Abstract
Background: Basilar invagination can be difficult
to diagnose with plain radiography in patients with rheumatoid arthritis.
Although numerous radiographic criteria have been described, few
studies have addressed the reliability of these parameters in the
rheumatoid population. The purpose of the present study was to validate
and compare the most widely accepted plain radiographic criteria
for basilar invagination in this patient population.
Methods: Cervical radiographs of 131 rheumatoid
patients were examined. Of these patients, sixty-seven (twenty-nine
with basilar invagination and thirty-eight without it) were also
evaluated with tomograms, magnetic resonance imaging, and/or sagittally
reconstructed computed tomography scans to detect the presence of
basilar invagination. Three observers who were blinded with regard
to the diagnosis independently scored each radiograph as positive,
negative, or indeterminate according to the established criteria
for invagination proposed by Clark et al., McRae and Barnum, Chamberlain, McGregor,
Redlund-Johnell and Pettersson, Ranawat et al., Fischgold and Metzger,
and Wackenheim. Interobserver and intraobserver variability, sensitivity,
specificity, total percentage of correct results, and negative and
positive predictive values were determined for each criterion as
well as for various combinations of the criteria.
Results: No single test had a sensitivity and a
negative predictive value of greater than 90% as well as a reasonable
specificity and a reasonable positive predictive value. The combination
of the Clark station, the Redlund-Johnell criterion, and the Ranawat
criterion, scored as positive for basilar invagination if any of
the three were positive, proved to be better than any single criterion;
the sensitivity of the combined criteria was 94%, and the negative
predictive value was 91%.
Conclusions: A screening test for basilar invagination
should have a high sensitivity and a high negative predictive value,
so that the disease will not be missed, and yet be specific, so
that the disease will not be overdiagnosed. Our data suggest that
none of the widely utilized plain radiographic criteria meet these
goals. We recommend that measurements be made according to the methods
described by Clark et al., Redlund-Johnell et al., and Ranawat et
al. and, if any of these suggests basilar invagination, tomography
or magnetic resonance imaging should be performed. Since approximately
6% of the cases of basilar invagination in rheumatoid patients would
still be missed with this approach, tomography or magnetic resonance
imaging should be performed on a rheumatoid patient whenever plain radiographs
leave any doubt about the diagnosis of basilar invagination.
Atlantoaxial and subaxial subluxations as well as basilar
invagination can develop in patients with rheumatoid arthritis.
Basilar invagination has been referred to as cranial settling, vertical
settling or migration, vertical atlantoaxial subluxation, atlantoaxial
impaction, superior migration of the odontoid process, and herniation
of the cervical spine into the foramen magnum. Although basilar
invagination is less common than atlantoaxial and subaxial subluxations,
it is more dangerous. Since the compression caused by basilar invagination
affects the brain stem at the level of the foramen magnum, the autonomic
centers may be compromised, resulting in labile blood pressures,
arrhythmias, or sudden death1-4. Early surgical intervention in
patients with basilar invagination has been recommended in more
than twenty studies5-9. Unfortunately, the diagnosis is often delayed,
with potentially devastating consequences, because of difficulties
in making the diagnosis on the basis of plain radiographs. In a
postmortem study of 104 rheumatoid patients, unrecognized cord compression
was found to be the cause of death in 10% (eleven)2. Detecting basilar invagination
on plain radiographs is difficult because the landmarks used to
make the diagnosis are often obscured on these images9-12. The cephalad
extent of the odontoid process cannot be identified on a high percentage
of lateral cervical radiographs of rheumatoid patients13. Since
many accepted criteria for invagination rely on identification of
the tip of the odontoid process, it is apparent why use of such
criteria often results in a missed or delayed diagnosis. A review
of several textbooks on the spine revealed that most authors give
three, four, or five criteria for diagnosing basilar invagination
without recommending a single best one14,15. A review of the literature
revealed few studies in which the various criteria were compared
with regard to accuracy of diagnosis10,11,13.
