The Goal of Imaging

The primary goal of imaging in the patient with spine or limb pain is to identify patients who are suffering from an undiagnosed systemic process causal of the pain/dysfunction syndrome. This is an uncommon phenomenon. An analysis by Jarvik and Deyo suggests that 95% of low back pain is due to benign processes. In patients presenting to a primary care setting with low back pain, only 0.7% suffer from undiagnosed metastatic neoplasm. Spine infection, including pyogenic and granulomatous diskitis, epidural abscess, or viral processes, is present in only 0.01% of subjects. Noninfectious inflammatory spondyloarthropathies, such as ankylosing spondylitis, account for 0.3% of presentations. Osteoporotic compression fractures are the most common systemic pathologic process to present as back pain, accounting for 4% of patients. Imaging seeks to identify the approximately 5% of patients with back or limb pain who have undiagnosed systemic disease as the etiology of their pain. A related imaging goal is to characterize and assist in therapy planning in the very small percentage of patients who have neural compressive disease resulting in radiculopathy or radicular pain syndromes that fail conservative therapy and require surgical or minimally invasive intervention.

Specificity: Asymptomatic Imaging Findings

The low prevalence of systemic disease as a cause of back pain implies that most imaging studies primarily describe what are often, and inappropriately, termed “degenerative” phenomena. These may include anterior and lateral vertebral body osteophytes, loss of T2 signal in the intervertebral disk, and structural changes of facet arthrosis. Degeneration is a pejorative term implying disease; these changes have no relationship to pain syndromes and correlate only with age. They are best referred to as age or age-related change. These age changes are typically relatively uniform across the spine, although the lowest lumbar segments are over-represented. Evidence for the lack of specificity of such imaging findings for spine pain syndromes is evident in cadaver studies, imaging studies in asymptomatic populations, and population studies.

Nathan described the presence of anterior and lateral osteophytes in 100% of cadavers at the age of 40, whereas posterior osteophytes are present only in a minority of cadavers at 80 years of age. Hult studied adults with spine radiographs and showed that by age 50 years, 87% will have radiographic evidence of disk age-related change (narrowing of the disk space, marginal sclerosis with osteophytes, vacuum phenomena). In a second study including a cohort of asymptomatic workers, Hult noted radiographic evidence of disk “disease” in 56% of those aged 40 to 44 years, which rose to 95% in subjects 50 to 59 years old. With the evolution of more sophisticated spine imaging techniques, this lack of specificity of degenerative findings has not improved. Hitselberger and Witten studied plain myelography of asymptomatic volunteers and noted that 24% showed abnormalities that would have been considered significant in a clinical context of back or leg pain. A study of lumbar spine CT in asymptomatic volunteers by Wiesel and colleagues showed that in patients older than 40 years, 50% had “significant” abnormalities. Similarly, Boden and colleagues evaluated MRI of the lumbar spine in asymptomatic volunteers; in patients older than 60 years, 57% had abnormalities that would have been considered significant in an appropriate clinical setting. Jarvik and colleagues studied a large patient population with MRI. This study noted that only extrusions, moderate to severe central canal stenosis, and direct visualization of neural compression were likely to be significant and would separate patients with pain from asymptomatic volunteers. Disk protrusions, zygapophysial joint (z-joint or facet joint) arthrosis, and anterolisthesis or retrolisthesis were virtually always asymptomatic findings. Imaging studies of asymptomatic volunteers are compiled in Table 15.1 .

A study by Kanayama in 200 healthy adults (mean age 40, with no current complaint or therapy for back pain nor any history of lumbar surgery) segregated lumbar MRI findings by segmental level ( Table 15.2 ). Asymptomatic T2 signal loss and disk herniations were most common at the L4 and L5 segmental levels; this series also had a high prevalence of asymptomatic high intensity zones (HZ) (24% at L4 and L5). More recent studies have addressed the prevalence of disk “degenerative” imaging findings (T2 signal loss, loss of disk space height) in younger populations, primarily in Scandinavian countries; these are MRI population-based studies without regard to symptomatology. Kjaer and colleagues, studying children age 13 years, found a 21% prevalence of disk “degeneration.” In a study of adolescents, Salminen and colleagues found a 31% prevalence of disk “degeneration” in 15-year-olds, which rose to 42% in 18-year-olds. Takatalo and colleagues evaluated 558 young adults aged 20 to 22 years. Using the 5-point Pfirrmann classification of disk degeneration, they noted disk degeneration of grade 3 or higher in 47% of these young adults. There was a higher prevalence in males (54%) than in females (42%). Multilevel degeneration was identified in 17%.

