SPINE Volume 25, Number 13, pp 1607–1616
©2000, Lippincott Williams & Wilkins, Inc.
Biomechanical Rationale for the Pathology of
Rheumatoid Arthritis in the Craniovertebral Junction
Christian M. Puttlitz, PhD,* Vijay K. Goel, PhD,* Charles R. Clark, MD,*
Vincent C. Traynelis, MD,† Jeffrey L. Scifert,* and Nicole M. Grosland, PhD*
Study Design. A finite-element model of the craniover-
tebral junction was developed and used to determine
whether a biomechanical mechanism, in addition to in-
flammatory synovitis, is involved in the progression of
rheumatoid arthritis in this region of the spine.
Objectives. To determine specific structure involve-
ment during the progression of rheumatoid arthritis and
to evaluate these structures in terms of their effect on
clinically observed erosive changes associated with the
disease by assessing changes in loading patterns and
degree of anterior atlantoaxial subluxation.
Summary of Background Data. Rheumatoid arthritis
involvement of the occipito-atlantoaxial (C0 –C1–C2) com-
plex is commonly seen. However, the biomechanical con-
tribution to the development and progression of the dis-
ease is neither well understood nor quantified. Although
previous cadaver studies have elucidated information on
kinematic motion and fusion techniques, the modeling of
progressive disease states is not easily accomplished us-
ing these methods. The finite-element method is well
suited for studying progressive disease states caused by
the gradual changes in material properties that can be
modeled.
Methods. A ligamentous, nonlinear, sliding-contact,
three-dimensional finite-element model of the C0 –C1–C2
complex was generated from 0.5 mm thick serial com-
puted tomography scans. Validation of the model was
accomplished by comparing baseline kinematic predic-
tions with experimental data. Transverse, alar, and cap-
sular ligament stiffness were reduced sequentially by
50%, 75%, and 100% (removal) of their intact values. All
models were subjected to flexion moments replicating
the clinical diagnosis of rheumatoid arthritis using full
flexion lateral plane radiographs. Stress profiles at the
transverse ligament-odontoid process junction were
monitored. Changes in loading profiles through the
C0 –C1 and C1–C2 lateral articulations and their associ-
ated capsular ligaments were calculated. Anterior and
posterior atlantodental interval values were calculated to
correlate ligamentous destruction with advancement of
atlantoaxial subluxation.
Results. Model predictions (at 0.3 Nm) fell within one
standard deviation of experimental means, and range of
motion data agreed with published in vitro and in vivo
values. The model predicted that stresses at the posterior
base of the odontoid process were greatly reduced with
From the Iowa Spine Research Center, *Departments of Biomedical
Engineering and Orthopaedic Surgery, and †Division of Neurosurgery,
University of Iowa, Iowa City, Iowa.
Supported in part by Surgical Dynamics, Inc., and the Iowa Spine
Foundation Research Fund.
Acknowledgment date: December 16, 1998.
Acceptance date: October 8, 1999.
Device status category: 1.
Conflict of interest category: 16.
transverse ligament compromise beyond 75%. Decreases
through the lateral C0 –C1 and C1–C2 articulations were
compensated by their capsular ligaments. Anterior and
posterior atlantodental interval values indicate that the
transverse ligament stiffness decreases beyond 75% had
the greatest effect on atlantoaxial subluxation during the
early stages of the disease (no alar and capsular ligament
damage). Subsequent involvement of the alar and capsu-
lar ligaments produced advanced atlantoaxial subluxa-
tion, for which surgical intervention may be warranted.
Conclusions. To the best of the authors’ knowledge,
this is the first report of a validated, three-dimensional
model of the C0 –C1–C2 complex with application to rheu-
matoid arthritis. The data indicate that there may be a
mechanical component (in addition to enzymatic degra-
dation) associated with the osseous resorption observed
during rheumatoid arthritis. Specifically, erosion of the
odontoid base may involve Wolff’s law of unloading con-
siderations. Changes through the lateral aspects of the
atlas suggest that this same mechanism may be partially
responsible for the erosive changes seen during progres-
sive rheumatoid arthritis. Anterior and posterior atlanto-
dental interval values indicate that complete destruction
of the transverse ligament coupled with alar and/or cap-
sular ligament compromise is requisite if advanced levels
of atlantoaxial subluxation are present. [Key words: bio-
mechanics, rheumatoid arthritis, upper cervical spine]
Spine 2000;25:1607–1616
Rheumatoid arthritis (RA) is a chronic, inflammatory
disorder of a largely unknown etiology characterized by
articular joint involvement. Prevalence data on RA in-
volvement among the United States general population is
estimated to be 0.3% to 1.5%,41 with women more pre-
disposed to RA affliction than men. Of those affected by
RA, involvement of the cervical spine ranges from 17%
to 86%,35 with the craniovertebral junction most often
affected. The most frequently reported abnormality is
anterior atlantoaxial subluxation (AAS). The literature is
replete with RA cohort studies reporting the prevalence
of RA-induced lesions involving the upper cervical
spine.1,3,6,7,10,19,22,23,25,27,28,33,34,37 Clark et al6 re-
ported that AAS was present in 20 (49%) of 41 patients
with RA. Similar results were found by Babic-Naglic et
al1 and Morizono et al,25 with AAS apparent in 44% and
49%, respectively, of patients with RA.
These studies are typical of the data reporting RA
involvement. From these data sets, various rationales
with respect to certain musculoskeletal structures have
been developed in support of the radiographic findings.
Regarding upper cervical involvement, most believe that
inflammatory destruction via synovitis of the atlantal
1607
1608 Spine • Volume 25 • Number 13 • 2000

