ARTICLE IN PRESS

Journal of Biomechanics 37 (2004) 479–485

Investigation of fixation screw pull-out strength on human spine

Q.H. Zhang, S.H. Tan*, S.M. Chou
School of Mechanical and Production Engineering, Nanyang Technological University, 50 Nanyang Ave, 639798 Singapore
Accepted 1 September 2003

Abstract

For the purpose of fixation and stabilizing the spine, the pull-out strength of the screw is one of the most important factors to be
considered. The material properties of the bone, coupled with the principal dimensions of the screw such as major diameter, minor
diameter, pitch and purchase length, may affect the pull-out strength of the fixation screw. In this study, the effects of various factors
on the bone screw pull-out strength were studied using finite element method of analysis. A three-dimensional finite element model
simulating the threaded connection of bone and surgical screw was constructed. The behavior of both the bone and the screw during
screw pull-out were analyzed and discussed. The effect of the screw parameters on the screw pull-out strength was also investigated.
The results showed that failure of the connection was due to bone shearing which occurred along a cylindrical surface determined by
the outer perimeter of the screw and penetrated length. The distribution of the shear stress along the threaded length in the bone was
nearly uniform. The effect of the major diameter on the screw pull-out strength was more significant than those of the minor
diameter and the pitch. The minor diameter hardly affected the pull-out strength.
r 2003 Elsevier Ltd. All rights reserved.

Keywords: Finite element modeling; Pull-out; Fixation screw; Spine; Factorial analysis

1. Introduction influence it (Hirano et al., 1997; Lim et al., 1995; Myers

et al., 1996; Zink, 1996; Zindrick et al., 1986). However,
The plate–screw system is one of the most popular all the studies were performed experimentally, with no
devices used in dealing with instability due to trauma or numerical study on the load and behavior during screw
previous surgery, scoliosis, spinal stenosis, tumors, and pull-out. Although experimental investigations can
infection. One of the main purposes of orthopaedic provide direct and practical in vitro results, there are
implant systems is to provide the desired biomechanical some problems that are difficult to overcome. The bone
outcome of the fixation and to avoid additional tissue property is an important factor affecting the screw pull-
trauma. For the purpose of stabilizing the inserted spine out strength. To compensate for individual variation of
section, the pull-out strength of the screw is one of the bone quality when studying the effect of screw design,
most important indices concerned by both manufac- comparisons could be performed by placing two screws
turers and surgeons. The material properties of the side by side within the same vertebra. However, this
bone, coupled with the principal dimensions of the screw method would limit the number of screws tested. In
such as major diameter, minor diameter, pitch and addition, since the parameters of the screw are
length, may affect the pull-out strength of the fixation correlative in reality, it is difficult to study the effect of
screw. Hence, it is crucial to identify the main factors each factor independently. Finally, the insertion techni-
and study their effects on the pull-out strength of the que of different operators will be quite different, and
screw. this may also influence the final result. These may be the
Studies have been done by numerous researchers to reasons why the results obtained from different
examine the pull-out strength and the factors that can researchers diverge so much (Chapman et al., 1996;
Lim et al., 1995; Ryken et al., 1995). Compared to the
experiment models, mathematical models can offer the
*Corresponding author. Tel.: +65-6790-4837. ideal opportunity from the point of view of control.
E-mail address: mshtan@ntu.edu.sg (S.H. Tan). They are characterized by absolute repeatability, with

0021-9290/$ - see front matter r 2003 Elsevier Ltd. All rights reserved.
doi:10.1016/j.jbiomech.2003.09.005
 ARTICLE IN PRESS
480 Q.H. Zhang et al. / Journal of Biomechanics 37 (2004) 479–485

