Abstract
Background: In 1989, a group of sixty-seven
asymptomatic individuals with no history of back pain underwent
magnetic resonance imaging of the lumbar spine. Twenty-one subjects
(31%) had an identifiable abnormality of a disc or of the
spinal canal. In the current study, we investigated whether the
findings on the scans of the lumbar spine that had been made in
1989 predicted the development of low-back pain in these asymptomatic
subjects.
Methods: A questionnaire concerning the development
and duration of low-back pain over a seven-year period was sent
to the sixty-seven asymptomatic individuals from the 1989 study.
A total of fifty subjects completed and returned the questionnaire.
A repeat magnetic resonance scan was made for thirty-one of these
subjects. Two neuroradiologists and one orthopaedic spine surgeon
interpreted the original and repeat scans in a blinded fashion,
independent of clinical information. At each disc level, any radiographic
abnormality, including bulging or degeneration of the disc, was
identified. Radiographic progression was defined as increasing severity
of an abnormality at a specific disc level or the involvement of
additional levels.
Results: Of the fifty subjects who returned the
questionnaire, twenty-nine (58%) had no back pain. Low-back
pain developed in twenty-one subjects during the seven-year study
period. The 1989 scans of these subjects demonstrated normal findings
in twelve, a herniated disc in five, stenosis in three, and moderate
disc degeneration in one. Eight individuals had radiating leg pain;
four of them had had normal findings on the original scans, two
had had spinal stenosis, one had had a disc protrusion, and one
had had a disc extrusion. In general, repeat magnetic resonance
imaging scans revealed a greater frequency of disc herniation, bulging,
degeneration, and spinal stenosis than did the original scans.
Conclusions: The findings on magnetic resonance
scans were not predictive of the development or duration of low-back
pain. Individuals with the longest duration of low-back pain did
not have the greatest degree of anatomical abnormality on the original, 1989
scans. Clinical correlation is essential to determine the importance
of abnormalities on magnetic resonance images.
The occurrence of radiographic abnormalities of the lumbar spine
in asymptomatic individuals has been described in a number of studies.
Lumbar myelograms have demonstrated abnormal findings in up to 24% of
normal subjects (seventeen of seventy-one subjects); computerized
tomography scans, in up to 36% (twenty-one of fifty-eight
subjects); and discograms, in up to 38% (twenty-seven)
of seventy-two disc levels1-3.
Magnetic resonance imaging studies by Boden et al. and Jensen et
al. demonstrated intervertebral disc abnormalities in 30% (twenty)
of sixty-seven asymptomatic subjects and in 29% (twenty-eight)
of ninety-eight asymptomatic subjects, respectively4,5.
Although radiographic abnormalities of lumbar discs in asymptomatic
subjects are an accepted clinical finding, their ability to predict
the development of low-back and radiating pain and the natural history
of the specific radiographic disc abnormalities in asymptomatic
individuals has not been established, to our knowledge1-8. The purpose of this prospective
investigation was to determine the association between abnormal
findings on magnetic resonance images of the lumbar spine and the
development and duration of low-back pain in the cohort of sixty-seven asymptomatic
individuals described by Boden et al. in their study published in
19904. The hypothesis of the present
study was that abnormalities on the magnetic resonance images would
not be predictive of the development of back or leg pain over a
seven-year period.
In 1989, volunteers were recruited for the study through advertising
in newspapers or by word-of-mouth4.
Review by the Institutional Review Board was not required by the
sponsoring institution at that time. Respondents and their spouses
were chosen to obtain a correct balance for gender in each age-range.
For a volunteer to be included in the study, he or she had to have
no history of pain in the back, sciatica, or neurogenic claudication.
Any episode of nonradiating low-back pain that had lasted for more
than twenty-four hours or had necessitated an absence from work
was grounds for exclusion from the study. Volunteers were also excluded
if they had sciatica (pain or sensory abnormalities in the buttocks
or lower limbs) or if walking caused pain or sensory deficits distal
to the knee.
Once the subject had been entered into the study in 1989, multiplanar
magnetic resonance imaging was done from the first lumbar to the
first sacral vertebra with a 1.5-tesla imaging system (Signa; General
Electric, Milwaukee, Wisconsin). Sagittal and axial localizing series
were performed. Technical details of the scans are described in
our previous report4.