transverse ligament leads to anteroposterior instability

of the atlantoaxial joint.7,23,27 The manifestation of this
instability is anterior subluxation of the atlas on the axis.
Erosion of the odontoid process frequently coincides
with the development of AAS. Protrusion of the odon-
toid process posteriorly into the spinal canal can result in
clinical symptoms or signs such as suboccipital pain and
myelopathy. Further progression of the disease involves
loss of alar and capsular ligament integrity (in addition
to further destruction of the transverse ligament) and
advanced osseous erosion of the odontoid process. Fi-
nally, loss of atlanto-occipital and atlantoaxial articula-
tion integrity via joint surface destruction and atlantal
lateral mass erosion and osteolysis27 leads to cranial set-
tling or basilar invagination of the odontoid process.
Occipitalization of the atlas has been shown to result in
irreversible paralysis and death.10
Radiographic diagnostic criteria have been developed
as descriptors of existence and advancement of AAS.
Two of the most commonly used criterion, anterior and
posterior atlantodental intervals (AADI and PADI, re-
spectively), measure the distances between the atlas and
the odontoid process,2,3,27 as shown in Figure 1. The
determination of AADI and PADI involves constructing
a line that connects the centroids of the anterior and
posterior rings of the atlas on a lateral plain radiograph
at maximal flexion. The AADI is the distance along this
line that measures the difference between the posterior
surface of the anterior aspect of the atlantal ring and the
anterior surface of the odontoid process. The PADI, the
quantity complementary to AADI, is measured as the
distance between the posterior surface of the odontoid
process and the anterior surface of the posterior ring of
the atlas. Most studies define the existence of AAS when
the AADI is greater than 3 mm or when the PADI is less
than or equal to 14 mm.
Although most of the radiographic findings have doc- Figure 1. Diagnosis of atlantoaxial subluxation via determination
umented the typical sequelae of osseous and soft tissue of the anterior and posterior atlantodental intervals (AADI and
PADI, respectively). A, The diagnosis is performed using lateral
abnormalities, there is a lack of biomechanical evidence, plane radiographs during full flexion. B, Model displacement plot
if any, supporting the occurrence of this sequence. That indicates the degree of flexion at 1.5-Nm flexion moment (darker
is, no studies have been published, known to the current elements indicate initial position before flexion load onset) and
authors, correlating the biomechanical degradation of shows how AADI and PADI were determined for this study.
specific spinal elements during the various stages of RA
to progression of the disease, and showing how the as-
and atlantoaxial (C1–C2) lateral articulation load trans-
sociated loading changes can lead to further progression.
missions. Also, the implications of erosive ligament dam-
This partly because it is not possible to use traditional in
age was evaluated with respect to clinical diagnosis and
vitro methods, such as cadaver specimens, to mimic the
intervention. The degree of AAS with ligament compro-
progressive nature of RA.
mise was determined by monitoring the radiographic di-
In light of this, the current study sought to evolve a
agnostic criteria, AADI and PADI, providing a biome-
biomechanical rationale for the clinically perceived le-
chanical basis for clinical intervention.
sions of RA using the finite-element (FE) method. The
authors hypothesized that the changes in loading that Methods
accompany the loss of mechanical integrity of specific Model Generation. A three-dimensional, nonlinear, FE
musculoskeletal structures favors progression of the dis- model of the occipito-atlantoaxial (C0 –C1–C2) complex was
ease. Specifically, the investigation sought to determine, generated from human cadaver data. The specimen, that of a
for the first time, the effect of transverse, alar, and cap- 54-year-old woman with no prior spinal ailments, included the
sular ligament destruction on the transverse ligament– intact cranium and entire cervical spine. Radiographs and dual-
odontoid process and on the occipito-atlantal (C0 –C1) energy radiograph absorptiometry were used to confirm the
 Pathology of Rheumatoid Arthritis • Puttlitz et al 1609

Table 1. Craniovertebral Junction Finite-Element Model

Characteristics, With Element Types and Number of
Elements Indicated
ABAQUS No. of
Structure Element Type Elements

Bone
Occiput C3D8 and C3D6 5186
Atlas C3D8 and C3D6 2303
Axis C3D8 and C3D6 2971
Ligaments
Transverse C3D8 60
Superior/inferior cruciform T3D2 6
Alar T3D2 22
Apical T3D2 4
Accessory T3D2 6
Nuchal T3D2 25
Tectoral membrane/ligament T3D2 29
Anterior longitudinal T3D2 22
Posterior C0–C1 capsular T3D2 11
Anterior C0–C1 capsular T3D2 13
Posterior C1–C2 capsular T3D2 12
Anterior C1–C2 capsular T3D2 20
Articulations
Lateral occipito-atlantal C3D8* 1098
Lateral atlantoaxial C3D8* 1456
Atlantodental C3D8* 291
Transverse ligament–odontoid C3D8* 210
process
* Surface elements were internally generated from the existing mesh, thus
they represent opposing faces of brick elements.
C3D8 ⫽ linear, isotropic eight-noded element (“brick”); C3D6 ⫽ linear, isotro-
pic six-noded element (“wedge”); T3D2 ⫽ linear two-noded element (no com-
pression transmission, (“cable”).
Both right and left lateral aspects of the occipito-atlantal (C0 –C1) and atlanto-
axial (C1–C2) articulations are modeled separately.

