Magnetic resonance imaging in low back pain: general principles and clinical issues.

Key Words: Back pain, Degenerative disk disease, Diagnosis, Intervertebral disk, Magnetic resonance imaging.

The clinical manifestations of low back pain (LBP) result from a complex interaction of biologic, psychologic, and sociologic factors, each of which must be addressed in the diagnostic process.[1-4] Although the linkage is often uncertain, anatomic and physiologic impairments are typically the primary mechanism for triggering and potentiating the symptoms of LBP.[3-6] Fundamental to detecting and quantifying these impairments is an accurate visualization of the morphology of the lumbar spine with high-resolution, in vivo images that allow examiners to detect morphologic variations consistent with disease, trauma, or aging.

Although numerous modalities are available, magnetic resonance imaging (MRI) has emerged as the procedure of choice for the diagnostic imaging of most painful conditions involving the lumbar spine.[7-10] Because lumbar MRI is noninvasive and does not require the use of ionizing radiation, it is safer than conventional plain film radiography or other commonly used procedures such as computerized tomography, myelography, diskography, or radionuclear bone scanning.[8,10,11] The high degree of soft tissue contrast and excellent spatial resolution characteristics of MRI have allowed examiners to observe lumbar anatomy in precise detail and to detect morphologic and biochemical abnormalities that were not observable previously. This dramatic increase in detection sensitivity has provided invaluable data for diagnosing degenerative, infectious, inflammatory, neoplastic, and metabolic disorders involving the lumbar spine.[2,12,13] Additionally, it has assisted in the generation of hypotheses relating to the role of many musculoskeletal abnormalities in the production of LBP or radiculopathy.[14-25] Although cost issues may sometimes limit its utilization, serial MRI provides an excellent mechanism to study the natural history of conditions such as disk herniation, degenerative disk disease, and spinal stenosis.[14,17,20,23] Magnetic resonance imaging also allows the assessment of changes in morphology resulting from various interventions.

As the use of lumbar MRI becomes more widespread, physical therapists are becoming increasingly interested in the interpretation of lumbar MRI findings and the role of lumbar MRI in clinical decision making. The goal of this article is to enhance the reader's understanding of MRI and its current role in the diagnosis and classification of patients with LBP. Topics to be discussed include the basic principles of MRI, normal lumbar MRI anatomy, and the concept of clinical usefulness as it relates to the use of MRI in the diagnosis and management of individuals with LBP.

General Principles of MRI

The MRI Signal

Magnetic resonance imaging provides high-resolution, multiaxial, multiplanar images that represent thin "slices" of tissue.[10,11] Although there are no known biohazards associated with the clinical use of MRI, contraindications include the presence of any mobile ferromagnetic implant in the orbit, skull, or spinal canal, as well as the presence of cardiac pacemakers, intracranial clips, or claustrophobia. Stable metallic implants, such as a prosthetic femoral neck and head, do not represent contraindications; however, the hardware distorts the magnetic field and dramatically reduces the quality of the image.

The primary components of an MRI system used for low back imaging include a large magnet (scanner), a radio frequency (RF) surface coil, and a minicomputer (Fig. 1).[11] The magnet creates a static, homogeneous magnetic field, which is referred to as [B.sub.o]. The strength of this field is expressed in units of tesla (T) and ranges from low field (~0.3 T) to high field (~1.5 T). Within the magnet are a series of "gradient coils," which act to create additional magnetic fields or gradients that assist in the encoding of signal data. The RF surface coil acts to emit and receive RF pulses traveling to and from the patient's lumbar region and is interfaced with a minicomputer that encodes the MRI "signals" into an anatomic image. This image is displayed on a video monitor.

[Figure 1 ILLUSTRATION OMITTED]

To create an MRI, a patient is positioned supine on a movable table with the RF surface coil under the lumbar spine. The patient is then placed into the magnet, at which time the effect of the magnetic field alters the normal random alignment of the spinning hydrogen nuclei within the patient's body by causing them to align themselves in a parallel or antiparallel direction to one another. The coil then emits a low-power RF pulse that causes the hydrogen nuclei to absorb energy and be pushed out of their parallel or antiparallel alignment. After the pulse is removed, the hydrogen nuclei realign themselves with the magnetic field while giving off energy in the form of an electromagnetic signal. This energy is received by the RF coil and transmitted to the minicomputer, where it is amplified and transformed into an image by a process known as 2-dimensional or 3-dimensional Fourier transformation.[26]

The signal from each anatomic structure varies as a function of the morphology and hydration of the tissues being imaged; thus, an image is created that illustrates the contrast between different tissues.[10,11,26] In an attempt to optimize the signal-to-noise ratio, optional RF mechanisms are used to reduce artifact from the pulsation of blood within the great vessels of thorax and abdomen and the pulsation of cerebrospinal fluid (CSF) within the spinal cord.[27] Improved contrast also may be obtained using paramagnetic contrast agents such as gadolinium, which are administered by intravenous injection prior to the examination.[28]

The signal intensity of anatomic structures will vary based on MRI scanner settings. To provide for the contrast necessary to detect a variety of abnormalities, a standard clinical lumbar MRI scanner uses a "spin echo" sequence. With this method, images are obtained with a variety of operator-selected scanner settings to accentuate different tissue characteristics based on biochemical properties referred to as [T.sub.1], [T.sub.2], and proton density (Tab. 1).[26] This process, known as "image weighting," can be achieved by adjusting the way the signal is obtained. The 2 primary variables are the time to echo or echo time (TE) and the repetition time (TR). When evaluating MRI results, the nature of the weighting can be determined by identifying the TE and TR for each image.

Table 1.
Signal Characteristics Based on Image Weighting(a)

                                       Proton
Structure             [T.sub.1]        Density        [T.sub.2]

Fat                    High            High            Low
Marrow                 High            High            Low
Cerebrospinal fluid    Low             Intermediate    High
Nerve                  Intermediate    Intermediate    Intermediate
Muscle                 Intermediate    Intermediate    Intermediate
Hyaline cartilage      Intermediate    Intermediate    Bright
Fibrocartilage         Low             Low             Low
Ligaments              Low             Low             Low
Nucleus pulposus       Intermediate    Intermediate    High

(a) A high signal is bright, an intermediate signal is gray, and a low signal is black.[8,10,27]

[T.sub.1]-weighted image. [T.sub.1]-weighted images are formed using a "short" TE (20-30 milliseconds) and a "short" TR ([is less than] 700 milliseconds) (Tab. 2). [T.sub.1]-weighted images result in high (bright) signals from bone marrow and fat and relatively low (dark) signals from cortical bone and CSF (Fig. 2A). Good contrast is provided between the intervertebral disk (IVD) and CSF. [T.sub.1]-weighted images are used to contrast tissues with different [T.sub.1] characteristics, such as neural foramina and their contents (eg, nerve roots).