As in the lumbar region, age changes in the cervical and thoracic spine are common, asymptomatic, and increase in prevalence with age. Matsumoto studied nearly 500 asymptomatic patients with MRI; he noted a loss of T2 signal within cervical disks in 12% to 17% of patients in their twenties, but in 86% to 89% of patients older than 60 years of age. Asymptomatic cervical cord compression was observed in 7.6% of patients, largely over the age of 50. Similarly, Boden studied 63 asymptomatic subjects with MRI and noted cervical disk “degeneration” in 25% of those younger than 40, and in excess of 60% of patients older than 40 years of age. Asymptomatic subjects older than 40 years of age had a 5% rate of disk herniations and a 20% rate of foraminal stenosis. Teresi studied 100 asymptomatic subjects with MRI and noted asymptomatic cervical cord compression in 7% and either disk protrusion or annular bulge in 57% of subjects older than 64 years of age. In the thoracic spine, Wood studied 90 asymptomatic patients with MRI. In this population, 73% of the patients had positive thoracic imaging findings, 37% had disk herniations, 53% demonstrated disk bulges, and asymptomatic cord deformity was present in 29%.

The evidence is deep and overwhelming. The structural spine imaging findings most commonly referred to as “degenerative changes” or “degenerative disk disease,” including anterior and lateral osteophytes, loss of T2 signal in the disk, loss of disk space height, disk bulges and protrusions, and facet arthrosis, are ubiquitous and unassociated with pain syndromes; their only association is with age. They can only be avoided by a youthful death. They are not a disease state and are best referred to as normal age change or age-related change.

A consequence of this high prevalence of asymptomatic age-related changes is that the imager must know the nature of the pain syndrome if he or she is to properly focus on findings significant to the unique patient under consideration. There must be concordance of the imaging finding and the pain syndrome it is postulated to elicit. Imaging cannot prove causation, hence the need for anesthetic and provocative procedures. Communication regarding the nature of the pain syndrome is essential, whether this occurs through a robust electronic medical record, an intake document at the imaging site, or direct interaction of the imager with the patient.

Specificity: Asymptomatic Imaging Findings

The low prevalence of systemic disease as a cause of back pain implies that most imaging studies primarily describe what are often, and inappropriately, termed “degenerative” phenomena. These may include anterior and lateral vertebral body osteophytes, loss of T2 signal in the intervertebral disk, and structural changes of facet arthrosis. Degeneration is a pejorative term implying disease; these changes have no relationship to pain syndromes and correlate only with age. They are best referred to as age or age-related change. These age changes are typically relatively uniform across the spine, although the lowest lumbar segments are over-represented. Evidence for the lack of specificity of such imaging findings for spine pain syndromes is evident in cadaver studies, imaging studies in asymptomatic populations, and population studies.

Nathan described the presence of anterior and lateral osteophytes in 100% of cadavers at the age of 40, whereas posterior osteophytes are present only in a minority of cadavers at 80 years of age. Hult studied adults with spine radiographs and showed that by age 50 years, 87% will have radiographic evidence of disk age-related change (narrowing of the disk space, marginal sclerosis with osteophytes, vacuum phenomena). In a second study including a cohort of asymptomatic workers, Hult noted radiographic evidence of disk “disease” in 56% of those aged 40 to 44 years, which rose to 95% in subjects 50 to 59 years old. With the evolution of more sophisticated spine imaging techniques, this lack of specificity of degenerative findings has not improved. Hitselberger and Witten studied plain myelography of asymptomatic volunteers and noted that 24% showed abnormalities that would have been considered significant in a clinical context of back or leg pain. A study of lumbar spine CT in asymptomatic volunteers by Wiesel and colleagues showed that in patients older than 40 years, 50% had “significant” abnormalities. Similarly, Boden and colleagues evaluated MRI of the lumbar spine in asymptomatic volunteers; in patients older than 60 years, 57% had abnormalities that would have been considered significant in an appropriate clinical setting. Jarvik and colleagues studied a large patient population with MRI. This study noted that only extrusions, moderate to severe central canal stenosis, and direct visualization of neural compression were likely to be significant and would separate patients with pain from asymptomatic volunteers. Disk protrusions, zygapophysial joint (z-joint or facet joint) arthrosis, and anterolisthesis or retrolisthesis were virtually always asymptomatic findings. Imaging studies of asymptomatic volunteers are compiled in Table 15.1 .