the additional advantage that any parameter can be Because of the complex geometry (especially the
varied in the desired degree. posterior elements), it is difficult to build an accurate
The purpose of the present study was to better 3-D model of a vertebra. In this study, an anterior screw
understand the behavior of the bone and screw during fixation was chosen to simplify the vertebra’s model
the process of screw pull-out, and the effects of the screw generation. In this condition, since the screw is
parameters (major diameter, minor diameter, pitch, implanted only in the vertebral body, all the posterior
purchase length) on the pull-out strength of fixation elements can be ignored and the focus is just on the
screw on the human spine. Three-dimensional (3-D) threaded connections when simulating the pull-out
finite element analysis was employed to study the procedure. Compared with the posterior elements, the
relationship of screw geometry and corresponding vertebral body is a geometrically regular solid. Using the
pull-out strength. The pull-out strength and stress data obtained by Chua (1999), the vertebral body was
distribution along the bone–screw joint during the modeled as a block of 16 mm wide (end-plate depth),
screws pull-out process were obtained and presented. 16 mm high (vertebral body height) and 25 mm long
(end-plate width). The position of the threaded hole was
assumed to be at the center of the block, with an axis
2. Method along the length of the block. The profile of the hole was
the same as that of the screw, but the size was a little
2.1. Geometry consideration larger. It is difficult to identify the bone–screw interface.
Because of the bone’s healing, the gap between the bone
A standard 6.5 mm cancellous bone screw with tissue and the screw will change at different periods after
spherical under-surfaces and asymmetrical thread was operation, intimate contact between them will occur
selected as the original screw model. The profile and after 6–12 weeks. Considering that the transfer of
dimensions of the screw thread were derived from the compressive force between screw and bone occurs
British Standards Institution (1991). Fig. 1 shows the mainly at the trailing edge of the screw during the screw
cross section of the thread. The detailed dimensions are pull-out, perfect contact was modeled at this interface
listed in Table 1. It should be noted that the crest of the (Fig. 2). A total of eight threaded joints were modeled to
thread is a fillet according to the standard. Due to its simulate a thread length of 22 mm.
small size (only 0.2 mm for 6.5 mm screw, and often it
will be 0.1 mm for smaller screws), it was modeled as a 2.2. Material properties
sharp angle formed by straight lines to avoid creating
bad elements. The data from Goel et al. (1988) were used to define
the material properties of the vertebral body in this
P e study. In addition, the bone was assumed to be perfectly
plastic after yield point to predict its plastic behavior. A
friction coefficient of 0.2 was designated to the contact
r2 r1

a1 a2 trailing Vertebral body

edge
d1

d2

(a) cross section

leading
edge

Screw
(b) crest in standard (c) crest in model F

Fig. 1. Cross-section of a standard 6.5 mm cancellous bone screw with Fig. 2. 2-D Sketch of the bone–screw interface. Considering that the
spherical under-surfaces and asymmetrical thread, which is derived transfer of compressive force between screw and bone occurs mainly at
from the British Standard. To avoid creating bad elements, the crest the trailing edge of the screw during the screw pull-out, perfect contact
was modeled as sharp angle. was modeled at this interface.

Table 1
Specification of the screw employed in this study

d1 (mm) d2 (mm) a1 ( ) a2 ( ) r1 (mm) r2 (mm) e (mm) P (mm)

6.5 3.0 5 25 1.2 0.8 0.2 2.75

The data were derived from British Standard.

 ARTICLE IN PRESS
Q.H. Zhang et al. / Journal of Biomechanics 37 (2004) 479–485 481

