-
directly or indirectly to the subject of this manuscript. atures ranging from 19°C to 46°C.4 They concluded that
Address correspondence and reprint requests to Robert S. Salzar, PhD,
Center for Applied Biomechanics, University of Virginia, 1011 Linden the mechanical properties of ligaments are not affected
Avenue, Charlottesville, VA 22902; E-mail: Salzar@virginia.edu by temperature, between room temperature and body
E436
Temperture-Dependent Porcine Lumbar Ligaments • Bass et al E437
temperature. However, the peak forces of the ligaments goal of this study was to determine the temperature-
decreased as the ligament temperature was increased dependent high-rate viscoelastic mechanical properties
from room temperature to body temperature. In con- of lumbar spinal ligaments.
trast, Woo et al investigated the temperature dependence
of dog medial collateral ligaments (MCLs). Noninjuri- Materials and Methods
ous cyclic extension tests were performed at tempera-
tures ranging from 2°C to 37°C. with 1 to 2 mm of Spines from 5 domestic pigs (Sus scrofa: 4 to 4.5 months old
with a mean live weight of 82.6 kg) were harvested from freshly
extension (3.5% maximum strain) at a strain rate of 0.33
killed specimens and were cryogenically frozen. Woo et al
seconds⫺1.5 Woo et al5 found that the ligaments were showed that postmortem freezing of rabbit MCLs has little
temperature dependent and that the tensile load was in- effect on the biomechanical properties of ligaments.9 Immedi-
versely proportional to the set ligament strain level with ately before preparation, the spine was allowed to thaw to
respect to temperature. They also found that the peak room temperature. The lumbar region of the porcine spine was
forces displayed relaxation during cyclic loading. The transected into functional spinal units (FSU) at 3 levels: T15–
relaxation was shown to be temperature dependent as L1, L2–L3, and L4 –L5. The anterior longitudinal ligament
the relaxation of the ligament decreased when the tem- (ALL) was then isolated from the FSUs by cutting through the
perature was decreased. An additional study was per- vertebral body along the coronal plane. Hydration of the ALL
formed by Lam et al6 using rabbit MCL. Tests were was maintained by placing a 0.9% saline soaked gauze on the
performed in cyclic extension for temperatures ranging ligament throughout the preparation process.
Each vertebral body was potted, using a fast curing resin
between 25°C and 55°C with a peak displacement of
(R1 Fast Cast, Goldenwest Manufacturing, Cedar Ridge, OH)
0.68 mm and a strain rate of 0.167 seconds⫺1. These in a separate metallic cup with the ALL spanning the distance
ligaments were found to be temperature dependent. In between the 2 cups. To provide hard anchor points for the
agreement with Woo et al,5 Lam et al6 found the liga- mounting procedure, small screws were inserted into the pos-
ments to be temperature dependent and that force relax- terior aspect of each vertebral body. The potted ligament was
ation for each cycle increased with temperature. mounted in a force test machine (Instron, Inc. #8874, Canton,
Although these tests were performed on porcine liga- MA) for uniaxial extension tests; the ligament was oriented so
ments, previous studies have related the mechanical be- that the vertebral endplates were parallel, the vertebral bodies
havior of human longitudinal ligaments with porcine were aligned so that they were parallel, and the ALL was
specimens. Hukins et al presented porcine ligament stiff- aligned so that it was perpendicular to the vertebral endplates.
ness, stress, and failure results, with evidence that human To warm the ligaments to the initial test temperature, the
ligaments, in a sealed plastic bag, were first placed in a temper-
longitudinal ligaments behave in a similar manner.7 This
ature controlled bath set to 37.8°C. The intervertebral disc was
study notes that the substructure of porcine longitudinal then carefully transected to isolate the ligament. The potted
ligaments is very similar to human tissue. They also note ligament was then mounted to the test fixture and pretensioned
that the modes of ligament failure are identical in both to a 2 N load. This procedure establishes a repeatable starting
human and porcine specimens. In addition, this study point for each ligament and is similar to that used by Lucas.8
demonstrated that both human and porcine longitudinal After the preload, the initial length of the ligament was mea-
ligaments store energy in a similar manner by comparing sured between the inferior and superior endplates of the verte-
hysteresis loops. bral bodies.