A study by Kanayama in 200 healthy adults (mean age 40, with no current complaint or therapy for back pain nor any history of lumbar surgery) segregated lumbar MRI findings by segmental level ( Table 15.2 ). Asymptomatic T2 signal loss and disk herniations were most common at the L4 and L5 segmental levels; this series also had a high prevalence of asymptomatic high intensity zones (HZ) (24% at L4 and L5). More recent studies have addressed the prevalence of disk “degenerative” imaging findings (T2 signal loss, loss of disk space height) in younger populations, primarily in Scandinavian countries; these are MRI population-based studies without regard to symptomatology. Kjaer and colleagues, studying children age 13 years, found a 21% prevalence of disk “degeneration.” In a study of adolescents, Salminen and colleagues found a 31% prevalence of disk “degeneration” in 15-year-olds, which rose to 42% in 18-year-olds. Takatalo and colleagues evaluated 558 young adults aged 20 to 22 years. Using the 5-point Pfirrmann classification of disk degeneration, they noted disk degeneration of grade 3 or higher in 47% of these young adults. There was a higher prevalence in males (54%) than in females (42%). Multilevel degeneration was identified in 17%.

As in the lumbar region, age changes in the cervical and thoracic spine are common, asymptomatic, and increase in prevalence with age. Matsumoto studied nearly 500 asymptomatic patients with MRI; he noted a loss of T2 signal within cervical disks in 12% to 17% of patients in their twenties, but in 86% to 89% of patients older than 60 years of age. Asymptomatic cervical cord compression was observed in 7.6% of patients, largely over the age of 50. Similarly, Boden studied 63 asymptomatic subjects with MRI and noted cervical disk “degeneration” in 25% of those younger than 40, and in excess of 60% of patients older than 40 years of age. Asymptomatic subjects older than 40 years of age had a 5% rate of disk herniations and a 20% rate of foraminal stenosis. Teresi studied 100 asymptomatic subjects with MRI and noted asymptomatic cervical cord compression in 7% and either disk protrusion or annular bulge in 57% of subjects older than 64 years of age. In the thoracic spine, Wood studied 90 asymptomatic patients with MRI. In this population, 73% of the patients had positive thoracic imaging findings, 37% had disk herniations, 53% demonstrated disk bulges, and asymptomatic cord deformity was present in 29%.

The evidence is deep and overwhelming. The structural spine imaging findings most commonly referred to as “degenerative changes” or “degenerative disk disease,” including anterior and lateral osteophytes, loss of T2 signal in the disk, loss of disk space height, disk bulges and protrusions, and facet arthrosis, are ubiquitous and unassociated with pain syndromes; their only association is with age. They can only be avoided by a youthful death. They are not a disease state and are best referred to as normal age change or age-related change.

A consequence of this high prevalence of asymptomatic age-related changes is that the imager must know the nature of the pain syndrome if he or she is to properly focus on findings significant to the unique patient under consideration. There must be concordance of the imaging finding and the pain syndrome it is postulated to elicit. Imaging cannot prove causation, hence the need for anesthetic and provocative procedures. Communication regarding the nature of the pain syndrome is essential, whether this occurs through a robust electronic medical record, an intake document at the imaging site, or direct interaction of the imager with the patient.