surface between bone and screw according to Liu et al. coarser mesh will suffice in areas of low or constant
(1998). Furthermore, only the cancellous bone was stress gradient. For a thread form, significant mesh
modeled in this study, the effect of cortical shell was not refinement is needed to accurately represent the stress
included. The material of the screw was assumed to be distribution around the thread root radius (Grewal and
316 l stainless steel. This material is commonly used for Sabbaghian, 1997; Tafrreshi and Dover, 1993). Further-
fixation implants. The material properties of the screw more, more elements should be used for bone than for
were derived from literature (Tencer and Johnson, screw due to the prediction that the bone will experience
1994). The material properties of the bone and screw much greater deformation during pull-out. Fig. 4 shows
are tabulated in Table 2. All the materials were assumed the mesh for the 3-D model. A total of 3970 elements
to be homogeneous and isotropic. with 5348 nodes were employed and no bad element was
found.
When the screw is subjected to a pull-out force, the
2.3. Finite element modeling bone and screw will contact and compress each other at
the trailing edge. To resolve contact problems, surface-
The finite element analyses were performed on a to-surface, rigid-to-flexible contact elements were em-
general purpose finite element software ANSYS Version ployed in the model. Eight contact pairs were defined
5.7. For the construction of 3-D model, the geometries between the bone and screw for the eight threaded
of the screw and the bone with a threaded hole were engagements. In each pair, the trailing edge of the screw
separately created using the Pro/Engineer software. Fig. was the target surface and that of the bone was the
3 shows the complete profile of a single thread. The contact surface.
created 3-D curves were imported into ANSYS to model As defined by the researchers doing experiments, the
the screw and bone individually. The meshed screw pull-out strength should be the peak load when pulling
model and bone model were then combined to generate the screw out of the bone. In most experiments, the
the final model. Because of the limitation of disk space, screw head was gripped with a chuck and a displacement
only a segment of 90 thread form was meshed for screw was induced along the axis of the screw at a low rate.
and bone. For the axial loading condition, this Load versus displacement diagrams were continuously
simplification will not influence the accuracy. Eight- recorded on a strip chart to get the maximum load
noded isoparametric brick solid element was chosen for reached. In this study, the considerations of the
the meshing. To optimize the mesh density, a fine mesh boundary conditions and loading conditions were
is needed in regions of high stress gradient while a identical with those applied in the experiments. The
outside surfaces of the bone were not allowed to move in
Table 2 any direction. All the surface areas on the inner and
Material properties of bone and screw
upper cross section (both the bone and screw) were
Material E (MPa) n sY (MPa) restricted in X and Y directions, respectively. Before the
Bone 100 0.2 2.0
loading condition was set, coupled nodes were defined in
Screw 193,000 0.3 250.0 order to apply degrees of freedom (DOF) constraints to
E—Young’s modulus; n—Poisson’s ratio; sY —yield stress.

Fig. 3. 3-D threaded profile of screw and thread hole of the bone were Fig. 4. A total of 3970 elements with 5348 nodes were employed to
all generated in Pro/Engineer. The threaded curves will then be mesh the eight thread joints of bone–screw system. No bad element
imported into ANSYS for construction of bone and screw models. was found.
 ARTICLE IN PRESS
482 Q.H. Zhang et al. / Journal of Biomechanics 37 (2004) 479–485

one node, called a ‘‘master’’ node, which was coupled to be discussed using the results obtained from these five
all the nodes on the front surface of the screw. Using this models.
method, all coupled nodes were forced to have the same The three main parameters of the screw geometry
displacement value along the Z direction. The master selected for the investigation are the major diameter,
node was then given a displacement value to calculate minor diameter and pitch. For an actual screw design,
the force required for pulling the screw out of the bone. these three parameters are usually interactional and
The load value obtained on the master node should be their effects on the pull-out strength cannot be studied
multiplied by 4 to work out the pull-out force because separately. Accordingly, eight different finite element
only a segment of 90 thread form were modeled in the models were constructed and 23 factorial analysis was
simulation. performed. Table 3 lists the levels of the three factors
and their combinations for the eight models analyzed.
The high/low level values (in mm) of the major diameter,
2.4. Parameters analysis minor diameter and pitch are 6.5/6.0, 3.5/3.0 and 3.667/
2.75, respectively. To coincide with the order of the
The effect of purchase length on the pull-out strength observations in a general 2k analysis, the eight screws are
was analyzed by simply changing the numbers of arranged as from (1) to abc. Except for the three
threaded connections between the bone and screw. Four parameters to be investigated, all other parameters were
additional 3-D models were constructed to compute the kept constant. Under this consideration, the four models
pull-out strength of screw with 6, 10, 12 and 14 threaded with pitch value of 2.75 mm have eight threads and the
connections. As the pitch value of the thread is 2.75 mm, other four have six threads to keep the length constant.
these models can represent a range of purchase length All the models have similar mesh formation.
from 16.5 mm (2.75  6) to 38.5 mm (2.75  14). These
four models were also developed on the basis of the
model used in Fig. 4. The effect of purchase length will 3. Results