single elastic modulus based on the work of Fielding et al,13

Figure 2. Finite-element model mesh of the occiput (C0), atlas (C1)
and axis (C2) shown from anterior (A) and lateral (B) viewpoints.
The lateral view also depicts some of the ligaments.
integrity and bony geometry of the cervical spine and occiput.
The occiput and upper cervical spine were scanned (settings:
120 kV, 200 mA, 1 second exposure time; slice thickness: 0.5 mm)
using computed tomography (Toshiba Xpress/SX, Tustin, CA).
In all, 188 transverse plane scans were produced. Each plane
was digitized for nodal output, and elemental connectivities
were assigned to the nodes. Stacking of the serial planes pro-
duced a three-dimensional model of the C0 –C1–C2 complex
(Figure 2). The cortical and cancellous regions of each bone
were delineated during model generation such that different
material properties could be assigned for each region. Bone ele-
ments were mainly eight-noded “brick” elements with some six-
noded “wedge” elements used for transition areas (Table 1).
The upper cervical spine ligaments (except the transverse
ligament) were modeled using three-dimensional “hypoelastic”
cable elements (no compression transmission). The transverse
ligament was modeled using eight-noded brick elements such
that modeling of the odontoid process–transverse ligament ar-
ticulation could be reproduced and represented, as dictated by
the FE code. This modeling technique was capable of being
implemented with no axial compression force transmission in
the longitudinal direction of the ligament.
In addition, the transverse ligament was modeled using a
wherein they described the ligament as “a very firm, inelastic
band of tissue.” Other modeled ligaments included the cruci-
form ligament, alar ligament, apical ligament, accessory liga-
ment, nuchal ligament, tectoral ligament/membrane, anterior
longitudinal ligament, occipito-atlantal capsular ligaments,
and atlantoaxial capsular ligaments. The C0 –C1 and C1–C2
anterior and posterior capsular ligaments were modeled as sep-
arate entities. The cruciform ligament was modeled as having
two segments: the transverse ligament and the superior/inferior
fascicles. Attachment sites and orientations for each ligament
were determined from direct anatomic dissection and spine
biomechanics/anatomy texts.5,39 Material and geometric prop-
erties for the transverse and alar ligaments were determined
from published data.9,26 All other ligaments were assigned
nominal material and cross-sectional area properties from an
earlier FE model of the lower cervical spine.15 The geometric
and material property data are given in Table 2.
Joint articulations were modeled using the ABAQUS Ver-
sion 5.7 (Hibbitt, Karlsson, & Sorenson, Inc., Pawtucket, RI)
deformable sliding formulation. Six sliding articulations were
incorporated into the model: right and left occipito-atlantal
joints, right and left atlantoaxial joints, dento-atlantal joint
(between the anterior aspect of the odontoid process and the
posterior aspect of the anterior ring of the atlas), and the trans-
verse ligament– odontoid process articulation (between the
transverse ligament and the base of the posterior aspect of the
odontoid process). The model consisted of 12,661 elements
and 19,185 nodes, with further details provided in Table 1. All
model solutions were obtained using ABAQUS.
1610 Spine • Volume 25 • Number 13 • 2000
Table 2. Material and Geometric Properties Used to Represent Various Components in the Model
Material Young’s Modulus (MPa)
Bone
Cortical bone 10,000
Cancellous bone 450
Ligaments*
Transverse (%) 20.0
Superior/inferior cruciform (%) 6.0 (⬍17), 10.0 (⬎17)
Alar (%) 5.0 (⬍17), 8.5 (⬎17)
Apical (%) 6.0 (⬍17), 10.0 (⬎17)
Accessory (%) 6.0 (⬍17), 10.0 (⬎17)
Nuchal (%) 12.0 (⬍17), 20.0 (⬎17)
Tectoral membrane/ligament (%) 6.3
Anterior longitudinal (%) 12.0 (⬍17), 20.0 (⬎17)
Posterior C0–C1 capsular (%) 6.0 (⬍17), 10.0 (⬎17)
Anterior C0–C1 capsular (%) 6.0 (⬍17), 10.0 (⬎17)
Posterior C1–C2 capsular (%) 6.0 (⬍17), 10.0 (⬎17)
Anterior C1–C2 capsular (%) 0.2 (⬍17), 1.25 (⬎17)
* All ligaments, except the transverse and tectoral membrane ligaments, were modeled as hypoelastic and have different stiffnesses during ligament strain
excursion. Transverse ligament representation using volume brick elements was necessary because of the inclusive modeling of the transverse ligament–
odontoid process articulation.
† When available, geometric data were derived from published studies9,15 and given nominal area properties from an earlier model of the lower cervical spine15
for all other ligaments.
Experimental Design. The model was validated by compar-
ing its kinematic predictions with cadaver data obtained at the
Iowa Spine Biomechanics Laboratory.14 The model was kine-
matically constrained for all nodes lying on the inferior surface
of the axis, thus replicating the in vitro study. Occipital mo-
ments of 0.3 Nm, the same location and load magnitude used in
the cadaveric investigation, were applied to the model. Finally,
the range of motion data predicted by the model were com-
pared to published data.
In an attempt to determine the role of transverse ligament
destruction in odontoid process erosion, the material proper-
ties of the transverse ligament were varied. Specifically, the
stiffness of the transverse ligament was reduced sequentially by
50%, 75%, and 100% (complete removal). Kinematic data,
including coupled motions, were reported to determine the ef-
fect of decreases in transverse ligament stiffness on motion
characteristics.
The stress profile at the transverse ligament– odontoid pro-
cess articulation was monitored to evaluate the change in the
local osseous tissue mechanical environment of the axis. In
addition, changes in occipito-atlantal and atlantoaxial articu-
lation force transmission were computed. The associated
change in C0 –C1 and C1–C2 capsular ligament forces was
determined from model output.
The role of specific ligament involvement during the devel-
opment and advancement of AAS was evaluated by calculating
the AADI and PADI after reductions in transverse, alar, and
capsular ligament stiffnesses. The AADI and PADI values were
determined by computing vector distances from nodal displace-
ment output, with nodal selection mandating that the inclusive
points fall along the vector connecting the centroids of the
anterior and posterior arches of the atlas. Each specific liga-
ment stiffness was reduced sequentially by 50%, 75%, and
100% (complete removal) while the other two remaining liga-
ments were held at their baseline (0% reduction) stiffness. Also,
combinations of two or three ligament stiffnesses were reduced
to predict changes in AADI and PADI. Finally, attempts were
made to correlate the model predictions of AADI and PADI
with values observed clinically during advanced AAS.
The applied boundary conditions and loadings were
Poisson’s Ratio Cross-Sectional Area (mm2)
0.3 —
0.3 —
0.3 18†
0.3 5
0.3 22†
0.3 5
0.3 5
0.3 5
0.3 5
0.3 6†
0.3 5
0.3 5
0.3 5
0.3 5
consistent for all models. The model was constrained at all
nodes located on the inferior surface of the axis. Pure flexion
moments of 1.5 Nm were applied to the occiput for all ligament
stiffness reduction models and the intact model (baseline data).
The rationale for choosing this load magnitude and for present-
ing only the flexion results are discussed in a latter section of
this article.
Results
The model was validated by comparing its kinematic
predictions with cadaver data14 at 0.3 Nm of pure mo-
ment loading. Table 3 shows the model predictions and
associated cadaver data. All model motion prediction
data, including coupled motions resulting from loading
in the sagittal (flexion– extension) and frontal planes (lat-
eral bending), fell within one standard deviation of the
cadaver data. The range of motion predicted by the FE
model, obtained by applying 1.5 Nm moments on the
occiput, was comparable with in vivo and in vitro pub-
lished data11,12,20,21,29,31,32,38 (Table 4). Specifically, the
model predicted the combined flexion– extension range
of motion at C0 –C1 and C1–C2 to be 45.5° and 31°,
respectively. Previous investigations11,32 have reported
combined occipito-atlantal (C0 –C1) sagittal plane mo-
tions of 50°. Furthermore, Penning31 reported total flex-
ion– extension at the atlantoaxial level (C1–C2) to be
30°. The model predicted frontal plane C0 –C1 and
C1–C2 range of motions to be 6.3° and 12.3°, respec-
tively. Panjabi et al29 reported lateral bending at the
C0 –C1 level to be 5.5°, and Penning31 reported a C1–C2
lateral bending value of 10°. Therefore, the model rea-
sonably represents, within anatomic variation, “nor-
mal” motion in flexion, extension, and lateral bending.
The kinematic model predictions indicate that flexion
at 1.5 Nm loading did not appreciably increase at either
level during the initial stages of transverse ligament de-
struction (Figure 3). However, the atlantoaxial flexion
 Pathology of Rheumatoid Arthritis • Puttlitz et al 1611