Table 2. Duration (in Milliseconds) Defining T[sub 1], T[sub 2], and Proton Density Weighting

                                    Repetition
                   Echo Time        Time

[T.sub.1]          Short (20-30)    Short (~500)
[T.sub.2]          Long (60-100)    Long (2,000+)
Proton density     Short (20-30)    Long (2,000+)

[Figure 2 ILLUSTRATION OMITTED]

[T.sub.2]-weighted image. [T.sub.2]-weighted images are formed using "long" TE (60-100 milliseconds) and "long" TR ([is greater than] 2,000 milliseconds) (Tab. 2). These images have lower signal-to-noise ratios than [T.sub.1]-weighted or proton density-weighted images but provide high contrast between tissues as a function of fluid content. This contrast is often referred to as a "myelographic effect."[9,27] Most diseases and pathologic conditions are associated with increased free water such as is contained in blood, pus, or synovial fluid and, therefore, have high signals on [T.sub.2]-weighted images. Because of the regional differences in hydration of the IVD, [T.sub.2]-weighted images are optimal for contrasting the nuclear and annular regions (ie, the normal nucleus pulposus has a moderately high signal intensity on [T.sub.2]-weighted images, whereas the annular region has a low or dark signal) (Fig. 2B).[10,27] Thus, [T.sub.2]-weighted images are quite useful for detecting and quantifying degenerative disk disease and its associated impairments.[29]

Proton density-weighted image. Proton density-weighted images can be conceptualized as an intermediate weighting because they are generated using a "short" TE (20-30 milliseconds) and a "long" TR (2,000 milliseconds) (Tab. 2).[9] Because of their excellent signal-to-noise ratio, these images often have high clarity and are useful for observing anatomy (Fig. 2C).

Viewing an Image Obtained with Lumbar MRI

To view an image obtained with MRI, an examiner should orient himself or herself to the anatomic location and position from which the view was obtained. As with most diagnostic imaging, to allow examiners to appreciate 3-dimensional relationships, a standard MRI lumbar spine series includes views of at least 2 anatomic planes of a given area. Although various oblique planes may also be imaged, the standard views include sagittal, axial (transverse-plane), and occasionally coronal (frontal-plane) views. A major advantage of MRI is the ability to image thin (3- to 5-mm) tissue "slices," thus allowing the examiner to visualize small or asymmetric structures, such as the neuroforamina, without the superimposition of adjacent structures. To assist examiners in determining which slice is represented in a given image, anatomical localizers (ie, images that label the slices) are typically included for each plane that is imaged (Fig. 3).

[Figure 3 ILLUSTRATION OMITTED]

Normal Lumbar MRI Anatomy

Numerous structures are visible on images obtained with lumbar MRI (Tab. 3)[9,10,27] The plane that is imaged and the tissue characteristics that are highlighted (weighted) are the fundamental concerns when viewing these images.

Table 3. Structures Visible on Midsagittal and Parasagittal Regions and Axial Views Transecting the Vertebral Body and the Intervertebral Disk

Midsagittal view

Vertebral body: trabecular bone, nutrient foramen, end plate Intervertebral joint: nucleus pulposus, annulus fibrosus Vertebral foramen and its contents: thecal sac conus

medullaris, cauda equina, cerebrospinal fluid, epidural fat Ligaments: anterior and posterior longitudinal ligaments,

"ligamentum flavum, interspinous ligaments Spinous process

Parasagital view

Intervertebral foramen and its contents: spinal nerve, radicular

vessels, forminal fat Facet joint: articular cartilage, joint space, capsule

Axial through pedicles (transpedicular)

Pedicle and vertebral body relationship Anterior and posterior longitudinal ligaments Vertebral foramen and its contents: thecal sac conus

medullaris, cauda equina, cerebrospinal fluid, epidural fat Lumbar muscles: Erector spinae, multifidus, quadratus lumborum Psoas muscle

Axial inferior to pedicles

Relationships of exiting nerve roots/spinal nerves Intervertebral foramen

Axial through middle of disk

Annulus fibrosus, nucleus pulposus Relationship of spinal nerves lateral to intervertebral foramen

Sagittal images: [T.sub.1]-weighted. Sagittal views include slices from the left to the right, which are typically at 3- to 5-mm intervals. Parasagittal views illustrate the lateral aspect of the vertebral bodies, the intervertebral foramina and their contents, the pars interarticularis, and the erector spinae and multifidus muscles (Fig. 4). Midsagittal views show numerous structures. Anteriorly, the vertebral body's shape and alignment and the IVDs, end plates, and longitudinal ligaments can be seen. In the middle portion, the spinal canal and its contents, including the spinal cord (at and above L1), nerve roots of the cauda equina (caudal to L1), CSF within the subarachnoid space, and epidural fat, are present. Posteriorly, the ligamentum flavum, spinous processes, interspinous and supraspinous ligaments, subdermal fat, and dermis are present (Fig. 2A).

[Figure 4 ILLUSTRATION OMITTED]

With [T.sub.1] weighting, adipose tissue has a high signal. Figure 2A gives examples of fat tissue in the spine such as epidural fat, subcutaneous fat, and fat associated with bone marrow in the vertebral bodies. With parasagittal images, fat is also noticeable in the intervertebral foramina (Fig. 4).

Intermediate (gray) signals are associated with the conus medullaris and the nerve roots of the cauda equina. The IVD also has a relatively homogeneous intermediate signal, making differentiation between the nucleus and the annulus difficult on [T.sub.1]-weighted images.

Cerebrospinal fluid within the subarachnoid space has a low signal on [T.sub.1]-weighted images. Cortical bone and the longitudinal ligaments both have low signals, making distinction between them somewhat difficult. Interspinous ligaments and muscle also have low signals. The spinal nerves, visible on parasagittal views, also give rise to low signals.

Sagittal images: [T.sub.2]-weighted. On sagittal [T.sub.2]-weighted images, the conus medullaris has a low to intermediate signal but is easily contrasted with the high signal from the CSF (Fig. 2B). The nuclear (or centroid) region of the IVD normally has a high signal in contrast to the low signal from the annulus fibrosus (Fig. 2B). Thus, [T.sub.2] weighting is optimal for studying IVD abnormalities.

An interesting finding commonly observed on midsagittal [T.sub.2]-weighted images relates to a roughly triangular bright signal in the posterior midportion of the vertebral bodies in the lumbar spine. This signal corresponds to the basivertebral venous plexus, which drains the well-vascularized vertebral body (Fig. 2B).

