We prospectively collected clinical, magnetic resonance imaging, and
electrodiagnostic data on persons with no symptoms, back pain, or clinical
evidence of spinal stenosis. Subject recruitment was complex, so it is
important to first briefly define the diagnostic categories used in the final
analyses. The final cohort comprised people fifty-five to eighty years of age
who had no contraindications to testing, had no evidence of cancer or
polyneuropathy, and had technically adequate results of magnetic resonance
imaging and electrodiagnostic tests as performed during the study. As will be
described, the final diagnostic categories were determined solely according to
the impression of a physiatrist who obtained a history and performed a
physical examination in a comprehensive and codified but unconstrained manner.
The results of the magnetic resonance imaging and electrodiagnostic testing
were masked for this clinician. The diagnostic categories were (1)
asymptomatic, (2) mechanical low-back pain, and (3) clinical spinal stenosis,
which included stenosis with or without low-back pain.
One arm of the recruitment began with advertisement in the community for
persons between the ages of fifty-five and eighty years old who claimed to
have no low-back pain. Potential subjects were screened by telephone for
exclusion criteria, including known polyneuropathy, diabetes, heavy alcohol
use, previous lumbar surgery, or relative contraindications to magnetic
resonance imaging or electrodiagnosis.
The second and third arms of the recruitment involved a review of serial
lumbar magnetic resonance imaging reports on patients who had been referred to
a university imaging center. Potential subjects were screened to ensure that
they were between fifty-five and eighty years of age and that they did not
meet any of the exclusion criteria noted above. The second group of potential
subjects was noted to have some anatomical evidence of spinal stenosis in the
radiologists' reports, which was confirmed by a review of the magnetic
resonance images by a study physician. The third group of potential subjects
had no apparent stenosis according to the magnetic resonance imaging reports
or the review of the images. The university's computerized medical records of
the third group of potential subjects were screened to exclude any person
whose record indicated pain that radiated below the knee. Telephone interviews
with the second and third groups of potential subjects excluded persons who
met any of the exclusion criteria noted above. In addition, since the
long-term goal of the project is to follow the natural history of spinal
stenosis, potential subjects who planned to have surgery were excluded.
The university's ethical review board approved the study. All subjects
provided informed consent and were compensated for participating in the study.
Of approximately 322 persons contacted by telephone, 150 (47%) underwent
testing. Others declined or were disqualified clinically.
All consenting subjects filled out a five-page clinical questionnaire and a
number of pain and disability
scales46-50.
Each was tested while walking at a comfortable speed for fifteen minutes, and
all wore a pedometer during waking hours for seven days. Physiatrists, all of
whom were subspecialty board-certified in pain medicine and in
electrodiagnostic medicine or were currently in a fellowship designed to
qualify them for these boards, obtained a spinal history and performed a
physical examination in a comprehensive and codified but unconstrained manner,
as mentioned above. The physiatrist's impression is termed the "clinical
diagnosis" throughout this paper. The validity of the clinical diagnosis
was examined by comparing it with the diagnosis made by a senior spine surgeon
(J.T.H.) who had reviewed the clinical and magnetic resonance imaging data
(but not the electrodiagnostic data) and by comparing it with a number of
clinical attributes thought to be related to spinal stenosis.
All subjects underwent a non-contrast lumbosacral spinal magnetic resonance
imaging scan (GE Signa Horizon LX; GE Medical Systems, Milwaukee, Wisconsin).
Images were reviewed at a workstation (Windows Advantage Workstation; GE
Medical Systems) by a neuroradiologist (V.V.P.) for whom other findings
concerning the patient were masked. Numerous anatomical measurements were made
with use of an electronic cursor. The test-retest reliability of the
measurements was excellent, with intraclass correlation coefficients of
=0.90 on repeated measures after one year.
Electrodiagnosis was performed with a Nicolet Viking II electromyography
machine (Nicolet Biomedical, Madison, Wisconsin). The masking technique,
detailed elsewhere, resulted in <6% of the electrodiagnostic tests having
any potential for clinically relevant
unmasking43. Nerve
conduction studies included bilateral H-waves and peroneal F-waves as well as
sural sensory and peroneal motor studies on the symptomatic side. All were
performed under temperature-controlled conditions and with use of standard
techniques. The needle examination included five limb muscles on the most
painful side or, if there was no difference between sides, the side chosen by
a coin toss51. In
each muscle, fibrillations were scored from 0 to 4+ after twenty-four
insertions52.