Since patients with rheumatoid arthritis are seen on a regular
basis by surgeons, the reliability of these methods needs to be
established. The purpose of the present study was to rigorously
compare the most widely accepted plain radiographic criteria for basilar
invagination. The diagnosis made with use of each criterion was
compared with the diagnosis established with the so-called gold
standard of conventional tomography, computerized tomography, or
magnetic resonance imaging. We then sought a single criterion or
a combination of measures that could combine a high sensitivity
(to avoid missing the diagnosis) with a high negative predictive
value (to avoid ordering too many unnecessary advanced imaging studies).
One hundred and thirty-one patients with a diagnosis of rheumatoid
arthritis and involvement of the cervical spine were identified
from the radiographic files of the senior author (H.H.B.). Both
the radiographs and the medical records of all patients were available
for review. A high-quality lateral plain radiograph of the cervical
spine and an anteroposterior open-mouth odontoid radiograph had
to be available for a patient to be included in the study. The lateral
radiographs were made with the head in neutral position and a standard
distance of 6 ft (1.8 m) maintained between the tube and the patient.
In addition to the plain radiographs, one or more of the following
studies of the cervicocranium (the occiput to the second cervical
vertebra) had to be available for each subject: plain anteroposterior
and lateral tomograms, computed tomography scans with sagittal reconstruction,
or magnetic resonance images.
Sixty-seven patients met the inclusion criteria and formed the
basis of this analysis. There were twenty-three men and forty-four
women. The average age at the time of the radiographic evaluation was
62.5 years (range, thirty-nine to eighty-six years).
Radiographic Analysis
The plain radiographs were evaluated by three spine surgeons
(K.D.R., A.S.H., and M.A.P.). All of the observers were familiar
with the radiographic anatomy of the cervicocranium and the various measurement
techniques for diagnosing basilar invagination. All measurements
were made with the same ruler. A diagram showing the relevant anatomic
landmarks and measurement techniques was available during viewing
of the radiographs to maintain consistency.
The three observers independently analyzed the plain radiographs
of each subject without prior knowledge of the results of tomography,
computerized tomography scanning, or magnetic resonance imaging.
First, an attempt was made to identify the following anatomic landmarks
on the lateral cervical radiograph: the opisthion (the dorsal border
of the foramen magnum), basion (the ventral border of the foramen
magnum), clivus, hard palate, atlas, pedicle of the axis, and tip
of the odontoid process (Fig. 1Fig. 1). The ability to visualize
each landmark was rated as definitely readable, partially readable,
and unreadable. The results were tabulated for each observer. By
averaging the values of the three observers, we calculated the percentage
of patients in whom each anatomic landmark was definitely readable,
partially readable, or unreadable.
Next, each observer utilized eight methods to determine the presence
or absence of basilar invagination on the plain radiographs.
Clark station5 (Fig. 2-AFigs. 2-A and 2-B2-B): The station
of the atlas is determined by dividing the odontoid process into
three equal parts in the sagittal plane. If the anterior ring of
the atlas is level with the middle third (station II) or the caudal
third (station III) of the odontoid process, a diagnosis of basilar
invagination is made.
McRae line16: A line is drawn
across the foramen magnum from the basion to the opisthion. Protrusion
of the tip of the odontoid process above this line indicates basilar
invagination.
Chamberlain line17: A line
is drawn from the posterior edge of the hard palate to the opisthion.
Protrusion of the odontoid tip more than 3 mm above this line is
indicative of basilar invagination.
McGregor line18: A line is
drawn from the posterosuperior aspect of the hard palate to the
most caudal point on the midline occipital curve. Basilar invagination
is diagnosed when the apex of the odontoid process rises more than
4.5 mm above this line.
Redlund-Johnell criterion12 (Figs. 3-AFigs. 3-A and 3-B3-B): The distance
between the McGregor line and the midpoint of the caudal margin
of the second cervical vertebral body is measured. A measurement
of less than 34 mm in males and less than 29 mm in females indicates
basilar invagination.