Ligaments and muscles generate a low to intermediate signal on [T.sub.2]-weighted images. Adipose tissue such as epidural fat and that associated with bone marrow within the vertebral body generates an intermediate signal on [T.sub.2]-weighted images as compared with a high signal on [T.sub.1]-weight images.

Axial images: [T.sub.1]-weighted/proton dens#y-weighted. Axial or transverse section views provide a cross-section of the lumbar spine at various levels. By comparing axial views at adjacent levels, a nerve root can be followed along its entire pathway while assessing its morphology and relationship to other structures.

This image clearly shows the vertebral foramen and its contents and the facet joints. A transpeduncular view transects the vertebral body and the pedicles and is useful for observing abnormalities of these bony structures (Fig. 5). The muscular components of the lumbar spine and their associated intramuscular and extramuscular fat, as well as the spinal cord or cauda equina and its associated CSF, can be seen on all axial views.

[Figure 5 ILLUSTRATION OMITTED]

Similar to the sagittal views, axial [T.sub.1]-weighted images result in high signals from epidural fat, bone marrow, and subdermal and intramuscular fat. Intermediate signals occur from nerve roots, dorsal root ganglia, the IVD, and the ligamentum flavum. Low signals are associated with CSF and cortical bone.

Axial images: [T.sub.2]-weighted. On axial [T.sub.2]-weighted images, the conus medullaris (low signal) is well contrasted with the CSF (high signal). The nuclear (or centroid) region of the IVD images brightly in contrast to the low signal from the annulus fibrosus. Thus, axial [T.sub.2]-weighted images provide excellent views to appreciate the relationship of IVD abnormalities to neural structures within the vertebral and intervertebral foramina (Fig. 6).

[Figure 6 ILLUSTRATION OMITTED]

The alignment and morphology of the facet joints can be seen on axial views. The articular cartilage associated with the facet joints gives rise to an intermediate signal on [T.sub.2]-weighted images. Epidural fat and vertebral body bone marrow also generate an intermediate signal on [T.sub.2]-weighted images as compared with a high signal on [T.sub.1]-weighted images.

Indications for Lumbar MRI

The utilization of lumbar MRI varies among practitioners.[2,30,31] As with any diagnostic test, one must consider the yield of the test in light of the patient's risk factors for the condition of interest.[31-32] Considering the favorable natural history, of "nonspecific" LBP, the primary indication for lumbar MRI in patients with a recent onset of LBP is to rule out life-threatening conditions.

The Agency for Health Care Policy and Research (AHCPR)[33] has generated guidelines for the utilization of diagnostic tests for patients who have had LBP of less than 3 months' duration. Lumbar MRI was strongly indicated for patients with history, and physical examination findings consistent with neoplastic disorders or infections. "Red flags" indicative of serious pathology included abnormal findings on hematologic testing or urinalysis, as well as either radiographic evidence of the condition in question or a positive radionuclear bone scan. For patients with lower-extremity radiculopathy, the AHCPR guidelines recommended that lumbar MRI be performed only if symptoms persist for more than 4 weeks from the time of onset, coupled with "abnormal" findings from electromyography. Deyo et al[2] argued that in absence of "red flags" on clinical examination, MRI should be used only as an anatomical localizer prior to lumbar surgery.

Herzog et al[30] have advocated a more widespread use for lumbar MRI and recommend that it should be used during the initial work-up in the presence of major trauma, signs of infection, neoplasia, or any progressive neural dysfunction (Tabs. 4 and 5). They also recommended the use of lumbar MRI for any patient with delayed recovery (ie, symptoms persisting longer than 7 weeks).

Table 4. Common Disease Processes Visible on Lumbar Magnetic Resonance Imaging[9,12,13]

Spinal cord

Tumors: primary and secondary Springohydromyelia Spinal cord trauma Ischemia and infarction Vascular malformations Inflammatory and degenerative diseases of the spinal cord Infections

Intradural, extramedullary

Tumor, nerve sheath tumors Infection and inflammation Arachnoiditis Guillain-Barre syndrome

Table 5. Anatomic Variations Frequently Associated With Degenerative Disk Disease[4,9,10,15,18,20,29]

Disk degeneration Annular tears Vertebral end-plate and bone marrow abnormalities Disk protrusion or herniation Free fragments of disk material Vertebral body osteophytes Facet arthrosis Synovial cysts Spinal stenosis Spondylolisthesis

Thus, although there is widespread agreement regarding the importance of MRI for diagnosing serious diseases, the role that lumbar MRI findings play in the evaluation and treatment of people with LBP and radiculopathy that is not associated serious disease is controversial.[2,7,23] For example, several authors[34-39] have identified a high prevalence of lumbar disk abnormalities in individuals without low back symptoms. This controversy has led clinicians to question whether the use of MRI has exaggerated disease prevalence relating to diskogenic LBP.

To explore the role of lumbar MRI in LBP, the following section will discuss concepts of test usefulness and provide examples from the literature relating to imaging findings of disk abnormalities and nerve compression.

Clinical Usefulness of MRI for Patients With LBP and Radiculopathy

The clinical usefulness of a test describes its overall impact and the benefits of its use.[40] It is important to note that clinical usefulness has several dimensions and that a test may be useful at one level but not at another. For example, Kent and Larson[40] have proposed a general framework to determine the clinical usefulness of MRI that has 3 primary dimensions: the disease in question, the quality of the research, and the level of assessment. This framework identifies 4 levels of assessment for MRI: technical capacity, diagnostic accuracy, therapeutic impact, patient outcome.

Technical Capacity

Technical capacity is determined by the ability of a test to measure a given phenomenon. This level of assessment includes the basic measurement properties of the system such as resolution and reproducibility. As with any developing technology, technical capacity has been the primary focus of basic science and clinical research on lumbar MRI. Case study or case series designs are typically used to compare MRI findings with direct observations noted during surgery, postmortem examination, or histochemical analysis of tissue samples. [17,19,24,29,41]

Intervertebral disk degeneration or degenerative disk disease is a global description of a complex process that results in morphologic, biochemical, and mechanical changes within the disk.[4] Strong links between MRI signals and degenerative disk disease have been demonstrated,[8,18,29,42] Two primary impairments are associated with disk degeneration and are readily visible on images obtained with MRI (Fig. 7).[10,29] A reduction in the water content throughout the disk is represented by a low signal corresponding to the nucleus pulposus on [T.sub.2]-weighted images.[10] Evidence of this reduction may range from a speckled appearance of the nucleus[23] to a completely low (dark) signal. An alteration in the shape of the disk, resulting in loss of vertical dimension and an increase in the horizontal dimension that causes bulging to occur, most noticeably in the posterior aspect, also is associated with disk degeneration (Fig. 7).