In 1996, the entire cohort of sixty-seven asymptomatic individuals
with scans from the previous study was contacted by letter. Written
informed consent was obtained from all patients. The study was approved
by the chairman of our Institutional Review Board. The letter requested
completion of an eighteen-item questionnaire. The questionnaire
asked for information concerning age, gender, episodes of low-back pain
since 1989, duration of pain, absence from work, sensory abnormalities
in the buttocks or lower extremities, abnormalities associated with
walking, a validated pain diagram, visits to a health-care provider,
specific diagnoses, tests performed, treatment received, treatment
outcome, coexistent illnesses, surgical procedures, and concurrent
medications. Eight individuals who did not wish to complete the
long questionnaire were asked for information concerning the presence
and duration of back pain and the location of pain in the lumbar
area and/or either leg during the past seven years. Also
included with the questionnaire was a request for a repeat magnetic
resonance scan of the lumbar spine. Individuals who did not respond
were contacted by telephone. A location service was used to find
the volunteers who had moved in the seven-year period.
In 1996, the thirty-one individuals who agreed to participate had
repeat multiplanar magnetic resonance imaging with use of a 1.5-tesla
unit (Magnetom Vision; Siemens Medical Systems, Erlangen, Germany).
Three sagittal image series (T1, proton density, and T2) were obtained
with a 4-mm slice thickness from the superior end plate of the third
lumbar vertebra to the superior end plate of the first sacral vertebra. Angled
axial cuts of the third lumbar to the first sacral disc spaces were
also performed. The image series were done without contrast enhancement.
A spine surface coil (quadrature form) was used for signal reception.
The fifty 1989 magnetic resonance studies of the individuals who
completed the questionnaire (or answered questions regarding back
pain) were randomly mixed with the thirty-one magnetic resonance
studies from 1996. The eighty-one scans were reviewed, in random
sequence, by two board-certified neuroradiologists (C.P. and D.S.)
and a board-certified orthopaedic spine surgeon (W.C.L.), who were
blinded with regard to the clinical status of the subjects. A determination
concerning the presence of a herniated nucleus pulposus, stenosis
of the canal, bulging, and degeneration was made at each disc level.
Each category was rated independently at each disc level. In addition
to rating the severity of the abnormality, the evaluators rated
their certainty about the diagnosis (definite, probable, or possible)
(Table I).
Each disc level was graded on an ordinal scale, ranging from 0
(normal) to 3 points (free disc fragment, severe stenosis of the
canal, diffuse disc bulging, or severe disc degeneration). At least
two evaluators needed to agree on the presence of an abnormality
(a score of 1 point) at a disc level in order for the level to receive
a score above 0 points. For example, a grading of severe disc degeneration
(a score of 3 points) by one evaluator was not adequate to categorize
a level as abnormal unless at least one other evaluator gave the
level a score of 1 point.
A change in the average score of 1 point was considered indicative
of progressive disease, whereas a decrease of 1 point was indicative
of improvement. Only the findings that the interpreters had labeled
as definitely or probably abnormal were tabulated. Therefore, an
abnormal ranking required two interpreters to give a score of 1
point with definite or probable certainty at the same disc level.
Once a level was scored as abnormal in this manner, the ratings
of all three evaluators were averaged.
Some individuals had more than one abnormality in the lumbar
spine on the magnetic resonance scan. The rank order of disc and
canal abnormalities used to place a subject in a radiographic category,
from the most to the least severe, was disc herniation, stenosis
of the canal, disc bulging, and disc degeneration. The degree of
severity of the abnormality did not preempt this rank order. For
example, an individual with moderate spinal stenosis (a score of
2 points), a diffuse disc bulge (a score of 2 points), and severe
disc degeneration (a score of 3 points) was listed once in the spinal
stenosis category for statistical analysis and was not included
in the disc-bulge or disc-degeneration categories.
The clinical condition of the subjects (the presence of back
or leg pain or sensory abnormalities, for example) in the seven-year
study period was matched with the abnormalities on the magnetic
resonance images made in 1989. For analysis of the correlation of
the magnetic resonance abnormalities and back pain and for analysis
of agreement among the evaluators of the magnetic resonance studies,
only the more severe alterations (an extruded or free fragment for
the herniated disc category, moderate or severe stenosis, diffuse
bulge, and moderate or severe degeneration) were considered anatomically
important abnormalities. Correlation was not measured for milder
anatomical abnormalities, such as disc protrusion, asymmetric disc
bulge, or mild disc degeneration.
Statistical Methods
Analysis of the correlation between the duration of pain (measured
on an ordinal scale from 0 points for no pain to 5 points for pain
of more than six weeks’ duration) and the magnetic resonance
abnormality consisted of the Spearman rank-order correlation. The
predictive correlative value of magnetic resonance abnormalities
and back pain was tested with chi-square analysis with significance
set at p < 0.05. The degree of progression of magnetic
resonance abnormalities from 1989 to 1996 was evaluated with use
of the Wilcoxon signed-rank test. The statistical method utilized
for rating agreement between pairs of evaluators was a 2 2 table
with a kappa coefficient. The three kappa coefficients were averaged.