Table 3. Validation of Model Kinematic Predictions With Cadaver Data for 0.3-Nm Occipital Moment Loading
C0–C1 C1–C2

Applied Moment Cadaver Model Cadaver Model

Rotation Direction Data (°) Prediction Data (°) Prediction

Flexion
Flexion* 6.5 ⫾ 2.5 5.4 4.9 ⫾ 2.0 3.2
Lateral bending 1.2 ⫾ 0.3 1.0 0.4 ⫾ 1.0 0.3
Axial rotation 0.1 ⫾ 1.3 0.3 0.6 ⫾ 1.5 0.6
Extension
Extension* 16.5 ⫾ 7.6 12.4 5.2 ⫾ 2.9 7.9
Lateral bending 1.4 ⫾ 1.3 1.6 0.6 ⫾ 1.6 0.5
Axial rotation 1.9 ⫾ 2.5 0.5 0.3 ⫾ 3.0 1.2
Lateral bending
Extension 1.2 ⫾ 1.2 1.4 0.4 ⫾ 1.0 0
Lateral bending* 3.4 ⫾ 2.8 1.3 4.2 ⫾ 2.8 3.6
Axial rotation 1.0 ⫾ 1.8 0.3 2.3 ⫾ 2.9 2.6
* Main and coupled motions for sagittal and frontal planes are given.
Values are mean ⫾ SD.

rotation increased from 10.6° to 13.7° (29%) when the
transverse ligament stiffness was reduced from 75% to
100% (complete removal).
The average contact stress, in the absence of any lig-
ament degradation, at the odontoid process–transverse
ligament junction was predicted to be 1281 kPa. Flexion
models (1.5-Nm loading) indicate that loss of transverse
ligament stiffness resulted in decreased contact stresses in
the axis at this site, with 50% and 75% reductions re-
sulting in decreased mean contact stresses of 880 kPa and
780 kPa, respectively. This was a manifestation of de-
creased force transmission through the contact patch.
The initial force resulting from contact was predicted to
be 25.2 N, with the 50% and 75% ligament stiffness
models predicting that the force transmission would de-
crease to 21.3 N and 18 N, respectively.
Occipito-atlantal (C0 –C1) contact forces were pre-
dicted by the FE model to be 52.6 N (right) and 26.7 N
(left) for the intact transverse ligament case at 1.5 Nm of
flexion. The same forces for the atlantoaxial (C1–C2)
lateral articulation were reported to be 35.2 N (right)
and 44.3 N (left) for the fully intact case. Both levels,
C0 –C1 and C1–C2, experienced greatly reduced contact
force transmission after transverse ligament removal
(Figure 4). Inconsequential increases (less than 2%)
in occipital condyle-atlantal superior facet force trans-
mission were predicted, with up to 75% reduction in
transverse ligament stiffness, whereas right and left
C0 –C1 lateral facet joints constantly decreased during
this time.
Flexion (1.5-Nm moment) models predicted that
changes in articular loading are compensated by contri-
butions of the associated capsular ligaments. In a fully
intact transverse ligament, the anterior and posterior oc-
cipito-atlantal capsular ligaments transmitted 2.6 N and
49.5 N, respectively, with the atlantoaxial capsular liga-
ments demonstrating tensions of 0.8 N (anterior) and
33.9 N (posterior). Complete transverse ligament disrup-
tion resulted in atlantoaxial load transfer increases of
approximately 50% through the posterior capsular liga-
ments (Figure 5). Occipito-atlantal capsular ligaments
experienced insignificant (less than 1%) increases in their
load-bearing role, with up to 75% transverse ligament
stiffness reduction.