[Figure 7 ILLUSTRATION OMITTED]

Evidence supports the relationship between other common anatomic impairments associated with disk degeneration and MRI signals. This evidence includes tears of the outer annulus,[15,18,42,43] loss of constraint of nuclear material (noncontained herniation, protrusion, extrusion)[17,19,36] (Tabs. 5 and 6), free fragments of nuclear material (sequestrum),[14] and ligamentous hypertrophy and osteophytes.[27]

Table 6. Operational Definitions for Anatomic Variations in the Intervertebral Disk as Determined by Magnetic Resonance Imaging(a)

Classification Criterion

Normal        No disk extension beyond the interspace
Bulge         Circumferential, symmetric extension of the disk
              beyond the interspace (around the vertebral
              end plates)
Protrusion    Local or asymmetric extension of the disk beyond
              the interspace, with the base against the disk
              of origin broader than any other dimension of
              the protrusion
Extrusion     More extreme extension of the disk beyond the
              interspace, with the base against the disk of
              origin narrower than the diameter of the
              extruding material itself or with no connection
              between the material and the disk of origin

(a) Adapted from Jensen et al.[36]

Establishing reliability of classifications of MRI findings requires clear operational definitions, sufficient sample sizes, independence between readings, and statistical analysis of concordance.[40,44,45] Two recent studies[44,45] have met these criteria. Raininko et al[44] reported moderate intrarater reliability (mean intraclass correlation coefficients [ICCs] and Kappa [[Kappa]] values = .59 - .66) for 3 examiners grading 12 impairments visible on thoracolumbar images obtained with MRI (N = 122 subjects). Interrater reliability was lower (mean ICCs and [Kappa] values = .40 - .51). In the lumbar spine, moderate to high interrater agreement was reported for disk height (ICC = .61 - .69), the presence of disk bulging (ICC= .57 - .61), the presence of osteophytes ([Kappa] = .47 - .60), and the signal intensity of the nuclear complex (ICC = .56 - .67). Poor agreement Was found for interpretation of the annulus ([Kappa] = .30 - .41) and for the presence of small disk herniations ([Kappa] = .30).

Brant-Zawadzki et al[45] reported similar reliability for 2 examiners who classified 625 IVDs on 27 people with LBP and 98 people without low back symptoms. Intrarater Kappa coefficients ranged from .69 to .71, and the interrater Kappa coefficient was .58.

The results of these studies indicate that, when clearly defined classification rules are used, examiners can reliably classify disk abnormalities relating to decreased signal in the nucleus (degenerative disk disease) and differentiate between normal and herniated disks. Thus, evidence of technical capacity exists for these lumbar MRI findings.

Diagnostic Accuracy

Diagnostic accuracy is the ability of a test to differentiate between individuals with and without a given condition.[31,32,40] Several statistical tests may be used to reflect diagnostic accuracy, including measures of sensitivity and specificity, positive and negative predictive values, likelihood ratios, and receiver operating characteristic (ROC) curves. This level of assessment is much more difficult to establish than technical capacity because one must have a mechanism to determine what is a true positive finding and what a true negative finding (ie, a "gold standard"). Determining the level of diagnostic accuracy has been problematic relative to LBP because there is no test that has been universally agreed on to accurately identify the tissues generating pain in the lumbar spine in a given patient.[3,31,32] The diagnostic accuracy of MRI findings in persons with LBP has been investigated as a component of prevalence studies and by symptom-reproduction case series studies.[14,15,17-20,23,30,36-39]

Prevalence studies. Prevalence studies have the advantage of sampling a wide spectrum of individuals and providing information regarding the frequency of abnormalities between people with and without LBP.[40] Numerous studies,[14,15,17-21,34-39,41,46] have provided age-referenced frequencies of MRI findings of disk degeneration, bulging, and herniation and the presence of nerve compression in patients with LBP and adults without low back symptoms.

The following trends emerge:

1. Disk degeneration is more pronounced with advancing age and is a common finding in both patients with LBP and individuals without low back symptoms. There is little or no difference in prevalence between the 2 groups.[34-39]

2. Disk bulging is independent of disk degeneration and is difficult to classify reliably. Disk bulging is prevalent in patients with LBP and in adults without low back symptoms.[18,36]

3. Disk herniation is independent of disk degeneration and, depending on the authors' definition, is present in 20% to 75% of adults with low back symptoms.[19,35.36]

4. Disk extrusion (an extreme extension of disk material beyond the interspace, with its anterior portion more narrow than its posterior portion) is rare in people without low back symptoms (~2%)[36] but is more prevalent in patients undergoing lumbar MRI (26%)[19] (Fig. 8).

[Figure 8 ILLUSTRATION OMITTED]

5. Minor nerve compromise (contact of the disk material with nerve root or thecal sac) is common in individuals without low back symptoms; however, major neural compromise (actual compression) is uncommon in individuals without low back symptoms but common in patients whose symptoms warrant lumbar MRI (~54%).[19,41]

In summary, severe disk extrusion and the presence of compressed nerves have a higher specificity for persons with LBP than do disk degeneration, bulging, and small herniation.

A weakness of prevalence studies is that causal relationships cannot be determined between findings and symptoms. The 2 phenomena may coexist, but whether a specific finding such as a disk herniation is responsible for the patient's symptoms cannot be shown.[40,47] Studies of the prevalence of MRI findings in patients with LBP and individuals without low back symptoms indicate that MRI cannot be used to identify a true positive or true negative finding and that MRI is of limited value in determining diagnostic accuracy.

Pain reproduction as a gold standard. By definition, LBP is a condition based on the report of pain rather than on any observable findings. It can be argued, therefore, that the gold standard used to compare MRI findings with the symptoms of LBP must be a pain-producing stimulus to the specific tissues that are being evaluated. One procedure that has been utilized as a gold standard for identifying painful lumbar IVDs is diskography.[15,18,23,42,43,48] In this procedure, a radiopaque dye is injected into a disk while the examiner assesses the patient's pain response. Ideally, if the patient reports a reproduction in symptoms following injection of dye into a disk, one source of the patient's pain is identified and a positive MRI finding corresponding to that disk would be classified as a "true positive." If injection of dye into a disk does not produce the customary, pain response, negative MRI findings corresponding to that disk would classified as a "true negative."

Kent and Larson[40] stated that studies investigating diagnostic accuracy must have the following characteristics: (1) at least 35 patients to allow appropriate calculation of confidence intervals, (2) independent assessment of disease presence (ie, the result of one test does not influence the examiner's decision on the second test), and (3) avoidance of referral bias (an example of referral bias would be that only those patients with severe disease would undergo the test, thus generating a high index of disease prevalence for the sample).