Informal motor unit analysis was performed on ten motor units per muscle.
Since most electrodiagnosticians consider motor unit changes a
"soft" finding, we did not include motor unit changes as a primary
outcome measure for this study.
The MiniPM abbreviated paraspinal mapping protocol was used for the
paraspinal needle
examination38,53.
Paraspinal mapping is an anatomically validated, quantified system for scoring
the paraspinal muscles, with an established range of normals and good
interrater reliability (r = 0.830, p = 0.041). Limited clinical evidence
suggests that it can localize the root level of a
lesion36,40,42,53.
Scores can theoretically range from 0 (no reproducible abnormal spontaneous
activity) to 96 (abnormal spontaneous activity occluding the oscilloscope
baseline in all locations tested). With evidence of higher scores in
asymptomatic older persons, we used an a priori cutoff of >4 as abnormal
for this
study38,54,55.
One year after testing, eight subjects had polyneuropathy or myopathy and
one subject had sacral cancer (not detected on the magnetic resonance imaging
performed for the study); all were excluded from the final analysis.
Inadequate or missing magnetic resonance imaging or electrodiagnostic findings
resulted in exclusion of fifteen other patients. Thus, the final cohort
included 126 persons, divided into three groups: clinical spinal stenosis
(fifty-one patients), mechanical low-back pain (forty-four), and asymptomatic
(thirty-one).
Data were initially entered in a Microsoft Excel database (Microsoft,
Redmond, Washington), where errors were checked and corrected. SPSS version
12.0 (SPSS, Chicago, Illinois) was used for statistical analysis. Group
differences in the various clinical measures were examined with either one-way
analysis of variance (for three-group comparisons) or a t test (for two-group
comparisons) for continuous measures and with a chisquare test of independence
for categorical measures. When analysis of variance was used, post hoc tests
with use of the Tukey honestly significant difference as the criterion were
conducted to determine how each of the three groups differed. Discriminant
function analyses were used to examine the ability of the canal size and
electrodiagnostic measures to predict the diagnosis given by the physiatrist.
A stepwise discriminant analysis was performed to statistically determine the
best combination of predictors. Significance was accepted at p < 0.05.
Additional details of the magnetic resonance imaging, electrodiagnosis, and
statistical methods are available in the Appendix.
The group with clinical spinal stenosis significantly differed from
the asymptomatic group (p < 0.05) in terms of lumbar tenderness, strength
deficits, nerve-root tension signs, pain severity, pain below the knee, and
walking deficits. They differed significantly (p < 0.05) from the group
with mechanical back pain in terms of pain below the knee, reflex deficits,
and nerve-root tension signs.
Although the spine surgeon reviewed the clinical data and examined the
magnetic resonance images and the clinical examiner did not examine the
magnetic resonance images, the diagnosis made by the clinical examiner had a
moderate relationship to the surgeon's diagnosis (kappa = 0.425, p <
0.001). Of the fourteen measurements made on the magnetic resonance images,
only two—the minimum anteroposterior canal diameter and the average of
the two smallest anteroposterior canal diameters—had significant mean
differences across the groups (F[2,123] = 4.26, p = 0.016, and F[2,123] =
5.040, p = 0.008, respectively).
Tables I and
II compare key
electrodiagnostic and magnetic resonance imaging findings between the group
with clinical spinal stenosis and each of the other two groups (mechanical
low-back pain and asymptomatic). The mean minimum spinal canal diameter was
15.03 mm in the asymptomatic subjects. One standard deviation below the mean,
a predetermined cutoff that we set for subsequent analyses, was 11.93 mm. This
cutoff is not dissimilar to the 90% cutoff of 13.0 mm used by Adamova et al.
in a study of twenty-four patients with
diabetes8. A more
stringent cutoff of two standard deviations below the mean (8.83 mm) would
have resulted in the measurements on the magnetic resonance images correctly
characterizing only five of the fifty-one persons thought clinically to have
stenosis.
