Patients with lumbar spinal stenosis usually undergo a staged diagnosis (see Figure 8). The first diagnostic stage is the physician visit, during which the patient receives a physical examination. Results of the physical are combined with information from the patient history in a preliminary diagnosis. Lumbar spinal stenosis is not definitively diagnosed at this stage, so the diagnostic results are described as consistent with spinal stenosis or not consistent with spinal stenosis.
Although clinicians consider a combination of results of the history and physical examination and imaging findings to be the most effective means of diagnosing lumbar spinal stenosis, no objective criteria for using the history and physical examination have been reported. In addition, there are no reported clinical trials of the effectiveness of such a composite diagnosis. The only quantitative evidence correlating diagnostic information with outcomes is for the imaging findings. This absence of evidence limits the analysis that can be performed for some of the research questions.
Katz et al. (1995) examined the value of the history and physical examination in the diagnosis of degenerative lumbar spinal stenosis. In this study, 93 patients over 40 years of age with symptoms of low back pain were examined by attending physicians who were then asked the extent to which they were certain the patient had lumbar spinal stenosis. The diagnostic impressions of expert clinicians and imaging, when available, were used as a reference standard to evaluate the attending physicians diagnosis. Severe lower extremity pain, absence of pain when seated, a wide-based gait, thigh pain following 30 seconds of lumbar extension, and neuromuscular deficits were all strongly associated with patients with lumbar spinal stenosis. No pain when seated and wide-based gait had the highest specificity, 93 percent and 97 percent, respectively. The highest sensitivity came from age greater than 65 (77 percent), pain below buttocks (88 percent), and no pain with flexion (79 percent) (Katz, Dalgas, Stucki et al., 1995).
Fritz et al. (1997) have developed a treadmill test as a clinical diagnostic tool for the differentiation of neurogenic claudication due to lumbar spinal stenosis from other pathologies that may produce similar symptoms. Spinal extension and weight bearing that occur during walking narrow the spinal canal and exacerbate the symptoms of lumbar spinal stenosis. Spinal flexion or nonweight-bearing postures that occur while sitting increase the dimensions of the spinal canal and reduce symptoms. The treadmill test involves having the patient walk on a level surface and an inclined surface. The time until onset of symptoms, total walking time, and time until symptoms return to baseline are recorded for each surface. Walking on an inclined plane produces spinal flexion and may be better tolerated by patients with lumbar spinal stenosis. The treadmill test was evaluated using 45 subjects with low back pain of varying etiologies and self-reported limitations in walking. Diagnostic images with MRI or CT were used as the gold standard for diagnosis. Twenty-six of the subjects were diagnosed by imaging as being stenotic. Self-reported sitting to relieve symptoms was significantly related to diagnosis. The sensitivity of this self-reported measure was 88.5 percent (95 percent confidence interval [CI] of 76.2 to 100), but specificity was 38.9 percent (95 percent CI of 16.4 to 61.4). For the treadmill test, earlier onset of symptoms with level walking, greater total walking time during inclined walking, and prolonged recovery after level walking were significantly related to a diagnosis of lumbar spinal stenosis. The sensitivity and specificity for earlier onset of symptoms with level walking were 68.0 percent (95 percent CI of 49.7 to 86.3) and 83.3 percent (95 percent CI of 66.1 to 100), respectively; for larger total walking time during inclined walking, they were 50.0 percent (95 percent CI of 37.5 to 62.5) and 92.3 percent (95 percent CI of 77.8 to 100), respectively; and for prolonged recovery after level walking, they were 81.8 percent (95 percent CI of 5.7 to 97.9) and 68.4 percent (95 percent CI of 47.5 to 89.3), respectively. The authors concluded that a two-stage treadmill test might be more useful in the differential diagnosis of lumbar spinal stenosis compared to patients self-reports of posture (Fritz, Erhard, Delitto et al., 1997).
Use of the treadmill-bicycle test for the differential diagnosis of neurogenic claudication was also examined by Tenhula et al. (2000). In this study, 32 patients with documented lumbar spinal stenosis were evaluated before and after surgery. Patients were found to have a significant increase in their symptoms from the start to the end of the treadmill test but fewer patients were found to have significant symptoms on bicycle testing. Two years after surgery, patients had an improvement in their walking ability on treadmill testing, but showed no improvement in their ability to bicycle. The authors believe the treadmill-bicycle test may be a useful tool for the differential diagnosis of neurogenic claudication (Tenhula, Lenke, Bridwell et al., 2000).