Sensitivity: Physiologic Imaging

There is also a major sensitivity fault associated with spine imaging. The majority of patient symptoms referable to the spine occur in axially loaded positions, either sitting or standing. A substantial portion of radiographs and most advanced imaging (CT and MRI) are obtained in a recumbent position, removing the effects of axial load and physiologic posture. This may fail to reveal the lesion responsible for the index pain.

There is ample evidence of the effect of axial load and physiologic posture on the biomechanical and structural characteristics of the spine, derived from biomechanical, cadaver, and imaging studies. Lumbar intradiskal pressures are higher when sitting or standing than when in a recumbent position. The cadaver study of Inufusa demonstrated a reduction in the cross-sectional area of the lumbar central canal and lateral recesses in extension, with an increase in flexion. Lumbar neural foraminal cross section is also diminished in extension and increased in flexion. Fujiwara noted a reduction in cadaveric lumbar neural foraminal area with side bending or rotation toward the index foramen; an increase in area was observed with side bending or rotation away from the foramen. Studying normal volunteers, Schmid observed a 40-mm reduction in the cross-sectional area of the dural sac at the L3-L4 level with movement from flexion to extension. The lumbar neural foraminal cross-sectional area was reduced by 23% in moving from an upright neutral to an upright extended position. Danielson noted a significant decrease in the dural sac cross-sectional area with axial loading in 56% of subjects, most commonly at L4-L5; this was more common with increasing age. Dynamic reduction in dural sac area with loading was less frequent in normal volunteers than in a population of patients with neurogenic intermittent claudication. Hansson and colleagues identified the ligamentum flavum as the most important structure resulting in dynamic reduction in the lumbar central canal area under physiologic loading. Physiologic posture (lumbar lordosis) is likely more important than axial load. Multiple studies in patients with upright imaging have demonstrated the enlargement of lumbar disk bulges or protrusions with axial load, which may be further exacerbated with extension. Synovial cysts, which may be provocative of radicular pain syndromes or contribute to neurogenic intermittent claudication, may be undetectable on recumbent imaging when synovial fluid remains in the facet joint space. Upon the assumption of axial load and apposition of the facet articular surfaces, the fluid is forced from the joint space into the cyst, where it may act as a neural compressive lesion.

The cervical spine similarly exhibits dynamic physiologic change with posture and load. Cadaveric studies have demonstrated increased disk bulging and buckling of the ligamentum flavum in cervical spine extension; the ligamentum flavum effect was most significant. MRI studies of patients noted an increase in central canal stenosis with both extension and flexion of the cervical spine relative to neutral posture; the decrease was most marked in extension. Cervical neural foramina diminish in cross section, width, and height in extension and increase in all these parameters in flexion.

In summary, physiologic extension and axial load reduce the area of all lumbar spine compartments; these are increased in flexion. The cervical central canal is diminished in areas in extension more so than in flexion; it is maximal in the neutral position. Cervical foramina increase in all dimensions in flexion and diminish in extension. These dynamic changes constitute the greatest sensitivity fault in imaging: conventional supine imaging may fail to reveal a lesion that is causal of the patient’s symptoms when that lesion is only expressed in physiologic postures.

Several methods have been devised to overcome this sensitivity fault. Radiographs should always be obtained upright; this allows assessment of sagittal and coronal balance in a physiologic posture. Flexion-extension radiographs may detect instability not observed on neutral upright views, but the yield of diagnostic information is very low. In studies in both the lumbar and cervical spine segments, less than 1% of flexion-extension radiographic studies provided information over that noted on static upright radiographs. The cost and radiation exposure is best deferred to a presurgical setting, not during an initial evaluation of the back or neck pain patient.

Advanced imaging can be performed with axial loading devices on conventional CT or MRI scanners. These devices can improve the sensitivity to the detection of clinically significant central canal compromise. The 2011 North American Spine Society’s evidence-based guidelines on evaluation and treatment of spinal stenosis suggest axially loaded imaging in the setting of suspected neurogenic intermittent claudication and stenosis unconfirmed by conventional imaging, with a canal diameter of less than 110 mm. Willén and colleagues demonstrated that surgical results of cases of occult lumbar spinal stenosis detected only by axially loaded MRI were comparable to those of stenosis observed in unloaded MRI examinations.