Table 3 3.1. Pull-out process prediction

Specification of the eight screws used for 23 factorial analysis with the
values selected for high level and low level of the three main screw The left and right part of Fig. 5 shows the Von Mises
design parameters—major diameter, minor diameter and pitch
stress contour of the bone and screw at the peak load,
Model no. Factors investigated Observation respectively. It can be seen from the former that the
A (mm) B (mm) C (mm)
bone experienced significant shear stress at the thread
root during screw pull-out. At the peak load, only those
1 6.0 ( ) 3.0 ( ) 2.750 ( ) (1) elements at the thread root region reached the yield
2 6.5 (+) 3.0 ( ) 2.750 ( ) a
point and failed. The gradient of the stress variation is
3 6.0 ( ) 3.5 (+) 2.750 ( ) b
4 6.5 (+) 3.5 (+) 2.750 ( ) ab great. The stress value of the outer circumference surface
5 6.0 ( ) 3.0 ( ) 3.667 (+) c can be much smaller than those at the root radius. In
6 6.5 (+) 3.0 ( ) 3.667 (+) ac this condition, it can be deduced that the pull-out failure
7 6.0 ( ) 3.5 (+) 3.667 (+) bc of the bone occurred along a surface around the major
8 6.5 (+) 3.5 (+) 3.667 (+) abc
diameter of the screw, where the screw threads cut a
A—major diameter; B—minor diameter; C—pitch. cylinder formed by those failed areas. The peak Von

Fig. 5. Von Mises stress in the bone and screw at the peak load. Failure of the bone occurred along the cylindrical surface formed by the major
diameter of the screw.
 ARTICLE IN PRESS
Q.H. Zhang et al. / Journal of Biomechanics 37 (2004) 479–485 483

Mises stress in the screw also occurred at the thread root is the most important factor, followed by the pitch and
radius, but the value is much lower than its yield stress, minor diameter. When considering the relative variation
i.e., only slight elastic deformation was experienced by between the low level and high level of each factor, the
the screw. The stress distribution along the thread length discrepancy will be clearer. For the major diameter, the
in the screw is also different from that in the bone. It can difference between the low level and high level is only
be seen from Fig. 5 that the pull-out force was almost 7.6% (6 versus 6.5 mm), while those of the minor
evenly carried by all the engaged threads in the bone, diameter and pitch are 16.7% (3.0 versus 3.5 mm) and
and the peak stress was distributed in each of them. For 33.3% (2.75 versus 3.667 mm), respectively. It is obvious
the screw, the stress value decreased gradually from the that a small alteration to the major diameter will result
thread near the head to those near the back. At the peak in a greater difference in the pull-out strength than the
load, the maximum stress (98.5 MPa) only occurred in pitch and minor diameter. In addition, it can be deduced
the first two thread roots and the stress in the last two that the effect of minor diameter and pitch is not
threads was only 34.7 MPa. Fig. 6 shows the stress significant. The effect of the major diameter is positive,
contour in the bone and screw when the screw was while the effects of the pitch and minor diameter are
pulled out 0.1 mm. The stress distribution of each bone negative. This means that a screw with larger major
thread is almost similar. Before the screw was pulled out diameter and smaller minor diameter and pitch will lead
0.2 mm, the bone mainly experienced elastic deforma- to an increase in the pull-out strength. The effects of the
tion. After that, the bone around the thread root area interactions (AB, AC, BC and ABC) are insignificant,
began to reach their yield point gradually and serious which means that the variation of one factor will not
plastic deformation could be seen at the contact influence the effect of other factors on pull-out strength.
surfaces. It can be concluded that the shear force is For example, regardless of the change in minor diameter
distributed almost equally among the bone threads, and pitch, the trend of larger major diameter leading to
which finally failed at the same time before the screw higher pull-out strength will remain the same.
was completely pulled out.

3.2. Parameter analysis

1000
Fig. 7 shows the screw pull-out strength versus thread 900
connection numbers from the five models representing
800
Pull-out Force (N)

different purchase length. It is obvious that a linear

correlation between the pull-out strength and thread 700
numbers can be obtained. This result further confirms 600
that for cancellous bone screw, more threaded connec-
tions can resist a higher force, i.e., longer purchase 500
length can effectively increase pull-out strength. 400
Table 4 lists the pull-out strengths of the eight screws
300
investigated. Fig. 8 shows the computed effects of the 6 8 10 12 14
three main factors and their interactions on the pull-out Thread Number
strength. In a 2k analysis, when the absolute magnitude Fig. 7. Linear correlation between the pull-out strength and thread
of an effect is large, its influence on response is relatively numbers can be obtained, which further confirms that longer purchase
more significant. It can be seen that the major diameter length can effectively increase pull-out strength.