Ranawat criterion9 (Figs. 4-AFigs. 4-A and 4-B4-B): The distance
between the center of the second cervical pedicle and the transverse
axis of the atlas is measured along the axis of the odontoid process. A
measurement of less than 15 mm in males and less than 13 mm in females
is indicative of basilar invagination.
Fischgold-Metzger line19: A
line is drawn between the tips of the mastoid processes on the anteroposterior
open-mouth odontoid radiograph. Protrusion of the tip of the odontoid
process above this line indicates basilar invagination.
Wackenheim line20: A line is
drawn along the superior surface of the clivus. Protrusion of the
odontoid tip posterior to the projection of this line indicates
basilar invagination.
At least twenty-four hours after the initial interpretation of
the radiographs, each observer reinterpreted the radiographs of
twenty randomly selected patients, using the same eight methods
to diagnose basilar invagination, in order to document intraobserver
reliability. The presence or absence of basilar invagination was
then determined for each patient by means of an analysis of the
plain tomograms, computerized tomography scans with sagittal reconstruction,
and/or magnetic resonance images. T1 and T2-weighted images in the
sagittal, axial, and coronal planes were available for the patients who
underwent cervical magnetic resonance imaging. On the basis of one
or more of these studies for each patient, the diagnosis of basilar
invagination was made if the odontoid tip extended above the foramen
magnum. In addition, if the sagittal magnetic resonance image showed
the disappearance of the ventral subarachnoid space and compression
of the medulla at the level of the foramen magnum by the odontoid
process, basilar invagination was diagnosed13. Of the sixty-seven
patients, twenty-nine (43%) had basilar invagination and thirty-eight (57%)
did not. For each observer and for each radiographic method of diagnosing
basilar invagination, we determined the sensitivity (the probability
of obtaining a positive finding for a patient with basilar invagination),
specificity (the probability of obtaining a negative finding in
an individual who does not have basilar invagination), positive
predictive value (the likelihood that a subject who has a positive
finding actually has basilar invagination), negative predictive
value (the likelihood that a subject who has a negative finding
does not have basilar invagination), and total percentage of correct determinations.
The values of the three observers for each parameter were then averaged.
Intraobserver and interobserver reliability were calculated for
each diagnostic method.
Determination of Landmarks
The hard palate was the most readily identifiable landmark, being
definitely visible on sixty-two (93%) of the sixty-seven radiographs.
The arches of the atlas were the next most clearly identifiable landmarks;
they were definitely seen on fifty-nine (88%) of the sixty-seven
radiographs. The tip of the odontoid process was the least readily
identifiable landmark; it was definitely seen on only twenty-three
(34%) of the sixty-seven radiographs and was completely unidentifiable
on thirteen (19%). On the remaining thirty-one radiographs (46%),
the location of the tip of the odontoid process could only be "guessed."
Results of Measurements
There was poor agreement among the various methods for the diagnosis
of invagination. None of the eight radiographic criteria for basilar
invagination had a sensitivity of greater than 90%. The most sensitive
criteria (those with the fewest false-negative results) were the
Wackenheim line (88%) and the Clark station (83%). The least sensitive
criterion was the McRae line; using this line, the investigators
were able to detect basilar invagination on only 43% of the radiographs
of patients with true disease. The greatest specificity (the fewest
false-positive results), which was 76%, was obtained with use of
the Redlund-Johnell criterion. The least specific measurement was
the Fischgold-Metzger line; only 8% of the radiographs that were
positive according to this criterion were actually of a patient with
basilar invagination.
Positive and negative predictive values were also calculated
for each of the criteria. The Redlund-Johnell criterion had the
greatest positive predictive value (68%). In comparison, the Wackenheim
line correctly predicted the presence of basilar invagination in
only 48% of the cases. The Fischgold-Metzger line had the greatest
negative predictive value; every patient who had a negative result
on this test proved to be free of disease. The McRae line had the
lowest negative predictive value (75%).