Table 4. Range of Motion Comparison of the Finite-Element (FE) Predictions With Previously Reported In Vitro and In
Vivo Occipito-Atlantal and Atlantoaxial Range of Motion Measurements for Combined Flexion–Extension and One-Side
Lateral Bending
Study C0–C1 C1–C2

Combined Combined
Flexion–Extension Lateral Bending Flexion–Extension Lateral Bending
Reference Type (degree) (degree) (degree) (degree)

Fick11 In vivo 50 34–40 0 0

Poirier and Charpy32 In vivo 50 14–40 11 —
Penning31 In vivo 30 10 30 10
Werne38 In vitro 13 8 10 0
Panjabi et al.29* In vitro 24.5 5.5 23.4 6.7

FE Model (1998) 45.5 6.3 31 12.3

29
* Range of motions reported by Panjabi et al. were obtained by 1.5-Nm occipital loading.
The model predictions fall within the reported ranges of motion at C0 –C1 and closely approximate the C1–C2 motions reported by Penning.31
1612 Spine • Volume 25 • Number 13 • 2000

Figure 4. The effect of transverse ligament compromise (0% ⫽

intact, 100% ⫽ complete removal) on percentage of change in
occipito-atlantal (C0 –C1) and atlantoaxial (C1–C2) lateral articula-
tion contact force corresponding to 1.5-Nm flexion loading.

decreasing to 15.28 mm at full flexion. As an isolated

entity, the model predicted that the transverse ligament
had the greatest effect on AADI (Figure 6) in the fully
flexed posture. Without transverse ligament disruption,
both alar and capsular ligament compromise did not
contribute significantly to the development of AAS. Sub-
stantial AAS, increases in AADI greater than 1.5 mm
(AADI greater than 4.4 mm) or decreases in PADI of 1.5
mm or more (PADI less than or equal to 13.8mm), was
not obtained until the transverse ligament stiffness was
reduced by 75%.
Finally, combinations of alar and capsular ligament
disruptions were modeled with transverse ligament re-
moval in an attempt to describe the interactive effect of
Figure 3. Rotation angle as a function of transverse ligament
compromise (0% ⫽ intact, 100% ⫽ complete removal) at C0 –C1
(A) and C1–C2 (B). Flexion is the main motion (1.5-Nm flexion
moment applied). Lateral bending and axial rotation are coupled
motions.

The center of contact force for the C0 –C1 and C1–C2

lateral articulation position was predicted by the model
at full flexion (1.5 Nm). All contact centroids shifted
medially after complete transverse ligament disruption,
with right and left atlantoaxial contact forces moving 1.4
mm and 1.9 mm, respectively. Three of the lateral artic-
ulations evidenced anterior movement, with the left
C1–C2 joint contact moving posteriorly (1.9 mm).
The AADI and PADI values were calculated for all
flexion models (1.5 Nm loading), and the data was com-
pared to the intact case. The baseline model (fully intact
transverse, alar, and capsular ligaments) exhibited atlan-
Figure 5. Effect of transverse ligament compromise (0% ⫽ intact,
todental contact (AADI ⫽ 0 mm) and a PADI of 18.20 100% ⫽ complete removal) on percentage of change in occipito-
mm in neutral (unloaded) posture. In flexion loading, atlantal and atlantoaxial posterior capsular ligament force trans-
the AADI increased to 2.92 mm, with this parameter mission corresponding to 1.5-Nm flexion loading.