MRI scanning in an upright position—sometimes referred to as dynamic, positional, or kinetic MRI—is now commercially available and widely marketed. The practical challenge is that currently available systems are of low field strength (0.6T), with an unavoidable reduction in image quality. This reduction in image quality can have important consequences for clinical image interpretation. If current low field strength dynamic systems were applied selectively to cases where conventional imaging failed to demonstrate a correlative lesion causal of the patient’s pain, the patient would likely benefit. All too often, however, the cost of these systems results in their routine use, or even promotion as the best available imaging tool for all spine conditions. This practice could harm patients, as diminished image quality reduces sensitivity to the detection of sinister lesions, which is the primary goal of imaging the back pain patient ( Fig. 15.1 ).

Diminished sensitivity of low field strength upright MRI to sinister lesions.

A 74-year-old female presented with bilateral lower extremity weakness and pain with a low field strength upright MRI from another institution. Sagittal T2 image (A) and axial T2 image (B) at the L4 level show degenerative spondylolisthesis at L4 with central canal compromise. She underwent an L3-L5 decompression without change in symptoms. A 1.5 Tesla MRI performed 3 weeks postoperatively demonstrated nodularity (arrows) in the cauda equina on a T2 sagittal image (C) . Gadolinium-enhanced fat-saturated T1 sagittal (D) and axial (E) images demonstrate diffuse leptomeningeal metastases from breast cancer. She died in 2 months. A high-quality preoperative MRI would have likely led to the diagnosis and avoided an unnecessary operation.

(From Khalil JG, Nassr A, Maus TP. Physiologic imaging of the spine. Radiol Clin North Am. 2012;50:599-611.)

Validity

Spine imaging must be undertaken with a full understanding of the specificity and sensitivity faults inherent in its use. The ultimate question, of course, is one of validity: Does performing an imaging study of the spine segment in question result in improved patient outcomes through a more timely and accurate diagnosis of the process causing the patient’s pain? This is well studied, particularly in the application of imaging to the acute presentation of back pain. Chou and colleagues performed a meta-analysis of the six randomized controlled trials (n = 1804) examining the role of imaging in the acute presentation of back or limb pain with no clinical features suggesting systemic disease. They identified no benefit in pain, function, quality or life, or patient-rated improvement in patients undergoing imaging (radiographs, CT, or MRI) at presentation versus those undergoing clinically directed conservative care. Although routine imaging might have been expected to provide reassurance, imaged patients did not have better psychological outcomes.

Carragee and colleagues elegantly demonstrated the lack of utility of imaging in the acute setting in a 5-year prospective observational study. A cohort of asymptomatic subjects deemed to be at risk for back pain resulting from labor-intensive vocations underwent lumbar spine MRI. This patient cohort was followed periodically over the subsequent 5 years; a subset of these subjects presented to a medical care provider with acute back or leg pain during this 5-year period and a second lumbar MRI was obtained. Less than 5% of the MRI scans obtained at the time of acute presentation with back or leg pain showed clinically relevant new findings; virtually all of the “positive findings” noted on the images at the time of presentation with back/leg pain had been present on the baseline studies obtained when the patient was asymptomatic. Only direct evidence of neural compression in patients with a corresponding radicular pain syndrome was considered to be useful imaging information. Of particular note, psychosocial factors, not the morphology seen on imaging, were the best predictors of the degree of functional disability caused by back or leg pain.

A study by Modic, also in the acute presentation of low back pain or radiculopathy, showed no relationship between the extent of disk herniations and presenting signs or symptom. The type, size, or location of a herniation at presentation, or change in size or type over time, did not correlate with clinical outcomes. MRI imaging characteristics did not have measurable value in planning conservative care. The study emphasized that the surgical decisions must be made on clinical grounds, given the inability of imaging to predict outcomes. The Modic study, like that of Carragee, demonstrated that psychosocial factors predict functional disability better than imaging parameters.