Imaging examinations appear to be used primarily in a second diagnostic stage. At this stage, some surgical intervention is usually under consideration, so the imaging examination is as much for surgical planning as it is for confirmation of the preliminary diagnosis. In chiropractic care, a plain x-ray image may initially be obtained to aid in chiropractic therapy even if surgery is not planned (DuPriest, 1993).
In clinical practice, the imaging results help to confirm the diagnosis of spinal stenosis after the history and physical indicate the likelihood of spinal stenosis. Clinical trials interchangeably use myelogram, CT, or MRI results for confirmation of the diagnosis. This is partly because spinal stenosis is defined in terms of the anatomy displayed in the images. Also, there is no other means of verifying the results, short of measurement of the spinal canal during surgery. As discussed in the next section, plain film radiography is not considered a definitive standard for use in diagnosing lumbar spinal stenosis (Widelec, Bacq, and Peetrons, 1999). Because no independent means is available to confirm that imaging results are right or wrong, assessment of the performance of any of the imaging modalities for diagnosis of lumbar spinal stenosis is not possible in the same way we usually assess diagnostics. This is particularly true for negative cases, in which there would not be subsequent surgery on the spine.
The first imaging modality used to diagnose and evaluate lumbar spinal stenosis was radiography: film-based x-ray imaging, colloquially known as plain film. The typical lumbar spine examination consists of AP (anteroposterior: front to back), lateral (side to side; see Figure 4), and oblique (diagonal; see Figure 5) views (Widelec, Bacq, and Peetrons, 1999). Film radiography has excellent spatial resolution for displaying small anatomic details but has several characteristics that limit its value in diagnosing spinal stenosis. First, it is a projection method: small anatomic details may be obscured by overlapping structures. Obtaining multiple views can sometimes resolve those structures. Second, the soft-tissue contrast of radiography is relatively low. Bones are depicted very clearly on plain radiographs, so they are frequently used to rule out vertebral fracture, if it is suspected.
Plain radiographs depict the spinal canal well enough for measurement of its diameter. The lateral radiograph is also useful for diagnosing spondylolisthesis, forward/backward displacement of a vertebra (Sackett, 1994; Wood, Popp, Transfeldt et al., 1994). For measuring displacement, radiography is considered the gold standard.
Intradural contrast-enhanced radiography of the spine is known as myelography. To obtain a myelogram, a radiopaque contrast agent is injected into the spinal canal, and x-ray images are taken. The contrast agent diffuses through the spinal canal, outlining it quite clearly on the myelogram, but if stenosis causes a complete blockage of the spinal canal, no contrast agent will flow below (inferior to) the stenosis. Modern contrast agents for myelography are water soluble and contain iodine. Some patients are sensitive to iodine and may have an allergic reaction to the contrast material. Myelography is contraindicated in these patients. Adverse side effects are common in myelography. About half of patients experience head or neck pain, and 15 percent experience nausea or dizziness. Nearly all patients find myelography uncomfortable and will not be able to resume some normal activities, like driving, until the day after the procedure (American College of Radiology, 1999; Mitchell, 2000; Ramsbacher, Schilling, Wolf et al., 1997).
Spinal stenosis is traditionally defined by its appearance on a myelogram. Complete blockage is as described above. Partial blockage is a stenosis where the film shows a complete interruption of the column of contrast agent at the point of focal narrowing (i.e., the stenosis), but enough contrast agent gets through the stenosis to opacify the spinal canal below (inferior to) the stenosis. Stenoses of lesser degree are defined by the diameter of the contrast agent column at the stenosis. This is usually measured in the AP dimension, as viewed on the lateral film. The threshold measurement used to define stenosis varies from investigator to investigator (see Table 17 and Question 7 in Chapter 3).
The myelogram has long been considered a gold standard for imaging spinal stenosis and helping to confirm the diagnosis, but it has been supplanted by the cross-sectional modalities CT and MRI. Recent clinical handbooks now recommend MRI rather than plain films or myelography as the primary imaging modality in cases of suspected spinal stenosis. Myelography is now rarely done in routine clinical practice (Eisenberg and Margulis, 2000b; Grossman, Katz, Santelli et al., 1994; Gundry and Heithoff, 1999; Mitchell, 2000).