Figure 6. Anterior atlantodental interval (AADI) calculated for the
intact model and models with stiffness reductions of the trans-
verse, alar, and capsular ligaments at the fully flexed posture
(1.5-Nm moment load). The stiffness of each ligament was altered
while the other two ligaments were held at the baseline value
(completely intact).
ligament compromise, which may lead to advanced AAS.
Alar ligament compromise with intact capsular liga-
ments markedly increased the level of AAS (Table 5).
Subsequent capsular ligament stiffness loss (50%) with
complete alar ligament removal led to an additional de-
crease in PADI of 0.92 mm. Simultaneous resection of
the transverse, alar, and capsular ligaments resulted in a
highly unstable situation, wherein force equilibrium was
not attainable by the model.
Discussion
Cadaver experimentation of the craniovertebral junction
has provided valuable knowledge about its kinematic
behavior characteristics.11,12,17,20,21,29,31,32,37 However,
because of the inherent nature in this type of testing,
cadaver investigation is somewhat limited in its ability to
model progressive disease states accurately. Application
of the FE method for modeling the occipito-atlantoaxial
region allows for investigation of specific structure involve-
ment during normal and diseased conditions. Furthermore,
Table 5. Combinations of Ligament Stiffness Reductions
With the Resultant Degree of AAS (AADI and PADI
Values) at Full Flexion (1.5-Nm Moment)
Reduction in Ligament Stiffness (%) Criteria (mm)
Transverse Alar Capsular AADI
0 0 0 2.92
100 0 50 5.77
100 0 75 6.21
100 50 0 7.42
100 75 0 7.51
100 100 50 8.43
AAS ⫽ anterior atlantoaxial subluxation, AADI ⫽ anterior atlantodental inter-
val, PADI ⫽ posterior atlantodental interval.
Zero (0) ligament stiffness values represent completely intact ligament stiff-
ness; 100 corresponds to total ligament destruction (via removal).
Pathology of Rheumatoid Arthritis • Puttlitz et al 1613
changes in the material properties of specific structures can
be evaluated without disrupting other spinal elements. This
is not always possible with physical testing methods. With
respect to the current study, isolated transverse ligament
disruption was difficult to produce in a cadaver model be-
cause it is encased posteriorly by the posterior atlantoaxial
membrane. Disruption of the transverse ligament necessi-
tates partial or complete resection of this membrane. Addi-
tionally, serial sectioning of the transverse (or alar and/or
capsular) ligament is not as representative of the progres-
sive nature of disease states such as RA, in which the entire
ligament may experience material property changes.
To the best of the authors’ knowledge, this article
represents the first description of an FE technique ap-
plied to upper cervical spine RA involvement. Validation
of the model was accomplished by comparison of model
kinematic predictions with cadaver data at 0.3-Nm pure
moment loading. In addition, the model range of motion
for flexion, extension, and lateral bending was obtained
by the application of 1.5-Nm loading. The ranges of
motion predicted by the model fell within the physiologic
range of motion determined by earlier published studies
(Table 5) using both in vivo and in vitro meth-
ods.11,12,20,21,29,31,32,37 These findings also are in agree-
ment with the work of Panjabi et al,29 in which they
determined that 1.5-Nm occipital moments produced
physiologic range of motion.
Furthermore, the clinical diagnosis and progression of
rheumatoid involvement of the upper cervical spine fre-
quently is obtained by measuring diagnostic criteria on
lateral plane radiographs during full flexion. Therefore,
in order to impose full physiologic flexion on the model,
1.5-Nm occipital pure flexion moment loading was used
for the evaluation of RA development and progression.
In reality, this moment value may change from person to
person because of such factors as age, sex, and muscle
conditions. However, the important point to keep in
mind is that model AADI and PADI values were calcu-
lated when the model was at full flexion, thus mimicking
the clinical evaluation of RA.
As with any analytical investigation, the model is not
without its limitations. The FE model was constructed
using geometric data from a single cadaver, and thus may
not represent the range of anatomic variation with re-
spect to bony dimensions. However, this limitation does
not seem to apply to the determination of AAS. In neutral
posture, the model PADI value was calculated to be 18.2
mm, with no AADI. In the study by Boden et al,3 the
PADI combined AADI and PADI values averaged 19.3 mm in
patients with isolated AAS (n ⫽ 9). These agreements
15.28
12.43
demonstrate that the model mimics the anteroposterior
11.99 dimension of space available for the spinal cord. Further-
10.79 more, possible deviations of AADI and PADI from spec-
10.71
9.83
imen to specimen and between specimen and radio-
graphic findings may exist, resulting in changes in AADI
and PADI model absolute values. However, the relative
change in AADI and PADI with transverse ligament stiff-
ness loss will remain the same, and the observation that
1614 Spine • Volume 25 • Number 13 • 2000
major changes occur with transverse ligament stiffness
loss greater than 75% will still hold.
Published material and geometric property data were
used in modeling the transverse and alar ligaments. all
other property data were extracted from a previous
model of the lower cervical spine.15 From kinematic
model predictions (Table 3), it seems that error intro-
duced by these approximations was minimal. Finally, the
FE model does not include any muscle contributions.