A kappa value of 0.5 indicated good agreement9.
The maximum rating of severity of any of the four anatomical changes
at any level of the spine was the value utilized for analysis. This
test was used for the fifty magnetic resonance scans made in 1989
and the thirty-one magnetic resonance scans made in 1996. Data analysis
was performed with SAS software (version 6.12; SAS Institute, Cary,
North Carolina).
During the study period, two male subjects died and two others
moved from the United States. Thirteen additional subjects did not
reply to repeated telephone calls and written requests for clinical
information. The average age of the fifty subjects from the 1989
study who answered the survey (or questions regarding back pain)
in 1996 was 43.6 years compared with 42.0 years for the total group
of sixty-seven subjects from the original study. The average age
of the cohort of thirty-one subjects who underwent a repeat magnetic
resonance scan was fifty-two years (range, twenty-six to sixty-eight
years).
Analysis of the fifty magnetic resonance scans made for the 1989
study revealed that thirty-two had normal findings and eighteen
had abnormal findings (see Appendix). Over the seven-year study
period, pain in the low back or leg developed in twenty-one individuals,
including twelve (38%) of those with a normal scan, five
of the six subjects with a herniated intervertebral disc at any
level, three of the four subjects with spinal stenosis, none of
the six subjects with disc bulging, and one of the two subjects
with moderate disc degeneration. Of the twenty-one individuals who
had pain, eight described discomfort radiating into the thigh and
leg. The 1989 magnetic resonance scans of these eight subjects revealed
an extruded disc between the fifth lumbar and the first sacral vertebrae
in one, a protruded disc between the first and second lumbar vertebrae
in one, and moderate stenosis of the canal between the fourth and
fifth lumbar vertebrae in two. The remaining four subjects had a
normal magnetic resonance scan or mild disc degeneration.
We analyzed the results for the thirty-one subjects who had a repeat
magnetic resonance imaging study in 1996 (see Appendix). In the
seven-year period, a herniated or protruded disc developed in ten
of them, mild-to-severe stenosis developed in ten, a bulging disc
developed in twenty, and moderate or severe disc degeneration developed
in fourteen. Eight of the thirty-one subjects continued to have
no magnetic resonance abnormality.
Five of the six subjects with a herniated intervertebral disc
on the magnetic resonance scan made in 1989 had a repeat study. Disc
protrusion or extrusion persisted at one level or more in all five
subjects. Two disc extrusions resolved to a protruded state, and
one totally resolved. Only one of the five subjects had radiating
pain lasting longer than two weeks. During the seven-year study
period, five new disc herniations developed in four subjects who
had had normal findings on the magnetic resonance images made in
1989 and in one subject who had had evidence of spinal stenosis.
Two of these subjects, including one with an extruded disc, had
no back pain. Two individuals had pain for one week, and another
had pain for six weeks.
In one of the four subjects in whom stenosis of the lumbar canal
had been the sole abnormality in 1989, repeated scans showed that
a protruded disc had developed at the level of mild stenosis. During
the seven-year period, stenosis of the canal developed in nine additional
individuals, four of whom had a concomitant herniated disc. Stenosis
was mild to moderate in nine individuals, and one of the nine also
had severe stenosis at another level but no pain. A single level
was involved in six subjects, and three levels were involved in four.
None of the subjects with only spinal stenosis on the scans made
in 1996 had pain.
A disc bulge was seen on the 1996 scans of twenty (65%)
of the thirty-one subjects and was the sole finding in eight subjects
(26%). New abnormalities had developed in eight individuals
during the study period. Three of these individuals with no other
abnormality had back pain lasting six weeks or longer. A single
disc level was involved in one, and two levels or more were involved
in the remaining subjects. The twelve subjects with disc bulges
noted in 1989 had progression of bulging, and there was involvement
at additional interspaces in seven individuals.
Disc degeneration was noted in twenty-three of the thirty-one subjects
who had scans made in 1996. Eighteen individuals had involvement
at two or more levels.