Because it is a three-dimensional modality, CT avoids the overlapping-structure problem of film-based radiography. But CT is still an x-ray modality, and bones will appear more distinct than soft tissue on a CT scan. CT is limited to axial images (cross sections as viewed from the patients head to toe; see Figure 3), but all currently available CT scanners have tilting gantries that allow the imaging plane to be tilted 20 to 30 in each direction (ECRI, 1999). For many patients, this permits acquisition of images parallel to the disks, so the entire disk is in one image, although the vertebral column is not straight up and down.
CT scans can be acquired following administration of contrast agent, just like radiographs. Because the three-dimensional images resolve overlapping structures, there is less need for contrast agent in CT scans of the spine. Also, smaller quantities of the contrast agent are necessary with CT, but CT can be done following a myelogram with its larger contrast agent dose.
While it depicts anatomy in cross-section as CT does, MRI is based on completely different physical principles. A full explanation of how MRI works is beyond the scope of this report. Briefly, MRI derives contrast primarily from differences in the T1 and T2 relaxation times of hydrogen nuclei in the body. Those differences stem from differences in the chemical environment of water in different types of cells. Parameters for magnetic resonance (MR) image acquisition (pulse sequences) can be varied to emphasize T1 or T2 or to base contrast on some combination of the two. Usually, MR examinations comprise several sets of images using several different pulse sequences. As with all other MRI applications, some pulse sequences are more effective than others for imaging the spine (Ramsbacher, Schilling, Wolf et al., 1997). Because technical characteristics of different MR scanners are different, the pulse sequences that are ideal for one scanner may not be ideal for another scanner.
Besides the advantage of soft-tissue contrast, MRI has the advantage of being able to acquire images in any user-selected plane. Sagittal sections (cross sections as viewed from the patients left to right; see Figure 4) are particularly useful in spine imaging, and it is quite easy to select axial image planes parallel to the disks. Contrast-enhancement agents are also available for MRI, but instead of containing iodine or other atoms of high atomic number like x-ray contrast agents, MR contrast agents contain paramagnetic ions like gadolinium. Because MR images inherently provide good contrast of the spine and its contents, contrast agents are not usually necessary in MRI of the spine. Studies described as MR myelography do not necessarily involve contrast agents (Ramsbacher, Schilling, Wolf et al., 1997).
The chief disadvantage of MRI is that its spatial resolution is not as good as that of film radiographs (Ramsbacher, Schilling, Wolf et al., 1997). MRI can also be susceptible to spatial distortions that would cause errors in quantitative measurements of the vertebrae, but adherence to routine quality assurance procedures at the imaging center will minimize distortions.
Ultrasound can also generate cross-sectional images, but it is not well suited to imaging of the spine and contents. Ultrasound waves do not penetrate bone, so imaging can only be done through windows like the ligamentum flavum (Pai, 1993). Adequate windows cannot be found in some patients, so their spinal canal cannot be visualized with ultrasound. One of the two clinical trials of spinal ultrasound reported that this problem is worse in individuals with spinal stenosis (Pai, 1993). In one-third of the measurements reported in this trial, ultrasound results differed from radiographic results (considered the reference standard) by 2 mm or more. The other published ultrasound trial did not report quantitative measurements (Tait, Charlesworth, and Lemon, 1985). Ultrasound is inexpensive and safe, but it lacks accuracy in quantitative measurement of the spinal canal.
Recent textbooks and clinical handbooks of radiology identify an important shift in clinical practice for diagnosis of spinal stenosis, herniation of the disks, and other conditions of the lumbar spine. The cross-sectional modalities, particularly MRI, are now considered the primary imaging tools for diagnosis of spinal stenosis (Eisenberg and Margulis, 2000a; Eisenberg and Margulis, 2000b; Grossman, Katz, Santelli et al., 1994; Gundry and Heithoff, 1999; Kirkaldy-Willis and Bernard Jr, 1999; Mitchell, 2000; Spengler, 2000; Widelec, Bacq, and Peetrons, 1999; Wilmink, 2000). Reasons cited for the shift include multiplanar imaging capabilities, better ability to show bulging or fragmentation of the disks, and the risks and discomfort of myelography. These books should not be considered evidence based. They represent conventional wisdom among radiologists, which is shaped in part by clinical trials and other evidence. In fact, Wilmink notes the absence of conclusive evidence that MRI is the most effective modality for diagnosing spinal stenosis (Wilmink, 2000).