Loading by pure flexion moments on the occiput resulted
in physiologic range of motion. Therefore, the effect of
muscle contributions has been reproduced as a whole
and, more importantly, should not change the trends
predicted by the model. Furthermore, because all models
were evaluated in flexion, specific muscles producing this
motion should contribute equally and are not likely to be
affected by ligament compromise.
The anterior atlantodental interval traditionally has
been used as an indicator of AAS. The inclusion of PADI
in this investigation was based on several factors. Boden
et al3 were able to demonstrate that preoperative PADI
values are a reliable predictor of the development and
severity of paralysis. They also showed that PADI was a
better predictor than AADI of whether postoperative
neurologic recovery was expected. This seems reason-
able because PADI has been shown to be correlated
closely with the space available for the spinal cord.30
Finally, because PADI values are by nature larger than
AADI values, the degree of error introduction from phys-
ical measurement will be (from lateral plane radio-
graphs) reduced if the PADI parameter is used.
The observation of erosive damage at the transverse
ligament– odontoid process is a commonly observed phe-
nomena in patients with RA. The model predicts that
stress reduction in the axis at the transverse ligament–
odontoid process junction is produced with transverse
ligament compromise. In light of this, the authors believe
that the phenomena observed in the clinical milieu, spe-
cifically erosion of the odontoid process, may not be
entirely the result of enzymatic degradation, but that a
coincident mechanical mechanism is contributing to the
observed bone loss. That is, the loss of bone at the pos-
terior base of the odontoid process may be caused par-
tially by resorption in accordance with Wolff’s law con-
siderations. It has been shown clinically that unloading
of bone leads to osteopenia and resorption.4,8,24,36,40
This mechanism has been demonstrated specifically in
the lumbar spine using mathematical modeling.16,18
Another commonly observed phenomenon in patients
with RA is loss of atlantolateral mass osseous tissue.
Involvement of the C0 –C1 and C1–C2 lateral articula-
tions leads to erosive changes in the atlas, resulting in
cranial settling and basilar invagination. The model has
shown that loss of the transverse ligament leads to an
associated decrease in the force transmitted through the
occipito-atlantal and atlantoaxial facet joints. As with
the aforementioned erosion of the odontoid process,
these loading changes favor resorption in the periarticu-
lar regions stemming from stress relief of the associated
underlying bone. Increased loading of the capsular liga-
ments compensates for decreased articular joint trans-
mission, perhaps leading to increased joint laxity via lig-
ament stretching. The data show that pronounced
flexion motion and anteroposterior translation, quanti-
tated by higher AADI and lower PADI values, occurs at
the atlantoaxial level after total transverse ligament dis-
ruption. This may expedite the erosion of the lateral joint
surfaces by allowing excessive motion, especially if sub-
stantial or complete cartilage erosion is present.
A mechanical basis for atlantolateral joint erosion is
supported further by the center of contact force data.
Progression of the occipito-atlantal and atlantoaxial
contact vectors medially unloads both the right and left
lateral aspects of these articulations. Therefore, it seems
that all of the model predictions with respect to loading
changes favor lateral mass resorption of the atlas.
Clinically, the evaluation of AAS in patients with RA
has been accomplished by measuring AADI and PADI on
lateral plane radiographs. Using this parameter, the au-
thors were able to confirm the observation that the trans-
verse ligament is the main stabilizer of anteroposterior
translation stability. Furthermore, the alar and capsular
ligaments function as secondary stabilizers to sagittal
plane translation. These findings agree in principle with
the cadaver investigation of Fielding et al,13 in which the
transverse (primary) and alar (secondary) ligaments were
implicated as the main resistors to anteroposterior trans-
lation.
The current authors extended these findings by dem-
onstrating that the capsular ligament also should be con-
sidered a secondary stabilizer to anteroposterior transla-
tion. Also, the level of AAS does not proceed to
pronounced levels (AADI greater than 6 mm, PADI less
than 12 mm) until the alar and/or capsular ligament stiff-
nesses are reduced by 75%. With complete transverse
ligament disruption, the model predicts that isolated alar
compromise will produce PADI values between 10 and
11 mm. Boden et al3 showed that surgical intervention in
patients with this degree of AAS (PADI ⬎ 11 mm) was
effective, with excellent potential for some degree of
postoperative neurologic recovery. Complete transverse
and alar ligament resection with capsular ligament in-
volvement (50% stiffness reduction) will result in PADI
values less than 10 mm. Overall, the findings suggest that
AAS cannot progress to pronounced levels without com-
plete disruption of the transverse ligament and signifi-
cant compromise of the alar, capsular ligaments, or both.
Summary
To the best of the authors’ knowledge, this study repre-
sents the first generation of a validated, fully three-
dimensional FE model for investigation of RA develop-
ment and progression. The motivation for this
investigation was a lack of information concerning the
biomechanical role, if any, in the progression of RA.
With the loss of transverse ligament stiffness, FE model