Intraobserver and interobserver consistency was calculated with
use of the duplicate measurements made by each of the three observers.
The Clark station was the most consistently measured criterion on
the same radiographs by the same observer (87% agreement). The Ranawat
criterion was consistently measured on the same radiographs by the same
observer in only 69% of the cases. The interobserver variability
was similar: the Clark station was the most consistently measured
by different individuals (79% agreement), whereas the Chamberlain
and Fischgold-Metzger lines were the least consistently measured
(65% agreement).
No single test had a sensitivity of higher than 90%, which is
the desired sensitivity for any plain radiographic criterion used
to screen for basilar invagination. However, this objective was
reached when the Clark station, the Redlund-Johnell criterion, and the
Ranawat criterion were measured. When at least one of these tests
was positive for basilar invagination, the sensitivity increased
to 94%, with a negative predictive value of 91%. Consequently, only 6%
of the patients with basilar invagination were falsely diagnosed
as not having that lesion. On the other hand, the positive predictive
value was 56%, which implies that 44% of those diagnosed as having
invagination with use of the test would have a false-positive test
and undergo unnecessary advanced imaging studies to rule out this
diagnosis.
Although basilar invagination is the least common type of subluxation
in rheumatoid patients, most investigators believe that it is the
most dangerous1,3,4,21-23. Thus, many authors have recommended surgical
treatment of rheumatoid patients with basilar invagination, regardless
of their symptoms5-9. This advice is based on the knowledge that
these patients can have rapid progression to paralysis or death.
In order to treat such patients, there must be an accurate and easy
method for establishing the diagnosis. Magnetic resonance imaging
fits these criteria, but it is prohibitively expensive to use for
the evaluation of every rheumatoid patient. Until a less expensive
diagnostic modality becomes readily available, plain radiography
remains the best diagnostic screening tool. Unfortunately, plain
radiographs of rheumatoid patients are often difficult to interpret
because of generalized osteopenia, multiple subluxations, and erosions4,8,9,11,24
. This difficulty is reflected in the multiple radiographic criteria
that have been proposed for diagnosing basilar invagination in this patient
population.
Many of the methods used to determine the presence of basilar
invagination necessitate the identification of the odontoid tip16-20,
which often is not clearly visible on radiographs of rheumatoid patients.
Magnetic resonance studies reveal erosion of the odontoid process
in the majority of rheumatoid patients24,25. Because of this, three
methods do not involve identification of the tip of the odontoid
process. Ranawat et al.9, in 1979, utilized a line connecting the
anterior and posterior arches of the first cervical vertebra. A
perpendicular line drawn to the center of the second cervical pedicle (sclerotic
ring) measuring less than 15 mm in males and less than 13 mm in
females correlated with invagination9. Redlund-Johnell and Pettersson12, in
1984, measured the distance from the McGregor line to a line demarcating
the inferior end plate of the second cervical vertebra. A length
of less than 34 mm in males and less than 29 mm in females was proposed
as being suggestive of invagination. Clark et al.5, in 1989, described
perhaps the simplest method. They divided the height of the odontoid
process into thirds and described three stations based upon the
location of the atlas. Station I, with the atlas level with the
superior third of the odontoid process, was normal. Stations II
and III, with the atlas level with the middle and caudal thirds
of the odontoid process, respectively, correlated with invagination.
The present study confirms that methods that rely on identification
of the tip of the odontoid process are unreliable. This is because
the odontoid tip was found to be clearly defined in only 35% (seventy) of
201 radiographic interpretations. In an additional 46% (ninety-three)
of the 201 interpretations the observer could only make a "best
guess," and in 19% (thirty-eight) the odontoid process was completely
obscured.