Official
guidelines developed by professional specialty societies like the
Plain films still have a role in diagnosis of spinal disorders; they may be all that is necessary to diagnose vertebral fracture and/or spondylolisthesis (Crosby and Brant-Zawadzki, 1994; Grossman, Katz, Santelli et al., 1994), and they are capable of measuring the diameter of the spinal canal (Sackett, 1994). Plain radiographs are considered appropriate by ACR in some clinical circumstances (6 or 7 on a 9 point scale) (American College of Radiology, 1998).
All the imaging modalities are adversely affected by surgical hardware and other metal objects implanted in the body. These objects are radiopaque; so they obscure overlapping structures in plain x-rays and myelograms. The effect is heightened in CT scanning, where the metal objects cause artifacts that can affect the entire image. Metal objects also affect MR scans, although most orthopedic devices are nonmagnetic and do not completely contraindicate MRI. Metal causes susceptibility artifacts: there is no signal from the object itself, and portions of the image near the object may be distorted or lose signal. The magnitude of this effect depends on the size and placement of the object and its composition (Fredrickson, 2000). Since diagnosis among patients who already have had surgery is outside the scope of this report, the studies we examined for this technology assessment did not deal with artifacts from metal implants.
All imaging modalities are subject to variation in image quality and diagnostic effectiveness and variation in interpretation by the radiologist. We located one study that addresses how variations in image quality affect the diagnosis of spinal stenosis. Jarvik et al. (2000) examined variations in the quality of lumbar spine MR images, and attempted to correlate them with characteristics like the magnetic field strength of the scanner and the ownership and siting of the scanner. Three readers rated the quality of 69 examinations from 17 centers. Readers were blinded to identifying information on the images, including information that could identify the center or the type of scanner used. The readers graded each examination on a four-point scale for the sharpness with which each of seven clinically important structures was depicted. They also assigned an overall quality score to each examination, using the same scale. The average overall quality rating for each center ranged from 1.96 to 3.56, while the average rating of the seven structures from each center ranged from 2.25 to 3.82 (Jarvik, Robertson, Wessbecher et al., 2000).
Jarvik et al. (2000) found that image quality was significantly decreased by the following characteristics: low magnetic field strength (less than 1.0 tesla), location of the scanner outside a hospital, and for-profit ownership of the scanner. The number of radiologists interpreting images at a center and the percentages of images reviewed by a physician before the patient was dismissed had smaller but still statistically significant effects on study quality. The authors note that they were unable to measure some characteristics that would be expected to affect study quality, such as training and experience of the radiologists. However, the results of this study are consistent with the idea that the quality of diagnostic imaging examinations, especially complicated ones like MRI, should not be taken for granted.
Finally, image quality should not be mistaken for diagnostic effectiveness, although poor image quality can impede effective diagnosis. While one modality may be able to depict structures with more detail and more contrast than another, the modality of lesser quality may still depict the anatomy clearly enough to permit a correct diagnosis. For this reason, we base our evaluation of diagnostic imaging modalities on their diagnostic results, not on subjective or objective measurements of image characteristics.
Imaging data are also necessary for planning surgical treatment. The surgeon needs to determine the extent of decompression, identify the bony anatomy, and measure the extent of the stenosis (Gunzburg and Szpalski, 1999). Accurate location of the stenosis is essential to avoiding operation at the wrong lumbar level. Identification of vertebral defects, anatomic landmarks, and other bone anatomy is necessary to minimize complications (Bernard and Yong-Hing, 1999). Selection of the correct type, size, and placement of orthopedic hardware is aided by preoperative imaging studies (Visarius, 2000). Both plain x-rays and CT or MRI are often needed in surgical planning (Lazennec, Ramare, Arafati et al., 1999); the specific combination of modalities varies from surgeon to surgeon and from patient to patient. There is no single preoperative imaging plan that is appropriate for all spinal stenosis patients.
Assessment of this indication for spinal imaging is complicated by the absence of any clinical trials measuring the effectiveness of the various imaging modalities in surgical planning.