Table I shows that the
findings on the magnetic resonance imaging could not differentiate persons
with clinical spinal stenosis from persons with mechanical low-back pain. This
was true whether the analysts compared mean values or used the 11.93-mm cutoff
for distinguishing normal from abnormal. In contrast, every electrodiagnostic
measure involving fibrillation potentials showed a significant difference
between patients with clinical spinal stenosis and those with mechanical back
pain or the asymptomatic controls. Motor unit changes and H-wave absence were
not associated with clinical spinal stenosis.
A significant main effect of group was demonstrated in a one-way analysis
of variance performed on smallest canal size and on paraspinal mapping scores
(Table III). Post hoc tests
with use of the Tukey honestly significant difference as the criterion
indicated that the subjects with clinical spinal stenosis had significantly
smaller canals than the asymptomatic subjects. No other groups differed
significantly from each other, with the numbers available. Abnormal findings
on electromyographic examination of the limb were also significantly related
to the diagnosis of clinical spinal stenosis.
We determined the ability of each of the measures to distinguish subjects
with clinical spinal stenosis from those with mechanical back pain or
asymptomatic subjects (Tables
IV and
V). Canal size, paraspinal
mapping, and needle examination of the limb all could be used to discriminate
persons with clinical spinal stenosis from those without any symptoms.
Paraspinal mapping and needle examination of the limb, but not canal size,
marginally discriminated the group with clinical spinal stenosis from the
group with mechanical back pain. Paraspinal mapping was more successful than
chance in predicting clinical spinal stenosis, but with the numbers studied
the needle examination of the limb and the spinal canal size did not predict
the group to which the patient belonged. Paraspinal mapping had a sensitivity
of 45.1% and a specificity of 84.1% in the study population.
Discriminant function analysis employing stepwise entry criteria (p <
0.05) was used to examine the ability of canal size, paraspinal mapping, and
needle examination of the limb to predict assignment to either the group with
clinical spinal stenosis or to a combined control group of asymptomatic
persons and those with mechanical back pain. This procedure constructs a
formula that uses variables to predict the group to which a subject belongs
(in this case, the diagnosis based on the results of the diagnostic testing)
and compares that group with a so-called gold standard (in this case, the
clinical diagnosis for the subject). Both needle examination of the limb and
canal size fulfilled the entry criteria (were significantly associated with
group membership), while paraspinal mapping scores were not significantly
associated with group membership after needle examination of the limb and
canal size were entered. The resulting discriminant function significantly
predicted group membership (chi square (1) = 14.23, p = 0.002). The
standardized canonical discriminate function coefficients indicated that the
needle examination of the limb again made the greatest contribution to the
prediction of group membership (0.814), followed by canal size (-0.646). The
discriminant function accurately classified 61.9% of the original cases;
specificity was 49.0%, while sensitivity was 70.7%. However, the
classification rate was not significantly better than that resulting from
chance.
Figures 1 and
2 provide a more clinically
intuitive demonstration of the sensitivity and specificity of different
choices for spinal canal and paraspinal mapping measures. Because limb
fibrillation potentials cannot be plotted in a linear fashion, they are not
included in these figures; however, as noted above, they would tend to
increase sensitivity moderately or slightly decrease specificity compared with
paraspinal mapping alone. The Appendix includes more detailed statistical
analyses.
The term "spinal stenosis" is used both for a measure of
the dimensions of the spinal canal and for a clinical syndrome. The study
results suggest that the relationships between these two are not strong enough
to be clinically useful. In contrast, electrodiagnostic testing can confirm
the presence of this disabling condition. It can also detect neuromuscular
disorders that might mimic stenosis. We believe that this is the first
adequately masked study of needle electromyography and that it included the
largest population of asymptomatic older persons to be tested with magnetic
resonance
imaging35.