predictions indicate that loss of load transfer in the odon-
toid process and lateral masses of the atlas match those
regions known to experience bone loss during RA. Al-
though inflammatory synovitis and pannus create a de-
grading enzymatic environment, it is the authors’ belief
that the mechanical environment associated with the ini-
tial lesion (transverse ligament compromise) also favors
bone resorption, and thus should not be overlooked. The
PADI data indicate that the transverse ligament is the
main anteroposterior translation stabilizer, and AAS
cannot proceed to pronounced levels without complete
transverse ligament destruction.
Further study is planned for incorporation of bone
remodeling algorithms into the model, allowing for spe-
cific site and quantification of bone loss. Also, investiga-
tion is warranted to determine whether the biomechani-
cal contribution hypothesis discussed in this article can
be extended to other clinically observed lesions occurring
during the latter stages of RA. Inclusion of the spinal
cord and vertebral artery would allow for predictions of
how these lesions affect the mechanical environments of
these structures. Finally, model predictions evaluating
the biomechanical efficacy of existing and new treatment
regimens could prove to be invaluable.
Key Points
● An osteoligamentous finite element model of the
craniovertebral junction (C0 –C1–C2) was gener-
ated and validated to study the biomechanical con-
siderations associated with rheumatoid arthritis.
● The model showed that loss of transverse liga-
ment structural integrity leads to odontoid unload-
ing, which may produce advanced resorption of
bone.
● Complete disruption of the transverse ligament
as well as alar and capsular ligament stiffness loss is re-
quired for advancement of atlantoaxial subluxation.
● Overall, the model showed that changes associ-
ated with the onset of rheumatoid arthritis result in
a mechanical environment that expedites progres-
sion of the disease.
Acknowledgment
The authors thank Dr. Yutaka Sato and Scot Heery for
their assistance in obtaining the computed tomography
scans, and Dr. Martin Cassell for his dissection of the
upper cervical spine.
References
1. Babic-Naglic D, Nesek-Madaric V, Potocki K, Lelas-Bahun N, Curkovic B.
Early diagnosis of rheumatoid cervical myelopathy. Scan J Rheumatol 1997;26:
247–52.
2. Boden SD, Clark CR. Rheumatoid arthritis of the cervical spine. In: Clark
CR, ed. The Cervical Spine. Philadelphia: Lippincott-Raven, 1998:693–703.
3. Boden SD, Dodge LD, Bohlman HH, Rechtine GR. Rheumatoid arthritis of
Pathology of Rheumatoid Arthritis • Puttlitz et al 1615
the cervical spine: A long-term analysis with predictors of paralysis and recovery.
J Bone Joint Surg [Am] 1993;75:1282–97.
4. Carter DR, Orr TE. Skeletal development and bone functional adaptation.
J Bone Miner Res 1992;7(Suppl 2):S389 –95.
5. Chase RA, Gosling JA, Dolph J, Glasgow EF, Mathers LH. Clinical Anatomy
Atlas. St. Louis: Mosby, 1996.
6. Clark CR, Goetz DD, Menezes AH. Athrodesis of the cervical spine in rheu-
matoid arthritis. J Bone Joint Surg [Am] 1989;71:381–92.
7. Conaty JP, Mangan ES. Cervical fusion in rheumatoid arthritis. J Bone Joint
Surg [Am] 1981;63:1218 –23.
8. Cowin SC. Bone stress adaptation models. J Biomech Eng 1993;115:528 –33.
9. Dvorak J, Schneider E, Saldinger P, Rahn B. Biomechanics of the craniocer-
vical region: Alar and transverse ligaments. J Orthop Res 1988;6:452– 61.
10. El-Khoury GY, Wener MH, Menezes AH, Dolan KD, Kathol MEH. Cranial
settling in rheumatoid arthritis. Radiology 1980;137:637– 42.
11. Fick R. Handbuch der Anatomie und Mechanik der Gelenke. Jena, Ger-
many: Verlag von Gustav Fischer, 1904.
12. Fielding JW. Cineroentgenography of the normal cervical spine. J Bone Joint
Surg [Am] 1957;39:1280 – 8.
13. Fielding JW, Cochran GVB, Lawsing JF, Hohl M. Tears of the transverse
ligament of the atlas. J Bone Joint Surg [Am] 1974;56:1683–91.
14. Goel VK, Clark CR, Gallaes K, Liu YK. Moment-rotation relationships of
the ligamentous occipito-atlanto-axial complex. J Biomech 1988;21:673– 80.
15. Goel VK, Clausen JD. Prediction of load sharing among spinal components
of a C5–C6 motion segment using the finite element approach. Spine 1998;23:
684 –91.
16. Goel VK, Ramirez SA, Kong WZ, Gilbertson LG. Cancellous bone: Young’s
modulus variation within the vertebral body of a ligamentous lumbar spine:
Application of bone adaptive remodeling concepts. J Biomech Eng 1995;117:
266 –71.
17. Goel VK, Winterbottom JM, Schulte KR, et al. Ligamentous laxity across
C0 –C1–C2 complex: Axial torque-rotation characteristics until failure. Spine
1990;15:990 –5.
18. Grosland NM, Goel VK, Grobler LJ, Griffith SL. Adaptive Internal Bone
Remodeling of the Vertebral Body Following an Anterior Interbody Fusion: A
Computer Simulation. The 24th Annual Meeting of the International Society for
the Study of the Lumbar Spine, Singapore, June 2– 6, 1997.
19. Halla JT, Hardin JG. The spectrum of atlantoaxial facet joint involvement in
rheumatoid arthritis. Arthritis Rheum 1990;33:325–9.
20. Holh M. Normal motions in the upper portion of the cervical spine. J Bone
Joint Surg [Am] 1964;46:1777–9.
21. Holh M, Baker HR. The atlanto-axial joint roentgenographic and anatom-
ical study of normal and abnormal motion. J Bone Joint Surg [Am] 1964;46:
1739 –52.
22. Kauppi M, Hakala M. Prevalence of cervical spine subluxations and dislo-
cations in a community-based rheumatoid arthritis population. Scand J Rheuma-
tol 1994;23:133– 6.
23. Kawaida H, Sakou T, Morizono Y, Yoshikuni N. Magnetic resonance im-
aging of upper cervical disorders in rheumatoid arthritis. Spine 1989;14:1144 – 8.
24. Lanyon LE. Control of bone architecture by functional load bearing. J Bone
Miner Res 1992;7(Suppl 2):S369 –75.
25. Morizono Y, Sakou T, Kawaida H. Upper cervical involvement in rheuma-
toid arthritis. Spine 1987;12:721–5.
26. Myklebust JB, Pintar F, Yoganandan N, et al. Tensile strength of spinal
ligaments. Spine 1988;13:526 –31.
27. Oda T, Fujiwara K, Yonenobu K, Azuma B, Ochi T. Natural course of
cervical spine lesions in rheumatoid arthritis. Spine 1995;20:1128 –35.
28. Paimela L, Laasonen L, Kankaanpaa E, Leirisalo-Repo M. Progression of
cervical spine changes in patients with early rheumatoid arthritis. J Rheumatol
1997;24:1280 – 4.
29. Panjabi M, Dvorak J, Duranceau J, et al. Three-dimensional movements of
the upper cervical spine. Spine 1988;13:726 –30.
30. Patel TC, Bell G, Panjabi MM, Cholewicki J. Atlanto-Axial Subluxation: A
Quantitative Analysis. Presented at the 23rd Annual Meeting of the Cervical
Spine Research Society, Santa Fe, New Mexico, 1995.
31. Penning L. Normal movements of the cervical spine. AJR Am J Roentgenol
1978;130:317–26.
32. Poirier P, Charpy A. Traite d’anatomie humaine, vol 1. Paris, 1926:74 – 89.
33. Rana NA. Natural history of atlanto-axial subluxation in rheumatoid arthri-
tis. Spine 1989;14:1054 – 6.
34. Ranawat CS, O’Leary P, Pellicci P, Tsairis P, Marchisello P, Dorr L. Cervical
spine fusion in rheumatoid arthritis. J Bone Joint Surg [Am] 1979;61:1003–10.
35. Rawlins BA, Girardi FP, Boachie-Adjei O. Rheumatoid arthritis of the cer-
vical spine. Trends Orthop Surg Rheum Arthritis 1998;24:55– 65.
36. Rubash HE, Sinha RK, Shanbah AS, Kim SY. Pathogenesis of bone loss after
total hip arthroplasty. Orthop Clin North Am 1998;29:173– 86.
37. Weissman BNW, Aliabadi P, Weifeld MS, Thomas WH, Sosman L. Prog-
1616 Spine • Volume 25 • Number 13 • 2000

nostic features of atlantoaxial subluxation in rheumatoid arthritis

patients.Radiology 1982;144:745–51. Address reprint requests to
38. Werne S. Studies in spontaneous atlas dislocation. Acta Orthop Scand 1957;
23:1– 89. Vijay K. Goel, PhD
39. White AA, Panjabi MM. Clinical Biomechanics of the Spine. 2nd ed. Phila- Iowa Spine Research Center
delphia: JB Lippincott, 1990. 1406 Seaman’s Center for the Engineering
40. Woo SLY, Akeson WH, Coutts RD, et al. A comparison of cortical bone Arts and Sciences
atrophy secondary to fixation with plates with large differences in bending stiff-
University of Iowa
ness. J Bone Joint Surg [Am] 1976;58:190 –5.
Iowa City, IA 52240
41. Zvaifler NJ. Rheumatoid arthritis: Epidemiology, etiology, rheumatoid fac-
tor, pathology, pathogenesis. In: Schumacher HR, ed. Primer on the Rheumatic
E-mail: Vijay-Goel@uiowa.edu
Diseases. Atlanta: Arthritis Foundation, 1988:83–7.