Buirski et al[18] compared lumbar MRI findings of disk degeneration with or without protrusion in 115 patients with chronic LBP who also underwent diskography and 63 people without low back symptoms who underwent MRI only. The authors reported a very high correlation between the MRI finding of a degenerative disk with associated protrusion and pain reproduction with diskography. There was no difference, however, in the prevalence of disk degeneration with or without protrusion on MRI between the patients and the control group. The authors concluded that there was no signal pattern that indicated whether a disk would be symptomatic.

Although the IVD has the largest aneural and avascular mass of any structure in the body, the outer portion of the annulus fibrosus is vascular and is innervated by pain-sensitive free nerve endings.[4,49,50] This observation has led authors to speculate regarding the outer annulus fibrosus as a source of LBP.[15,18,42,43,48] Tears in the annulus fibrosus may be visible as a zone of high [T.sub.2] signal on midsagittal and axial views of the lumbar IVD. Several authors[15,18,43,48] have used symptom reproduction with diskography to determine the relationships between this signal, known as the high-intensity zone, and symptoms.

Aprill and Bogduk[15] studied 500 patients with intractable LBP and found that 28% had a high-intensity zone at one or more levels. The authors reported a high degree of agreement (66/67) between 2 examiners in identifying this diagnostic sign. The positive predictive value of the high-intensity zone to predict symptom reproduction with diskography was .90.

Schellhas and colleagues,[43] in a study of 63 people with LBP, compared the high-intensity zone with diskography in abnormal versus normal disks within the same patient. Of 100 disks meeting the criterion for annular tear, 87 disks were painful during diskography. Of the 67 control disks, only 2 disks reproduced the patients' symptoms during diskography. The authors concluded that the high-intensity zone was a useful indicator of outer annular disruption, which may be a primary source of LBP.

Recently, Ricketson et al[48] have disputed this conclusion. They reported no correlation between the presence of the high-intensity zone on MRI and symptom reproduction with diskography (N = 29 patients with LBP with or without radiculopathy). Interestingly, no high-intensity zones were found in morphologically normal disks (as determined by diskography). The authors hypothesized that the high-intensity zone was not representative of nuclear or annular material but was actually granulation tissue. No histochemical analysis was reported. Ricketson et al concluded that the high-intensity zone was a nonspecific sign for LBP and that further research should investigate the high-intensity zone using gadolinium enhancement. Thus, the role of the high-intensity zone in pain production remains inconclusive.

Studies utilizing diskography have several limitations. Because diskography is invasive, it is often confined to use in individuals with severe disk disease and it is difficult to justify its use in other populations. This "referral bias" greatly limits the external validity of many studies. The actual validity of diskography as a gold standard is questionable, with some authors[31] reporting very high false-positive rates.

Because of methodological limitations (ie, a questionable gold standard and a limited study population), the diagnostic accuracy of MRI relating to most impairments to LBP remains unclear. Considering the limitations of diskography and surgical findings as gold standards, Deyo and colleagues[31,32] and Thornbury et al[41] have proposed that a superior method would be to have multiple expert clinicians review a patient's clinical course while considering the evolution of clinical findings and treatment. This method would allow a broader based population to be studied and would reduce examiner and diagnostic biases.

Therapeutic Impact

Therapeutic impact relates to the effect of diagnostic test results on patient care.[40] This construct encompasses not only the impact of early diagnosis but also issues of reduced cost and patient morbidity. For example, strong associations between MRI signal abnormalities and symptoms may allow clinicians to institute treatment and avoid further workup.[23] Tests with high false-positive rates are especially problematic because they may lead to aggressive and inappropriate treatments. Well-established negative tests, however, can benefit patients by providing reassurance that they do not have a serious illness.[40]

Studies that examine therapeutic impact use patient outcome as the primary dependent measure. Designs may include case-control, cohort, or randomized clinical trails. The use of diagnostic procedures in this category often relates to screening (eg, mammography). Because of cost restraints, however, lumbar MRI is not frequently used for screening.

Patient Outcome

The highest level of test usefulness is the effect of test results on patient outcome.[40] This construct may be viewed from a global perspective such as a quality-of-life improvement. Deyo et al[32] point out that although outcome usefulness should be closely linked with diagnostic accuracy and therapeutic impact, this may not always occur. For example, the choice of intervention may not be necessarily linked to the severity, and stage of the disease.

Study designs to determine the degree to which lumbar MRI has affected outcome would include nonrandomized concurrent comparisons and randomized clinical trials.[40, 47] Because most imaging studies have been limited to patients with delayed recovery of LBP or with radiculopathy who are undergoing advanced workup, surprisingly little work has been published describing outcome efficacy of lumbar MRI.

Gill and Blumenthal[21] illustrated a way that lumbar MRI may improve outcomes. These authors reported that of patients with a normal MRI who underwent lumbar interbody fusion, only 50% (7 of 14) reported improvement in symptoms postoperatively, whereas 74% (29 of 39) of those with an abnormal MRI reported improvement. Thus, a negative MRI may be a predictor of poor outcome following surgical intervention.

Ackerman et al[14] reported outcomes from a multicenter study of patients with persistent LBP (N = 1,084, median time since onset of symptoms = 6 months) who had undergone extensive diagnostic workup tot a suspected herniated nucleus pulposus based on clinical examination. The presence of a sequestered or free fragment of disk was present on imaging in 21% of the patients during the initial MRI. Patients where followed for 2 years and were asked to list the number of days they were unable to perform their usual work-related activities. Those patients who underwent surgery, for this condition reported only one fifth the number of disability days at 2 years compared with those patients treated nonoperatively. Thus, the finding of a free fragment may be a predictor of poor outcome following nonoperative treatment.

Discussion

As with any evolving technology, lumbar MRI for patients with LBP continues to be developed and refined. The technical capacity of MRI to image tissue is quite good and continues to improve due to advances in both MRI hardware and software. The ability to accurately reflect disk morphology, nerve compression, and bone marrow makes MRI an excellent tool to study the natural history, of degenerative disk disease. The use of MRI to evaluate muscle morphology following exercise or injury also has shown promising results.[51]

Unfortunately, the establishment of the diagnostic accuracy of MRI findings in patients with LBP remains problematic. Because of its high sensitivity to morphology, and biochemical composition, MRI illustrates anatomy in greater detail in vivo than do previously used diagnostic tools. Examiners, therefore, must question whether a visible variation from normal anatomy indicates what is causing symptoms, what is a risk factor for future symptoms, or what is simply a benign variation.[52] Currently, one must rely on clusters of findings such as the results of a physical examination or neurodiagnostic testing to improve judgments based on lumbar MRI.