Although the depth of evidence regarding the inability of imaging to improve outcomes in back and limb pain patients is most profound in the acute setting, it applies more broadly as well. Chou and colleagues explored the seemingly counterintuitive finding that routine imaging does not lead to better outcomes in back or limb pain. This lack of utility can be attributed to the favorable natural history of back and limb pain, the low prevalence of sinister disease as causal of back pain, the weak correlation between imaging findings and symptoms (the specificity fault), and the minimal impact of imaging on clinical decision making. Given the demonstrable modest utility of imaging, the decision to undertake this path must be a carefully reasoned one.

Imaging Risk/Benefit

The decision to proceed with any medical test or procedure should be preceded by a consideration of likely benefit weighed against risk or actual harms. Certainly there are benefits to be derived from imaging. Foremost, imaging may suggest the diagnosis of previously unsuspected systemic disease. In the patient with a radicular pain syndrome or radiculopathy that has not responded to conservative therapy, imaging may supply invaluable information that allows planning of minimally invasive or surgical procedures. Negative imaging should also have value in providing reassurance that there is no sinister disease present and in stopping further workup in appropriate circumstances. Finally, in patients with chronic pain syndromes, imaging may assist in the identification of the structural or inflammatory cause of such pain. Only when a specific pain generator is identified can a specific plan of therapeutic intervention, whether it be conservative or invasive, be developed.

There are direct harms and potential risks associated with imaging, which must be balanced against potential benefits. These include radiation dose, cost, the labeling effect, and the downstream risk of provoking minimally invasive or surgical interventions of dubious efficacy.

Radiation dose in radiography, CT or CT/myelography, and nuclear medicine studies constitutes a direct patient harm. Radiation exposure from radiographs, CT, and nuclear medicine studies carries a cumulative risk of neoplasm induction. This risk becomes particularly problematic when serial studies are performed. The biologically effective absorbed radiation dose is measured by the Sievert (Sv); in North America, the average annual natural background exposure is approximately 3 mSv. A frontal and lateral chest radiograph is often considered the common currency of radiation exposure, incurring a dose of approximately 0.1 mSv. A three-view lumbar spine radiographic series is then worth approximately 15 chest radiographs, or 1.5 mSv; cervical spine radiographs incur a dose 0.2 mSv. A dose of 6 mSv is typical for a lumbar spine CT scan, a value of 60 chest radiographs. Cervical spine CT incurs 2 mSv, or 20 chest radiographs. A technetium bone scan has a dose of 6.3 mSv. The cumulative radiation exposure creates real harm; the 2.2 million lumbar spine CT scans performed in the United States in 2007 were projected to result in 1200 future cancers. Although less radiation intensive, lumbar radiography is performed much more frequently than CT and contributes nearly fivefold the cumulative radiation burden to the U.S. population. The average annual radiation exposure from lumbar radiographs is 75 times that of chest radiography. Radiation risks of spine imaging are made more acute by the necessary inclusion of radiosensitive tissues, the gonadal structures in the pelvis, and the thyroid in the neck.

Imaging studies of the spine are costly. In the United States, the medical imaging community incurs more than $100 billion of societal cost per year. The 2012 Medicare reimbursements for lumbar spine imaging include radiographs: $41; noncontrast CT: $264; myelogram: $506; noncontrast MRI: $439; whole-body positron emission tomography (PET)/CT: $1183; bone scan with single-photon emission CT (SPECT): $261. Nominal fees are typically 3 to 5 times the Medicare reimbursements. It is easy to appreciate how quickly imaging costs can accrue.

The labeling effect refers to the inevitable identification of age-related change, usually described as “degenerative change,” or “degenerative disk disease” on imaging studies obtained in evaluation of back or limb pain. The patient may then perceive that he or she suffers from a degenerative spine condition; the term degenerative has only negative connotations. Fearing further damage to their “degenerative” spine, they may give up favorite activities and exercise, resulting in deconditioning and contributing to depression. These fear-avoidance behaviors can be a major impediment to recovery. In a study in which back pain patients were randomized to either receiving or being blinded to the results of their MRI imaging, those who were privy to the results (which were all benign) had a lesser sense of well-being. A study of subacute or chronic back pain patients showed that those who underwent spine radiography reported more pain, had a diminished global health state, and consumed more follow-up care than those who were not imaged. These findings emphasize the need for patient education regarding the insignificance of age-related imaging findings; imaging professionals should carefully choose descriptive language in imaging reports and avoid the use of pejorative terms such as “degenerative change.”