Other studies have found porcine ligaments to be a The ligaments were tested in an environmental chamber
good model for human ligamentous tissue. In a previous capable of maintaining humidity at ⬃100% RH and tempera-
study, it has been shown that porcine ligamentous tissue ture between ⫺30°C and 45°C (Figure 1). The temperature
inside the chamber was monitored using “K” type thermocou-
behaves similarly to human ligamentous tissue both
ples (OMEGA Engineering, Inc., Stamford, CT) at 2 locations
qualitatively as well as quantitatively, as long as the inside the chamber: one in the ambient air and another on the
properties are scaled by vertebral depth and width. In a surface of the ligament. The temperature on the surface of the
study by Lucas,8 viscoelastic behavior of porcine and ligament was controlled to within 0.6°C. Additional ligament
human cervical spinal ligaments is compared. This study hydration beyond chamber hydration was provided by spray-
found that the mechanical behavior of porcine longitu- ing a mist of 0.9% saline onto the ligament between tests. The
dinal ligaments is similar to human ligaments across dif- ligaments were strained to 20% of their initial length at an
ferent cervical levels and different ligament types. average rate of 78 seconds⫺1 using a protocol that included fast
In this study, a viscoelastic model with separable time- ramp/hold tests and sinusoidal cyclic tests.
dependent relaxation and displacement-dependent in- The test series consisted of a battery of tests at 5 different
stantaneous elastic force behavior was used to determine temperatures steps in order from 37.8°C, 29.4°C, 21.1°C,
12.8°C, and 4.4°C. Before testing, specimens were maintained
the temperature-dependent components of the porcine
at rest for a 30-minute period at each condition to allow the
uniaxial ligament mechanical response. In a previous ligament temperature to equilibrate with the chamber. At each
study, porcine model ligaments have been shown to pro- temperature, a 20% preconditioning strain fast ramp and hold,
vide an acceptable representation of human viscoelastic a 20% strain fast ramp and hold, and a 20% oscillatory strain
ligamentous material response.8 The ligaments were was applied to the ligament. Each fast ramp and hold test was
tested using an environmentally controlled uniaxial test held for 60 seconds. The oscillatory test consisted of 2 fre-
fixture at high rates typical of automobile impact. The quency inputs, 20 cycles at 20 Hz and 20 cycles at 2 Hz, with a
E438 Spine • Volume 32 • Number 16 • 2007
total duration of 11 seconds. The ligaments were allowed to 1.8 By using the same time constants for each of the individual
relax for 10 minutes between each preconditioning and ramp ligament models, the relaxation function coefficients can be
test. averaged to create a general model to describe the mean behav-
ior of the ligaments.
Data Analysis. In this study, the time-force history of the The model’s coefficients were then optimized numerically
ligament was initially modeled using a linear viscoelastic model using a direct integration technique.11 For this optimization,
based on the Boltzmann superposition principal.10 The form of the objective was to minimize the sum squared error between
this model is: the actual force and the predicted model force.
t
冕
Statistical Analysis. The fit of the viscoelastic model to the
dF共␦共兲兲 experimental fast-ramp data was evaluated using the R2 coef-
F共␦,t兲 ⫽ G共t ⫺ 兲 d (1)
d ficient. The model was then validated by comparing the re-
0
sponse of the model to data from the sinusoidal tests.
The coefficients from the model fits were statistically an-
where F is the force response at time t, F is the instantaneous alyzed using 3 methods. First, the model coefficients deter-
elastic function, G(t) is the reduced relaxation function, ␦ is the mined from each ramp test were plotted in box and whisker
displacement, and is the change in time. This hereditary inte- plots to assess the variability of the data. Second, a Student
gral accounts for the previous time history of the ligament and t test was performed on the mean values of each viscoelastic
is derived from the force response of the ligament. Based on coefficient to assess whether the change in each coefficient
previous work, the ligaments are assumed to have linear elastic between temperatures was statistically significant. For this
response to 20% strain. So, the instantaneous elastic function is assessment, statistical significance was assumed to hold for
solely a function of displacement, while the reduced relaxation ␣ ⫽ 0.05 if P ⬍ 0.05. Third, a regression analysis was per-
function is solely a function of time. The relaxation function is formed using the slopes of lines generated using the temper-
normalized so that the sum of the relaxation coefficients equals ature-dependent viscoelastic model coefficients for each lig-
one. ament model. The slope of the line represents the degree of
In Equation 1, the instantaneous elastic force Fe is assumed temperature dependence while the “no slope” reference rep-
to be linear as resents no temperature dependence. The null hypothesis of
the regression is that coefficients are temperature dependent.
Fe ⫽ A␦ (2)
If the P value is greater than 0.05, the null hypothesis is
and the reduced relaxation function G(t) is defined as rejected and the coefficients are considered not temperature
dependent with 95% certainty.