The high prevalence oranatomic variation in the IVDs in individuals without low back symptoms suggests several questions. For example, why does only 1 of 2 patients with identical disk abnormalities have symptoms? Various hypotheses have been proposed. Ricketson et al[48] have suggested that a slight alteration in the fluid concentration within the annulus, such as neovascularity associated with the presence of granulation tissue, may trigger symptoms. Local venous distension within the outer annulus may stimulate nocioceptors, resulting in LBP. As the detection sensitivity of MRI improves, this question may be answered.

Although disk bulges and small disk displacements are common, extrusions are uncommon in individuals without low back symptoms.[19,36] Large disk herniations tend to have greater reduction in size over time than do small herniations.[17] Considering the natural history as well as surgical outcomes, it would make sense that MRI findings of large herniations are quite likely to indicate what generates symptoms, whereas MRI is not useful in identifying what generates symptoms with small herniations.

The MRI signal arising from a herniation may be an important predictor of outcome. For example, disk herniations giving off a bright [T.sub.2] signal have a high fluid content, which may have a high potential for resorption when related to the vascular and lymphatic tissues of the spinal canal. Herniations giving off a low [T.sub.2] signal have a low water content and may have a lower potential for resorption. This hypothesis is supported by Ackerman et al,[14] who observed that free fragments of disk are associated with poor outcome with nonoperative treatment.

The MRI finding of nerve displacement by disk material as a source of LBP and radiculopathy logically appears to have a direct relationship to the degree of distortion of the nerve.[19,41] A reliable classification scale will greatly assist in clarifying this relationship. Some studies[5,6,53,54] suggest that a chemical interface between the dorsal root ganglia and extravasated disk material is a source of symptoms in radiculopathy. As biologic markers of this event are identified, examiners may be able to differentiate between benign nerve compression and nerve compression that is clinically relevant. This discriminative ability could have an important effect as a "surgical localizer" for patients with multilevel disease.

Several barriers exist to research in lumbar MRI. Summarizing the literature is difficult because of a lack of operational definitions, as well as technological variations between researchers (eg, different magnet and coil capabilities).[55] Because of the high prior probability of abnormalities in the patients frequently studied, disease prevalence may be inflated. "Fast scan" technology, which reduces the time and expense required to perform MRI, may allow researchers to examine populations with a large variety of symptoms and help to clarify the natural history of disk disease.[56] The use of panels of experts and operational definitions will allow the establishment of a gold standard to improve diagnostic accuracy. Improved utilization of statistical analysis such as ROC curves will allow researchers to determine cutoff points between relevant and nonrelevant findings.[40]

In this issue, Riddle presents various classifications and discusses the role of pathology in patient classification. Knowledge of the anatomic impairment resulting in symptoms is highly desirable to provide a mechanistic basis for developing interventions. Unfortunately, this knowledge is often lacking for patients with LBP.

The relationship of pain to the presence of an anatomic impairment detectable on imaging is indicated in Table 7. For the first combination (ie, no pain with no physical impairment), the patient typically would be classified as having no low back symptoms. The second combination (ie, pain, but no observable anatomic impairment on MRI) is commonly encountered in physical therapy. In this situation, one must consider the possibility of nonspondylogenic sources (eg, vascular origin) or the potential for pain to be referred from another anatomic source that is not in the imaging field of view. Assuming that the source of the symptoms is of musculoskeletal origin, patients with this clinical profile often lead practitioners to theorize the existence of less tangible diagnoses such as facet impingement, sacroiliac joint dysfunction, and pain from a myofascial origin.[57-59]

Table 7.
Four Combinations of Pain and Anatomic Impairment

Clinical Finding              Consideration

No pain, no anatomic          Normal examination
  impairment
Pain, but no anatomic         Nonspondylogenic?
  impairment
                              Referred from a distant location?
                              Local impairment, but not
                                detectable on magnetic
                                resonance imaging
Anatomic impairment, but      Naturally occurring variation?
                                no pain
                              Risk factor for future onset of
                                symptoms?
                              Is the patient self-modulating
                                symptoms?
Anatomic impairment, pain     Is there a causal relationship
                                between anatomic impairment
                                and the patient's symptoms?
                              Is the anatomic impairment a
                                naturally occurring variation
                                that is not responsible for the
                                patient's symptoms?

The third combination includes those people with obvious lumbar anatomic impairments but no pain. The fundamental question is: Why are there no symptoms? Are the findings naturally occurring variations, and, if so, are they benign or is the patient at elevated risk for subsequent onset of symptoms? Answering this question obviously will have a great impact diagnostically, as well as psychologically.

The last combination, the presence of pain along with the existence of an anatomic impairment, may illustrate a causal relationship, but it can also be misleading. For example, the finding of an anatomic variation may not be relevant to the patient's symptoms. The results of the clinical examination, therefore, must be in agreement with the imaging findings. Care must be taken to detect subtle findings that may be overshadowed by more noticeable findings. For example, consider a case study[60] in which a patient was initially diagnosed as having diskogenic LBP based on the MRI finding of a bulging IVD at the L5-S1 level. Physical examination findings were inconsistent with the MRI finding. On closer evaluation, a zone of high [T.sub.2] signal was noted in the pars interarticularis at [L.sub.4], which was suggestive of an acute stress fracture. The presence of an acute stress fracture was later confirmed with radionuclear bone scanning.

The often discordant relationship between MRI findings and symptoms leads to the question, "When a patient has LBP and anatomic impairments visible on images, to what degree can this combination of clinical findings identify the cause or uniquely classify the individual?" Unfortunately, this information does not exist for most persons with LBP, and it marginally exists for patients with radiculopathy. The use of presumptive diagnoses obtained from imaging to classify patients is problematic because it can drive treatments that are ineffective, costly, and, in some instances, physically or psychologically harmful.[1] For example, a patient with no observable pathology may be wrongly labeled as seeking "secondary gain." Conversely, disability assessments may be spuriously inflated for a patient with an anatomic abnormality, such as a disk herniation, even though it may be a benign variation.