The most easily identified landmarks are the hard palate and
the atlas, so it would appear that a diagnostic method based on
these landmarks would be more accurate. The surprising finding from
the present study was that even methods that do not rely on identification
of the odontoid process were unreliable as screening tests. The
sensitivities of the Redlund-Johnell and Ranawat criteria were 61% and
71%, respectively, and the negative predictive values were 77% and
87%, respectively. The ideal screening test should have a high sensitivity,
so that the vast majority of patients with invagination will be
identified, as well as a high negative predictive value, so that
a negative test reliably indicates the absence of invagination.
In addition, it should have a relatively high positive predictive
value in order to avoid an excessive number of costly magnetic resonance
imaging scans of patients without the condition. While a high specificity
is desirable, it is less important than a proper diagnosis of all patients
with the disease. On the other hand, if a screening test has no
specificity, it loses all value as a screen. Given these criteria,
the measurements with the highest combined sensitivity and negative predictive
value were the Clark station (a sensitivity of 83% and a negative
predictive value of 85%) and the Wackenheim line (a sensitivity
of 88% and a negative predictive value of 90%). The specificity of
Wackenheim's method was so low (20%), however, that one would end
up diagnosing invagination in 80% of all patients with rheumatoid
arthritis. With the Clark station, one would miss the diagnosis
in 17% of patients with invagination, but the specificity is such
that a false-positive diagnosis would be made approximately 50%
of the time.
In an effort to determine if we could improve the sensitivity
and negative predictive values of the methods without compromising
their specificity, we evaluated various combinations of measurements.
By scoring the test as positive if any of the methods led to a diagnosis
of invagination, we hoped to increase the sensitivity of the test.
The best test turned out to be a combination of the Clark station,
the Redlund-Johnell criterion, and the Ranawat criterion. If all
three of these tests were negative, there was a 91% chance that
the patient did not have invagination. In addition, the high sensitivity
(94%) meant that disease would be missed in few people. Finally,
using this combination as a screening test, one would perform unnecessary magnetic
resonance imaging scans on only 45% of patients. This positive predictive
value compared favorably with those of all of the other methods.
Based upon the results of this study, we recommend that the workup
for basilar invagination begin with high-quality lateral cervical
radiographs. Any patient with equivocal findings for basilar invagination
on these plain radiographs should be evaluated with magnetic resonance
imaging, tomography, or reconstructed computed tomography images.
In addition, if the patient's clinical picture suggests cervicomedullary
or brain-stem dysfunction despite three negative measurements for
basilar invagination, then we recommend erring on the side of caution
and ordering an advanced imaging study to rule out this potentially
life-threatening condition.
Davidson RC; Horn JR; Herndon JH; and Grin OD: Brain-stem compression in rheumatoid arthritis. JAMA,1977.238: 2633-4, 2382633
1977
[PubMed]
Mikulowski P; Wollheim FA; Rotmil P; and Olsen I: Sudden death in rheumatoid arthritis with atlanto-axial
dislocation. Acta Med Scand,1975.198: 445-51, 198445
1975
[PubMed]
Rana NA; Hancock DO; Taylor AR; and Hill AG.: Upward translocation of the dens in rheumatoid arthritis. J Bone Joint Surg Br,1973.55: 471-7, 55471
1973
[PubMed]
Slatis P; Santavirta S; Sandelin J; and Konttinen YT: Cranial subluxation of the odontoid process in rheumatoid
arthritis. J Bone Joint Surg Am,1989.71: 189-95, 71189
1989
[PubMed]
Clark CR; Goetz DD; and Menezes AH: Arthrodesis of the cervical spine in rheumatoid patients. J Bone Joint Surg Am,1989.71: 381-92, 71381
1989
[PubMed]
Conaty JP, and Mongan ES: Cervical fusion in rheumatoid arthritis. J Bone Joint Surg Am,1981.63: 1218-27, 631218
1981
[PubMed]
Parish DC; Clark JA; Liebowitz SM; and Hicks WC: Sudden death in rheumatoid arthritis from vertical subluxation
of the odontoid process. J Natl Med Assoc ,1990.82: 297-9, 302-4, 82297
1990
[PubMed]
Pellicci PM; Ranawat CS; Tsairis P; and Bryan WJ: A prospective study of the progression of rheumatoid arthritis
of the cervical spine. J Bone Joint Surg Am,1981.63: 342-50, 63342
1981
[PubMed]
Ranawat CS; O'Leary P; Pellicci P; Tsairis P; Marchisello P; and Dorr L: Cervical spine fusion in rheumatoid arthritis. J Bone Joint Surg Am,1979.61: 1003-10, 611003
1979
[PubMed]
Hinck VC; Hopkins CE; and Savara BS: Diagnostic criteria of basilar impression. Radiology,1961.76: 572-85, 76572
1961
[PubMed]
Kauppi M; Sakaguchi M; Konttinen YT; and Hamalainen M: A new method of screening for vertical atlantoaxial dislocation. J Rheumatol,1990.17: 167-72, 17167
1990
[PubMed]
Redlund-Johnell I, and Pettersson H: Radiographic measurements of the cranio-vertebral region.