The control populations in this study included a group of asymptomatic
persons. Since physicians do not typically evaluate asymptomatic subjects in
the clinic, we also included persons who appeared to have only mechanical
low-back pain. Although the anteroposterior diameters of the spinal canals
ranged from 5.6 to >20 mm, our exclusion of persons who planned to have
surgery means that the relationships found in this study might not apply to
persons with a very small canal or very severe clinical disease. This remains
to be proven. Since our group of subjects with clinical spinal stenosis ended
up being older than the asymptomatic volunteers, it is possible that even the
weak relationships found between radiographic evidence of spinal stenosis and
symptoms may have been an artifact of the age difference.
A physician's clinical impression alone is not considered to be the gold
standard for the diagnosis of the clinical syndrome of spinal stenosis.
However, its relationship in this study to factors such as leg pain and loss
of reflexes as well as its relationship to the independent surgeon's
impression suggest that it is a reasonable assessment that probably reflects
the judgment of other physicians.
Electrodiagnosis as a test for spinal stenosis was further validated by an
analysis of forty-eight of our subjects for whom the diagnostic category had
been agreed on by the radiologist, examining clinician, and spine surgeon
(with the findings of each masked for all of the
others)44. In that
analysis, electrodiagnosis proved to be highly specific and moderately
sensitive for revealing the consensus diagnosis of spinal stenosis.
Paraspinal mapping, a newer technique devised by our group, was a key
component of the electrodiagnostic examination. We have shown that paraspinal
examination with a more simplistic method results in irrational
results56. Data
from the current study have been used to show that paraspinal mapping
diminishes examiner bias compared with that associated with needle examination
of the limb57. We
believe that, in the future, one should consider using paraspinal mapping to
improve the objectivity of clinical testing and certainly of electrodiagnostic
research.
Electrodiagnostic data are of value in two ways. First, fibrillation
potentials discriminate clinical spinal stenosis from other disorders with
excellent specificity and moderate sensitivity. Whether the
"missed" cases were actually those of persons with an insufficient
spinal canal can be debated. Second, the post hoc exclusion of eight persons
with neuromuscular diseases indicates that alternative or comorbid disorders
are common and are not reliably detected on clinical examination. These
disorders are certainly not detectable on magnetic resonance imaging.
Authors of previous small studies of magnetic resonance imaging scans of
asymptomatic older persons did not report actual measures; they reported
subjective impressions. Boden et al. detected stenosis in three of fourteen
volunteers between the ages of sixty and eighty but did not comment on any
specific measurement or level of
severity18. Jensen
et al. found that 67% of twenty-seven asymptomatic persons over the age of
fifty years had multiple disc abnormalities, but they did not attempt to grade
central stenosis19.
Tong et al. analyzed our data and found that radiologists reported
degenerative changes in 85% of the thirty-one asymptomatic
subjects58.
One might contend that the gestalt impression of a neuroradiologist, as
used in the previous studies described above, is more critical than individual
objective measures. However, in another analysis, we found that the
radiologist's impression had no relationship to the clinical impression (p
= 0.80)59.
We do not question that the spinal canal can become too small for its
occupants. However, other factors come into play. Segmental hypermobility
(perhaps related to, and exacerbated by, paraspinal atrophy or spontaneous
denervation60,61),
local neurovascular compromise, venous congestion, foraminal stenosis, or
facet joint effusion may also be important. Given the lack of specificity of
imaging for the clinical syndrome of spinal stenosis in the current study, the
clinician should interpret radiographic findings in light of the clinical
picture.
We believe that the results of this study suggest a need for separate
terminology for the benign anatomical variation and the disabling disease.
Radiologists will undoubtedly continue to use the word "stenosis"
to describe the anatomy. Treating physicians should recognize that this
terminology does not indicate the presence or severity of symptoms. We have
found that the most accurate objective pathophysiological correlate with the
clinical syndrome of spinal stenosis is paraspinal denervation. Consequently,
we prefer the term "paraspinal denervation syndrome" rather than
"spinal stenosis."
A table presenting the measurements on magnetic resonance imaging as well
as text providing additional details of the methodology and the statistical
analyses are available with the electronic versions of this article, on our
web site at
(go to
the article citation and click on "Supplementary Material") and on
our quarterly CD-ROM (call our subscription department, at 781-449-9780, to
order the CD-ROM). ?