Patients with physical impairment may be wrongfully counseled regarding the severity of their condition. This situation could lead to inappropriate activity and treatment modification and could iatrogenically cause abnormal illness behavior.[1]

The goal of identifying various sources of LBP and creating examination protocols that allow meaningful classifications based on causal relationships is potentially achievable in the future. Tremendous advances are being made in biometrics, such as dynamic MRI, by which a structure may be exercised as the patient actively moves within the scanner. New paradigms are focusing on the concept of multiple tissues interacting to create symptoms. Chemical markers of certain diseases may be detectable on MRI. For example, the presence of fibrinolyric disorders may predispose some patients to delayed recovery by causing fibrin deposits around a nerve root.[61] Other issues such as self-modulation of pain interacting with various psychological factors must be considered.[1,6]

Important future work will develop meaningful classification systems to cluster patients with similar clinical findings and responses to treatment. From this work, mechanistic questions can be explored to determine whether specific pathoanatomic findings are a cause of symptoms, a comorbid condition, or a benign variation.

Summary

Lumbar MRI has a good technical capacity to provide high-resolution images that contrast soft tissues. It is an excellent tool for detecting the presence of anatomic impairment. In the absence of serious disease, reasonable evidence exists to support an association between disk extrusion, free fragments of disk, and major nerve compression with LBP. Disk degeneration, bulging, and small herniations are common in both individuals with and individuals without low back symptoms.

The clinical manifestations of LBP result from a complex interaction of biologic, psychologic, sociologic, and environmental factors that must all be considered when classifying the patient. Although certain MRI findings are critical to determining diagnostic classification and intervention, inappropriate interpretation of abnormalities visible on MRI may lead to erroneous patient classification and interventions that are inappropriate. As with any diagnostic test, lumbar MRI must be related to clinical findings to be meaningful. In the absence of corroborating physical examination findings, MRI findings are not adequate to classify patients for treatment.

Acknowledgment

We thank Paul W Stratford, PT, for his help with this work.

References

[1] Feuerstein M, Beattie PF. Biobehavioral factors affecting pain and disability in low back pain: mechanisms and assessment. Phys Ther. 1995;75:267-280.

[2] Deyo RA, Rainville J, Kent DL. What call the history and physical examination tell us about low back pain? JAMA. 1992;268:760-765.

[3] Haldeman S. North American Spine Society: Failure of the pathology model to predict back pain [presidential address]. Spine. 1990;15: 718-724.

[4] McNab I, McCulloch J. Backache. 2nd ed. Baltimore, Md: Williams & Wilkins; 1990.

[5] Cavanaugh JM. Neural mechanisms of lumbar pain. Spine. 1995;20: 1804-1809.

[6] Siddall PJ, Cousins MJ. Spine update: spinal pain mechanisms. Spine. 1997;22:98-104.

[7] Ackerman SJ, Steinberg EP, Bryan RN, et al. Trends in diagnostic imaging for low back pain: Has MR imaging been a substitute or add-on? Radiology. 1997;203:533-538.

[8] Enzmann DR, DeLaPaz RL, Rubin JB. Magnetic Resonance Imaging of the Spine. St Louis, Mo: CV Mosby Co; 1990:108-110.

[9] Stoller DW, Genant HK, Lang P. The spine. In: Stoller DW, ed. Magnetic Resonance imaging in Orthopaedics and Rheumatology. Philadelphia, Pa: JB Lippincott Co; 1989:322-375.

[10] Modic MT. Degenerative disorders of the spine. In: Modic MT, Masaryk TJ, Ross JS, eds. Magnetic Resonance Imaging of the Spine. Chicago, Ill: Yearbook Publishers; 1989:75-119.

[11] Berquist TH, Morin RL. General technical considerations in musculoskeletal MRI. In: Berquist TH, ed. MRI of the Musculoskeletal ,System. New York, NY: Raven Press; 1990:53-73.

[12] Berquist TH. Musculoskeletal neoplasm. In: Berquist TH, ed. MRI of the Musculoskeletal System. New York, NY: Raven Press; 1990:435--474.

[13] Berquist TH. Musculoskeletal infection. In: Berquist TH, ed. MRI of the Musculoskeletal System. New York, NY: Raven Press; 1990:475-489.

[14] Ackerman SJ, Steinberg EP, Bryan RN, et al. Persistent low back pain in patients suspected of having herniated nucleus pulposus: radiologic predictors of functional outcome--implications for treatment selection. Radiology. 1997;203:815-822.

[15] Aprill C, Bogduk N. High intensity zone: a diagnostic sign of painful lumbar disc on magnetic resonance imaging. Br J Radiol. 1992;65: 361-368.

[16] Beattie PF, Brooks WM, Rothstein JM, et al. Effect of lordosis on the position of the nucleus pulposis in supine subjects: a study using magnetic resonance imaging. Spine. 1994;19:2096-2102.

[17] Bozzao A, Gallucci M, Masciocchi C, et al. Lumbar disk herniation: MR imaging assessment of natural history in patients treated without surgery. Radiology. 1992;185:135-141.

[18] Buirski G, Silberstein M. The symptomatic lumbar disc in patients with low-back pain: magnetic resonance imaging appearances in both a symptomatic and control population. Spine. 1993;18:1808-1811.

[19] Boos N, Rieder R, Schade V, et al. The diagnostic accuracy of magnetic resonance imaging, work perception, and psychosocial factors in identifying symptomatic disc herniations. Spine. 1995;20:2613-2625.

[20] Fraser RD, Sandhu A, Gogan WJ. Magnetic resonance imaging findings 10 years after treatment for lumbar disc herniation. Spine. 1995;20:710-714.

[21] Gill K. Blumenthal SL. Functional results after anterior lumbar fusion at L5-S1 in patients with normal and abnormal MRI scans. Spine. 1992; 17:940-942.

[22] Hashimoto K, Akahori O, Kitano K, et al. Magnetic resonance imaging of lumbar disc herniation: comparison with myelography. Spine. 1990;15:1166-1169.

[23] Horton WC, Daftari TK. Which disc as visualized by magnetic resonance is actually the source of pain? A correlation between magnetic resonance imaging and discography. Spine. 1992;17(suppl 6):S164-S171.

[24] Kim KY, Kim YT, Lee CS, et al. Magnetic resonance imaging in the evaluation of the lumbar herniated intervertebral disc. Int Orthop. 1993; 17:241-244.

[25] Komori H, Shinomiya K, Nakai O, et al. The natural history of herniated nucleus pulposus with radiculopathy. Spine. 1996;21:225-229.

[26] Horowitiz AL. MRI Physics for Physicians. New York, NY: Springer-Verlag New York Inc; 1989.

[27] Modic MT, Ross JS. Magnetic resonance imaging in the evaluation of low back pain. Orthop Clin North Am. 1991;22:283-301.

[28] Tullberg T. Grane P, Isacson J. Gadolinium-enhanced magnetic resonance imaging of 36 patients one year after lumbar disc resection. Spine. 1994;19:176-182.

[29] Hickey DS, Aspden RM, Hukins DWL, et al. Analysis of magnetic resonance images from normal and degenerate lumbar intervertebral discs. Spine. 1986;11:702-708.