冘
n
Figure 2. Representative force-time histories. Figure 4. Plot of a representative viscoelastic model fit.
37.8 0.0385 ⫾ 0.01 0.0351 ⫾ 0.021 0.0422 ⫾ 0.035 0.657 ⫾ 0.048 0.227 ⫾ 0.034 216 ⫾ 65
29.4 0.0352 ⫾ 0.01 0.0361 ⫾ 0.022 0.0378 ⫾ 0.046 0.646 ⫾ 0.056 0.245 ⫾ 0.042 221 ⫾ 70
21.1 0.0255 ⫾ 0.01 0.0350 ⫾ 0.016 0.0279 ⫾ 0.037 0.632 ⫾ 0.028 0.280 ⫾ 0.031 299 ⫾ 78
12.8 0.0218 ⫾ 0.007 0.0346 ⫾ 0.014 0.0209 ⫾ .03 0.637 ⫾ 0.019 0.286 ⫾ 0.024 361 ⫾ 93
4.4 0.0193 ⫾ 0.007 0.0350 ⫾ 0.012 0.0132 ⫾ 0.024 0.642 ⫾ 0.019 0.290 ⫾ 0.021 434 ⫾ 110
has a  of 1 Hz that corresponds to a timescale of 1000 dominated by this coefficient and has small temperature
milliseconds. The average value of the G1 coefficients are dependence, especially for short-term relaxation. The G⬁
less than 4% of the total relaxation response and in- coefficient has a larger influence on the relaxation re-
crease with temperature with an average slope of 6E- sponse than the G1, G2, and G3 coefficients with the G⬁
4°C⫺1 (R2 ⫽ 0.96). Consequently, the G1 coefficients coefficient offsetting the temperature dependent behavior
have small temperature dependence and have a small of G1 and G3 (generally increasing with temperature)
effect on the relaxation function and resulting viscoelas- over a large portion of the relaxation spectrum.
tic model. The G2 coefficients have a  of 10 Hz that An example instantaneous elastic response for a single
corresponds to a 100 milliseconds timescale. The average ligament is shown in Figure 6. As the temperature of the
value of the G2 coefficients are less than 4% of the total ligament decreased, the slope of the instantaneous elastic
relaxation response and increase with temperature with response increased. Since the instantaneous elastic re-
an average slope of 9E-6°C⫺1 (R2 ⫽ 0.40). The G2 co- sponse of the ligaments is approximately linear, the in-
efficient has a small effect on the total relaxation behav- stantaneous elastic stiffness of the ligaments at each tem-
ior of the ligament as it is less than 4% of the normalized perature may be calculated by finding the slope of
relaxation response. The G3 relaxation coefficients have instantaneous elastic function versus displacement as
a  of 100 Hz that corresponds to a 10-millisecond time- shown in Figure 7. The average values for this stiffness,
scale. The average value of the G3 coefficients is less than A, are shown in Table 2. As the temperature of the liga-
4% of the total relaxation response and increase with ment decreased, the value of the stiffness increased over
temperature with an average slope of 9E-4°C⫺1 (R2 ⫽ 100%. Thus, the instantaneous elastic response is
0.99). The G4 coefficients have a  of 1000 Hz that strongly temperature dependent. The slope of this depen-
corresponds to a 1-millisecond timescale. The average dence is smaller near 37.8°C than near 4.4°C (Table 3).
value of the G4 coefficients is approximately 64% of the
Statistics
total relaxation response and increase with temperature
A representative box and whisker plot by spinal level and
with an average slope of 4E-4°C⫺1 (R2 ⫽ 0.23). The G⬁
temperature is shown for the G2 coefficients in Figure 8.
coefficients represent the long-term behavior of the liga-
The box represents the average of the coefficients with
ments over the timescale tested (60 seconds). The aver-
standard deviation where the line through the middle of
age value of the G⬁ coefficients is approximately 27% of
the box is the average value. The outliers are the “whis-
the total relaxation response and decreases with temper-
kers” and are seen extending from the box. The repre-
ature with an average slope of ⫺2E-3°C⫺1 (R2 ⫽ 0.91).
sentative G2 and G3 plots show several outliers, pre-
Since the G4 coefficient is the largest of the relaxation
dominantly tests involving ligaments 11 and 12.
coefficients, the relaxation behavior of the ligament is
The relaxation coefficient G1 shows statistically sig-
nificant differences in coefficient values for temperature