Designed for evaluation of abnormalities in rheumatoid arthritis. Acta Radiol Diagn (Stockh).,1984.25: 23-8, 2523
1984
[PubMed]
Kawaida H; Sakou T; and Morizono Y: Vertical settling in rheumatoid arthritis. Diagnostic value
of the Ranawat and Redlund-Johnell methods. Clin Orthop,1989.239: 128-35, 239128
1989
[PubMed]
Hensinger RA. Congenital anomalies
of the cervical spine. In: Herkowitz HN, Garfin SR, Balderston RA,
Eismont FJ, Bell GR, Wiesel SW, editors. The spine. 4th
ed, volume 1. Philadelphia: WB Saunders; 1999. p 222
Lubicky JP.The spine in osteogenesis
imperfecta. In: Bridwell KH, DeWald RL, editors..The textbook
of spinal surgery. 2nd ed, volume 1. Philadelphia: Lippincott-Raven; 1996.
p 329
McRae DL, and Barnum AS: Occipitalization of the atlas. . Am J Roentgenol,1953.70: 23-45, 7023
1953
Chamberlain WE: Basilar impression (platybasia). A bizarre developmental
anomaly of the occipital bone and upper cervical spine with striking
and misleading neurologic manifestations. Yale J Biol Med,1939.11: 487-96, 11487
1939
[PubMed]
McGregor M: The significance of certain measurements of the skull
in the diagnosis of basilar impression. Br J Radiol,1948.21: 171-81, 21171
1948
[PubMed]
Fischgold H, and Metzger J: Etude radiotomographique de l'impression basilaire. . Rev Rhum Ed Fr,1952.19: 261, 19261
1952
Wackenheim A.Roentgen
diagnosis of the cranio-vertebral region. New York: Springer;
1974
Henderson DR: Vertical atlanto-axial subluxation in rheumatoid arthritis.. Rheumatol Rehabil,1975.14: 31-8, 1431
1975
[PubMed]
Mathews JA: Atlanto-axial subluxation in rheumatoid arthritis. . Ann Rheum Dis,1969.28: 260-6, 28260
1969
[PubMed]
Vogelsang H; Zeidler H; Wittenborg A; and Weidner A: Rheumatoid cervical luxations with fatal neurological
complications.. Neuroradiology,1973.6: 87-92, 687
1973
[PubMed]
Einig M; Higer HP; Meairs S; Faust-Tinnefeldt G; and Kapp H: Magnetic resonance imaging of the craniocervical junction
in rheumatoid arthritis: value, limitations, indications. . Skeletal Radiol,1990.19: 341-6, 19341
1990
[PubMed]
Glew D; Watt I; Dieppe PA; and Goddard PR: MRI of the cervical spine: rheumatoid arthritis compared
with cervical spondylosis. Clin Radiol,1991.44: 71-6, 4471
1991
[PubMed]