Finally, and most ominously, imaging may precipitate interventions that have little evidence of efficacy and expose the patient to harm. Jarvik and colleagues documented that obtaining advanced imaging (MRI) early in a patient’s spine pain syndrome leads to increased surgical interventions despite equivalent pain and disability profiles, when compared with un-imaged patients. Likewise, Lurie and colleagues examined the dramatic regional variation (12-fold) in the rate of surgical intervention for central canal stenosis. These investigators noted that the rate of surgical intervention correlated directly with the intensity of CT and MRI use. Webster and colleagues, evaluating subjects with acute job-related back pain, noted that undergoing an MRI in the first month of symptoms was associated with an eightfold increased risk for surgery and a fivefold risk for more medical care consumption than that noted for clinically matched, un-imaged subjects. Many of the interventions directed toward spine pain, both surgical and minimally invasive, have only modest, if any, evidence basis.

Imaging Recommendations

Imaging is performed to identify sinister disease as a cause of a patient’s back or limb pain. Spine imaging has major specificity and sensitivity faults. The decision to initiate imaging must occur as a reasoned decision, weighting potential benefit against real harms and risk. The evidence is clear that there is no benefit to imaging in the acute presentation of back or limb pain in the absence of signs or symptoms of systemic disease. These principles are firmly based on evidence and have led to the imaging guidelines promulgated by several scientific societies.

These guidelines are not new. In 1994, the Agency for Health Care Policy and Research recommended against imaging patients with back pain within the first month of a pain syndrome in the absence of signs of systemic disease. The American College of Radiology (ACR) consensus practice guidelines were restated in 2009. Imaging is not indicated in the patient who presents with acute low back pain with or without radiculopathy except in the presence of “red flag” features including the following:

• •
  

  Recent significant trauma, minor trauma in a patient older than 50 years

  
  
• •
  

  Unexplained weight loss

  
  
• •
  

  Unexplained fever

  
  
• •
  

  Immunosuppression

  
  
• •
  

  History of neoplasm

  
  
• •
  

  Prolonged steroid use or osteoporosis

  
  
• •
  

  Age greater than 70 years

  
  
• •
  

  Known intravenous drug abuse

  
  
• •
  

  Progressive neurologic deficit with intractable symptoms

  
  
• •
  

  Duration longer than 6 weeks

A 2007 joint recommendation from the American College of Physicians (ACP) and the American Pain Society stated that imaging should not be obtained in patients with nonspecific low back pain. Imaging should only be performed when severe or progressive neurologic deficits are present or when serious underlying systemic disease is suspected. Furthermore, patients with signs or symptoms of radiculopathy or spinal stenosis should be imaged only if they are candidates for surgical or minimally invasive intervention (e.g., epidural steroid injection). In a further elaboration on these guidelines, the ACP, in its initiative to promote high-value medical care, provided more specific imaging recommendations based on clinical scenarios ( Table 15.3 ).

There is evidence that of the small proportion of subjects with sinister disease as the cause of their back or limb pain, virtually all have risk factors that trigger imaging under these guidelines. A retrospective study of 963 patients presenting with acute low back pain noted that the 8 subjects with neoplasm all had clinical risk factors. In a prospective study of 1170 acute low back pain patients without clinical risk factors, no cases of neoplasm were found. No sinister disease was missed in the absence of clinical risk factors in a subsequent systematic review.

Despite these well-supported, evidence-based guidelines, clinical practice in the United States remains greatly divergent from this ideal. By one estimate, between one third and two thirds of all advanced spine imaging is inappropriate when measured against existing guidelines. Utilization of spine imaging is accelerating despite a complete lack of evidence of its effectiveness in improving the outcomes of back and limb pain patients. Chou and colleagues have enumerated causes of this overutilization: inappropriate patient expectations, direct and indirect financial incentives on the part of providers, defensive medicine, and provider time constraints. These issues present great challenges to imaging professionals and those who utilize imaging to improve the clinical state of their patients. Solutions will undoubtedly be multifactorial, but education of the patient, the imaging professional, and imaging consumers would seem to be at the heart of the matter. It is hoped that the evidence presented here will assist in more rational decision making regarding imaging the spine pain patient ( Box 15.1 ).