Fig. 6. Von Mises stress in the bone and screw before the peak load. The stress distribution of each bone thread is almost similar.
 ARTICLE IN PRESS
484 Q.H. Zhang et al. / Journal of Biomechanics 37 (2004) 479–485

Table 4
Pull-out strengths for the eight models for 23 factorial analysis

Observation (1) a b ab c ac bc abc

Pull-out strength (N) 447.7 531.2 438.4 510.9 424.0 505.3 395.4 472.4

A 1993), the failure pattern can be quite different if the

80
B magnitudes of the mechanical properties of the two
C
60 AB
engaged materials are close or similar. In this condition,
AC most of the applied load will be carried by only the first
BC
three engaged threads and the purchase length will not
Computed Effects

40 ABC
contribute to the increase in pull-out strength after a
20 certain value.
The results from parametric study showed that a
0
larger major diameter and longer purchase length could
increase the pull-out strength significantly. Reducing the
-20
minor diameter or pitch can also increase the pull-out
-40
strength, but the effects are relatively insignificant.
Main Factors and Their Interactions These conclusions are coincident with the finite element
Fig. 8. Effects of major diameter, minor diameter, pitch and their stress analysis. As illustrated in the stress plots, the bone
respective interactions on the pull-out strength. experienced serious shear stress at the thread root radius
when the screw was pulled out. In this case, the pull-out
4. Discussion strength should be directly related to the shear strength
of the bone along the cylinder whose diameter is equal
In current study, 3-D finite element models were to the major diameter of the screw. Different minor
developed to investigate the failure behavior of the bone diameter or pitch does not change the area of this
during the screw pull-out and the effect of screw sheared circumference, i.e., the influence on the shear
geometric parameters on the pull-out strength. To the strength is relatively not significant. In the study of
best of the authors’ knowledge, no comprehensive Chapman et al. (1996), polyurethane foams, whose
studies are currently available to model the detailed 3- apparent densities and shear strengths encompassed that
D thread connections, including the helical curve, using of human cancellous bones, were employed to study the
finite element method. specific effects of screw thread geometry on pull-out
Since the main focus of the present study is on the strength. The study concluded that in porous bone, the
effects of screw geometries on the pull-out strength, the major diameter, purchase length, the shear strength of
material property of the cancellous bone was simplified the material into which the screw is embedded will
to be homogeneous and isotropic. It is well known that, determine the pull-out strength.
as a typical cellular solid, the stiffness and strength of Generally, the importance and significance of the
the cancellous bone are related to the bone density. major diameter of screw on the pull-out strength have
However, since the magnitude of the mechanical been accepted by most researchers, regardless of the
properties of the human cancellous bone are much vertebra section where the screw is inserted. However,
lower than that of the screw, the discrepancy of the the effect of purchase length on the pull-out strength
different sections of the cancellous bone can be ignored. reported in literatures are still inconclusive or contra-
The results showed that cancellous bone screw pull-out dictory. For example, the results obtained by Krag et al.
causes failure of the connection due to bone shearing (1986) presented a greater fixation strength for pedicle
with little or no damage to the screw. The force needed screws with 80% penetration depth compared with
for pull-out is related to the surface area of the cylinder screws with 50% penetration depth, while Zindrick et al.
which is determined by the major diameter of the screw (1986) reported that no difference was found in pull-out
and purchase length. All the results agree well with the strength of screws having penetration depth of 50%
observations from experiments (Chapman et al., 1996; versus 100%. However, the authors did not quantify the
Tencer and Johnson, 1994) and this indicates that the variation of the mineral densities (or mechanical
simplification is reasonable. However, the prediction of properties) of the bone specimens. As discussed by
the pull-out strength should not be applied to other Hirano et al. (1997), the area of the cancellous section of
fixation type, such as pedicle screw or cortical screw the pedicle was only 27% of the whole cross-sectional
insertions, due to the much stronger materials in these area and the screw threads are usually engaged in the
vertebra sections. As discussed by other researchers denser subcortical section. Furthermore, the regional
(Grewal and Sabbaghian, 1997; Tafrreshi and Dover, bone mineral density of the pedicle is significantly denser
 ARTICLE IN PRESS
Q.H. Zhang et al. / Journal of Biomechanics 37 (2004) 479–485 485