The average severity of involvement increased significantly in all
four magnetic resonance categories. Fifteen individuals with disc
degeneration and fifteen with a bulging disc had the greatest increases
(p < 0.0001). An increase of one grade or more was noted
in five subjects with a herniated disc (p < 0.003). These
anatomical alterations were identified by all three evaluators of
the magnetic resonance images. The duration of pain in these subjects
ranged from no pain to one week of pain. Eight subjects had an increase
in spinal stenosis of one grade or more (p < 0.0002). The
duration of pain in these individuals ranged from no pain to six
weeks of pain. During the seven-year period, four subjects, ranging
in age from thirty-nine to forty-seven years, had no progressive
change in any category of abnormality of the intervertebral discs.
Three additional individuals, who were forty-seven, fifty-two, and fifty-seven
years old, had only mild degenerative changes at a single disc level.
None of these seven individuals had low-back pain during the study
period.
There was radiographic improvement of some of the intervertebral
disc abnormalities; however, there was only one instance in which
all three evaluators reached a consensus with regard to a one-grade
improvement in the rating for disc bulging.
Predictive Value of Magnetic Resonance Scans
With regard to the ability of magnetic resonance scans to predict
low-back pain, a positive trend was noted in each category but none
were significant.
The relationship between low-back pain and the presence of magnetic
resonance abnormalities on the 1996 scans was investigated. A significant
correlation between the duration of low-back pain and the presence
of a herniated nucleus pulposus (p = 0.01) or moderate
degenerative disc changes (p = 0.04) was found. Although
the presence of a herniated disc or degenerative disc disease on
the original (1989) scan did not predict the development of back
pain, there was a clinical correlation between back pain and the
simultaneous existence of disc herniation and degeneration. The
relative risk that low-back pain would develop in individuals with
worsening abnormalities on magnetic resonance scans was 3.5.
Agreement Among Evaluations of Magnetic Resonance Scans
The three evaluators of the 1989 and 1996 magnetic resonance
scans agreed most frequently on the anatomical changes associated
with disc degeneration (k = 0.801) and disc herniation
(k = 0.633) in 1989. The greatest degrees of disagreement,
with kappa averages of <0.5, included stenosis of the canal
in 1989 (k = 0.4341) and 1996 (k = 0.448), disc herniation
in 1996 (k = 0.244), and disc bulges in 1989 (k = 0.362).
Clinical data gathered from the questionnaires revealed that four
of the fifty subjects missed an average of twelve days (range, one
to twenty-one days) of work. No patient had persistent pain at the
time that the questionnaire was completed. A total of nine new specific
diagnoses were made during the seven-year study period. The diagnoses
included back strain (three subjects), sciatica (two), osteoarthrosis
(two), lumbar radiculopathy (one), and sacroiliac joint dysfunction
(one).
Correlation between the symptoms and signs and the magnetic resonance
findings is necessary to determine the clinical importance of anatomical
abnormalities identified by this radiographic technique. Questions
remain concerning whether radiographic findings can be used to predict
the development of future low-back pain in individuals with anatomical
abnormalities of the lumbar spine. A study by Wilberger and Pang suggested
that asymptomatic individuals with disc herniations identified on
spinal myelography may be at risk for low-back pain10. They reported that symptoms of
lumbosacral radiculopathy developed in 64% of 108 patients
with asymptomatic lumbar disc herniations who were followed for
three years.
The present study demonstrated the inability of magnetic resonance
scans of the lumbar spine to predict the development of low-back
pain over a seven-year period. Magnetic resonance scans confirm
the presence of anatomical abnormalities that are suspected on the
basis of the history and physical examination of a patient. Our
study suggests that the anatomical findings most closely associated
with clinical symptoms and duration of back pain are a herniated
nucleus pulposus and moderate degenerative disc disease.
Low-back pain developed in a total of twenty-one of the fifty subjects
over a seven-year period. The largest group of individuals in whom
low-back pain developed during this period consisted of those who
had had no specific magnetic resonance abnormalities in 1989. The
predictive value of magnetic resonance imaging was diminished by
this large group of subjects who had pain but no anatomical abnormalities
on magnetic resonance imaging.
In five of the six individuals with an extruded or protruded disc
on the 1989 magnetic resonance scan, pain developed in the back
(three subjects) or in the leg (two subjects). A positive trend
was noted between the presence of a herniated disc and the development
of pain (p = 0.2), but it was not significant with the
numbers evaluated. Individuals with a herniated or protruded disc
appear to be at risk for the development of pain.
A clinical association was found between the development of back
or leg pain and the presence of a herniated nucleus pulposus or
disc degeneration. Four of the eight patients with leg pain had
an extruded disc or moderate spinal stenosis on the 1989 scans.
The leg pain corresponded with these magnetic resonance findings.
However, one subject in whom an extruded disc developed and two
subjects in whom moderate or severe lumbar stenosis developed during
the study period remained pain-free.