[30] Herzog RJ, Guyer RD, Graham-Smith A, Simmons ED Jr. Magnetic resonance imaging: use in patients with low back or radicular pain. Spine. 1995;20:1834-1838.

[31] Deyo R. Understanding the accuracy of diagnostic tests. In: Weinstein JN, Rydevik BL, Sonntag VKH, eds. Essentials of the Spine. New York, NY: Raven Press; 1995:55-69.

[32] Deyo R, Haselkorn J, Hoffman R, Kent DL. Designing studies of diagnostic tests for low back pain or radiculopathy. Spine. 1994; 19(suppl 18):2057S-2065S.

[33] Bigos S, Bowyer O, Braen G, et al. Acute Low Back Problems in Adults. Rockville, Md: US Dept of Health and Human Services, Agency for Health Care Policy and Research; December 1994. Clinical Practice Guidelines, Quick Reference Guide, No. 14. AHCPR Publication No. 95-0643.

[34] Beattie PF. The relationship between symptoms and abnormal magnetic resonance images of lumbar intervertebral disks. Phys Ther. 1996;76:601-608.

[35] Boden SD, Davis DO, Dina TS, et al. Abnormal magnetic-resonance scans of the lumbar spine in asymptomatic subjects: a prospective investigation. J Bone Joint Surg Am. 1990;72:403-408.

[36] Jensen MC, Brant-Zawadzki MN, Obuchowski N, et al. Magnetic resonance imaging of the lumbar spine in people without back pain. N Engl J Med. 1994;331:69-73.

[37] Paajanen H, Erkintalo M, Kuusela T, et al. Magnetic resonance study of disc degeneration in young low-back pain patients. Spine. 1989; 14: 982-985.

[38] Powell MC, Wilson M, Szypryt P, et al. Prevalence of lumbar disc degeneration observed by magnetic resonance in symptomless women. Lancet. 1986;2:1366-1367.

[39] Weinreb JC, Wolbarsht LB, Cohen JM, et al. Prevalence of lumbo-sacral intervertebral disk abnormalities on MR images in pregnant and asymptomatic nonpregnant women. Radiology. 1989;170:125-128.

[40] Kent DL, Larson EB. Disease, level of impact, and quality of research methods: three dimensions of clinical efficacy assessment applied to magnetic resonance imaging. Invest Radiol. 1992;27:245-254.

[41] Thornbury, JR, Fryback DG, Turski PA, et al. Disk-caused nerve compression in patients with acute low back pain: diagnosis with MR, CT myelography, and plain CT. Radiology. 1993;186:731-738.

[42] Osti OL, Fraser RD. MRI and discography of annular tears and intervertebral disc degeneration: a prospective clinical comparison. J Bone Joint Surg Br. 1992;74:431-435.

[43] Schellhas KP, Pollei S, Gundry CR, Heithoff KB. Lumbar disc high-intensity zone: correlation of magnetic resonance imaging and discography. Spine. 1996;21:79-86.

[44] Raininko R, Manninen H, Battie MC, et al. Observer variability in the assessment of disc degeneration on magnetic resonance images of the lumbar and thoracic spine. Spine. 1995;20:1029-1035.

[45] Brant-Zawadzki MN, Jensen MC, Obuchowski N, et al. Interobserver and intraobserver variability in interpretation of lumbar disc abnormalities: a comparison of two nomenclatures. Spine. 1995;20: 1257-1263.

[46] Sward L, Hellstrom M, Jacobsson B, et al. Disc degeneration and associated abnormalities of the spine in elite gymnasts: a magnetic resonance imaging study. Spine. 1991;16:437-443.

[47] Elwood JM. The importance of causal relationships in medicine. In: Elwood JM, ed. Causal Relationships in Medicine: A Practical System for Critical Appraisal New York, NY: Oxford University Press Inc; 1988:3-9.

[48] Ricketson R, Simmons JW, Hauser BO. The prolapsed intervertebral disc: the high-intensity zone with discography correlation. Spine. 1996;21:2758-2762.

[49] Bogduk N. The innervation of the lumbar spine. Spine. 1983;8: 286-293.

[50] Humzah MD, Soames RW. The human intervertebral disc: structure and function. Anat Rec. 1988;220:337-356.

[51] Fleckenstein JL, Canby RC, Parkey RW, Peshock RM. Acute effects of exercise on MR imaging of skeletal muscle in normal volunteers. AJR Am J Roentgenol. 1988;151:231-237.

[52] Videman T, Battie MC, Gill KG, et al. Magnetic resonance imaging findings and their relationships in the thoracic and lumbar spine: insights into etiopathogenesis of spinal degeneration. Spine. 1995;20: 928-935.

[53] Howe JF, Loeser JD, Calvin WH. Mechanosensitivity of dorsal root ganglia and chronically injured axons: a physiological basis for the radicular pain of nerve root compression. Pain. 1977;3:25-41.

[54] Bogduk N, Twomey LT. Clinical Anatomy of the Lumbar Spine. New York, NY: Churchill Livingstone Inc; 1987:11-24, 130-138.

[55] Reicher MA, Gold RH, Halbach VV, et al. MR imaging of the lumbar spine: anatomic correlations and the effects of technical variations. AJR Am J Roentgenol. 1986;147:891-898.

[56] Jarvik J, Maravilla KR, Haynor DR, et al. Rapid MR imaging versus plain radiography in patients with low back pain: initial results of a randomized study. Radiology. 1997;204:447-454.

[57] Jackson RP. The facet syndrome: Myth or reality? Clin Orthop. 1992;279:110-121.

[58] Schwarzer AC, Aprill CN, Bogduk N. The sacroiliac joint in chronic low back pain. Spine. 1995;20:31-37.

[59] Schwarzer AC, Aprill CN, Derby R, et al. The relative contributions of the disc and zygapophyseal joint in chronic low back pain. Spine. 1994;19:801-806.

[60] Beattie PF. MRI findings and low back pain. PT--Magazine of Physical Therapy. 1997;5 (6) :68-76.

[61] Nachemson A. The future of low back pain. In: The Lumbar Spine. Vol II. 2nd ed. Philadelphia, Pa: WB Saunders Co; 1996:28-42.

PF Beattie, PhD, PT, OCS, is Physical Therapist, University Sports Medicine, Department of Orthopaedics, School of Medicine and Dentistry, University of Rochester, 2180 S Clinton Ave, Rochester, NY 14618 (USA). Address all correspondence to Dr Beattie.

SP Meyers, MD, PhD, is Associate Professor of Radiology, School of Medicine and Dentistry, University of Rochester.