than that of the anterior vertebral body. This may be the Krag, M.H., Beynnon, B.D., Pope, M.H., Frymoyer, J.W., Haugh,
reason why deeper insertion of the pedicle screw had L.D., Weaver, D.L., 1986. An internal fixator for posterior
little effect on fixation stiffness. application to short segments of the thoracic, lumbar, or
lumbosacral spine. Design and testing. Clinical Orthopaedics and
Related Research 203, 75–98.
Lim, T.H., An, H.S., Evanich, C., Hasanoglu, K.Y., McGrady, L.,
Wilson, C.R., 1995. Strength of anterior vertebral screw fixation in
References relationship to bone mineral density. Journal of Spinal Disorder 8,
121–125.
British Standards Institution, 1991. Implants for Osteosynthesis. Part Liu, C.L., Chen, H.H., Cheng, C.K., Kao, H.C., Lo, W.H., 1998.
5, Bone Screws and Auxiliary Equipment. British Standards Biomechanical evaluation of a new anterior spinal implant. Clinical
Institution, London (Section 5.3, Specification for the dimensions Biomechanics 13, S40–S45.
of screws having hexagonal drive connection, spherical under Myers, B.S., Belmont Jr., P.J., Richardson, W.J., Yu, J.R., Harper,
surfaces and asymmetrical thread). K.D., Nightingale, R.W., 1996. The role of imaging and in situ
Chapman, J.R., Harrington, R.M., Lee, K.M., Anderson, P.A., biomechanical testing in assessing pedicle screw pull-out strength.
Tencer, A.F., Kowalski, D., 1996. Factors affecting the pullout Spine 21, 1962–1968.
strength of cancellous bone screws. Journal of Biomechanical Ryken, T.C., Clausen, J.D., Traynelis, V.C., Goel, V.K., 1995.
Engineering 118, 391–398. Biomechanical analysis of bone mineral density, insertion techni-
Chua Hu Chuan, 1999. Quantitative 3-D anatomy of the human que, screw torque, and holding strength of anterior cervical plate
vertebrae (Asian population). M.Eng. Thesis, Nanyang Technolo- screws. Journal of Neurosurgery 83, 325–329.
gical University, School of Mechanical and Production Engineer- Tafrreshi, A., Dover, W.D., 1993. Stress analysis of drillstring
ing. threaded connections using the finite element method. Interna-
Goel, V.K., Kim, Y.E., Lim, T.H., Weinstein, J.N., 1988. An tional Journal of Fatigue 15, 429–438.
analytical investigation of the mechanics of spinal instrumentation. Tencer, A.F., Johnson, K.D., 1994. Biomechanics in Orthopedic
Spine 13, 1003–1011. Trauma Bone Fracture and Fixation. Martin Dunitz, London, pp.
Grewal, A.S., Sabbaghian, M., 1997. Load distribution between 86–88.
threads in threads connections. Journal of Pressure Vessel Zindrick, M.R., Wiltse, L.L., Widell, E.H., Thomas, J.C., Holland,
Technology, Transactions of the ASME 119, 91–95. W.R., Field, B.T., Spencer, C.W., 1986. A biomechanical study of
Hirano, T., Hasegawa, K., Takahashi, H.E., Uchiyama, S., Hara, T., intrapeduncular screw fixation in the lumbosacral spine. Clinical
Washio, T., Sugiura, T., Yokaichiya, M., Ikeda, M., 1997. Orthopaedics and Related Research 203, 99–112.
Structural characteristics of the pedicle and its role in screw Zink, P.M., 1996. Performance of ventral spondylodesis screws in
stability. Spine 22, 2504–2509. cervical vertebrae of varying bone mineral density. Spine 21, 45–52.