With regard to the progression or regression of magnetic resonance
abnormalities, repeat magnetic resonance scans demonstrated a general
progression in all categories of spinal abnormalities. Of the five
extruded discs, one resolved entirely, two regressed to a protruded
status, and two remained extruded. Stenosis and degeneration did
not resolve during the seven-year study period. Of the five individuals with
a repeat scan who had had a diffuse bulge on the 1989 scan, one
had severe stenosis when they were reexamined in 1996.
The present study has a number of limitations. Seventeen of the
sixty-seven subjects did not complete a questionnaire (or answer
questions regarding back pain). The loss of younger individuals
in that group may have biased the data by decreasing the number
of subjects with a herniated disc who had pain compared with those
with a herniated disc who may have been asymptomatic. Herniated
discs were identified in younger subjects in this study. The inclusion
of these younger subjects who were lost from the 1989 study may
have increased the number of herniated discs identified by magnetic
resonance imaging in 1996. The predictive power of magnetic resonance
imaging in the present study might have been increased if these
younger individuals had remained in the study. Another possibility
is that younger individuals did well and chose not to participate
because of an absence of symptoms. Another limitation of the study
is the fact that patients were asked questions concerning back pain
after a seven-year hiatus. The recall of the patients with regard
to the duration or location of pain may have been faulty. Also,
two different magnetic resonance machines were used for this study.
Magnetic resonance scans from the same machine would have been preferable;
however, both machines contained high-field-strength magnets. Finally,
the individuals who evaluated the original (1989) magnetic resonance
scans were not the same ones who interpreted the 1989 and 1996 scans
in the present study.
In conclusion, magnetic resonance scans of the lumbar spine did
not predict the development or duration of low-back pain. Individuals
with leg pain frequently had a spinal lesion in a location on a
magnetic resonance scan that corresponded to the clinical symptoms
and signs. The findings discovered by magnetic resonance technology
can only confirm the clinical suspicions of the clinician. Therapeutic
or prophylactic interventions should not be based solely upon magnetic
resonance abnormalities in the absence of clinical indicators.
A table showing the age and gender of the patients, the findings
on the 1989 and 1996 magnetic resonance images, the location and
duration of pain, and the location of the abnormalities is available
with the electronic versions of this article, on our web site at
www.jbjs.org (go to the article citation and click on "Supplementary
Material") and on our CD-ROM (call 781-449-9780, ext. 140,
to order).
Note: The authors thank Sam Simmens for his statistical analysis
of the data presented in this article.
Hitselberger WE,Witten RM. Abnormal myelograms in asymptomatic patients. J Neurosurg,1968;28: 204-6. 28204
1968
[PubMed]
Holt EP Jr. The question of lumbar discography. J Bone Joint Surg Am,1968;50: 720-6. 50720
1968
[PubMed]
Wiesel SW, Tsourmas N, Feffer HL, Citrin CM,Patronas N. A study of computer-assisted tomography. I. The incidence
of positive CAT scans in an asymptomatic group of patients. Spine,1984;9: 549-51. 9549
1984
[PubMed]
Boden SD, Davis DO, Dina TS, Patronas NJ,Wiesel SW. Abnormal magnetic-resonance scans of the lumbar spine
in asymptomatic subjects. A prospective investigation. J Bone Joint Surg Am,1990;72: 403-8. 72403
1990
[PubMed]
Jensen MC, Brant-Zawadzki MN, Obuchowski N, Modic MT, Malkasian D,Ross JS. Magnetic resonance imaging of the lumbar spine in people without
back pain. N Engl J Med,1994;331: 69-73. 33169
1994
[PubMed]
Modic MT. Lumbar disc herniation: Natural history and imaging. Semin Spine Surg,1997;9: 2-8. 92
1997
Powell MC, Wilson M, Szypryt P, Symonds EM,Worthington BS. Prevalence of lumbar disc degeneration observed by magnetic resonance
in symptomless women. Lancet,1986;2: 1366-7. 21366
1986
[PubMed]
Weinreb JC, Wolbarsht LB, Cohen JM, Brown CE,Maravilla KR. Prevalence of lumbosacral intervertebral disk abnormalities on
MR images in pregnant and asymptomatic nonpregnant women. Radiology,1989;170: 125-8. 170125
1989
[PubMed]
Landis JR,Koch GG. The measurement of observer agreement for categorical
data. Biometrics,1977;33: 159-74. 33159
1977
[PubMed]
Wilberger JE Jr,Pang D. Syndrome of the incidental herniated lumbar disc. J Neurosurg,1983;59: 137-41. 59137
1983
[PubMed]