Radiographs

With the failure of clinically directed conservative care, and having made a reasoned decision to initiate imaging of the spine pain patient, imaging should begin with upright, weight-bearing radiographs of the appropriate spine segment. The ACR and the ACP are consistent in their recommendations that patients with “red flag” features of recent trauma, osteoporosis, age greater than 70 years, or clinically suspected inflammatory spondyloarthropathy should initially undergo radiographs; advanced imaging should be reserved for patients with progressive neurologic deficits or a strong clinical suspicion of infection or neoplasm. Radiographs provide a modest sensitivity screen for sinister conditions, establish spine enumeration, and when performed in physiologic positions allow assessment of sagittal and coronal plane balance.

Spine enumeration is a critical but underappreciated role of radiographs. The typical spine morphology of 24 mobile, presacral spine segments (7 cervical, 12 rib-bearing thoracic, and 5 lumbar type vertebral bodies) is not uniformly present; deviation can result in confusion in establishing the origin of pain syndromes, or wrong segment minimally invasive or surgical interventions. It can be reasonably assumed in the human species that there are 7 cervical vertebrae. There is considerable variation in the number and distribution of thoracic and lumbar segments that may be difficult to appreciate on MRI alone; radiographs can establish this enumeration and serve as the foundation for subsequent advanced imaging description.

In a study by Carrino and colleagues using complete spine radiographs, 91.8% of subjects had 24 presacral vertebral segments: 4.8% had 23 and 3.4% had 25. Akbar and colleagues, using full spine sagittal MRI localizer images, identified 23 presacral segments in 3.3% of subjects and 25 in 3.3%; these anomalies of spine enumeration were not mentioned in the radiology report in nearly 70% of cases. In Carrino’s study, if one considers both anomalous number and distribution (e.g., 13 rib-bearing vertebrae + 4 lumbar type vertebrae) of thoracolumbar spine segments, 10.9% of subjects have nonclassical anatomy. For the spine interventionalist, this implies these situations are in the procedure room regularly; failure of recognition virtually guarantees wrong segment procedures.

Anomalous segmentation is predicted by the presence of transitional thoracolumbar or lumbosacral vertebral bodies, which may be a source of confusion in their own right. Thoracolumbar transitional segments have hypoplastic ribs at the lowest rib-bearing segment; they were present in 4.1% of subjects in Carrino’s study. Lumbosacral transitional segments have characteristics of both the L5 lumbar body and the superior sacral segment; their described prevalence ranges from 4% to 30%. The Castellvi classification ( Fig. 15.2 ) describes the morphologic types, ranging from an expanded (height > 19 mm) dysplastic transverse process (type I), through a pseudoarticulation between the transverse process and sacral ala (type II), to osseous fusion between the transverse process and the sacrum (type III). Type IV denotes the presence of a type II transition on one side and a type III on the other. Unilateral (a) and bilateral (b) subtypes are also described. On lateral radiographs or sagittal MRI images, a transitional segment can be suggested in the presence of a perceived narrow S1-S2 disk, which extends for the entire anterior-posterior (AP) width of the sacral body with parallel end plates and a squared upper sacral segment ( Fig. 15. 3 ). The presence of a transitional lumbosacral segment increased by sevenfold the likelihood of an anomalous number of presacral segments. In patients with radicular pain syndromes and lumbosacral transitional anatomy, the pain practitioner must also be aware of the possibility of an extraforaminal nerve entrapment between the enlarged transverse process and sacral ala.

Castellvi classification of lumbosacral transitional segments.

Transitional segments.

Lateral radiograph (A) and sagittal T2 MRI image (B) demonstrate typical findings of a transitional segment interspace (arrows): a narrow disk space with parallel end plates and normal T2 signal intensity. Castellvi IIa transitional segment is demonstrated in frontal (C) and lateral (D) radiographs. Note right pseudoarticulation. In another patient, frontal radiograph (E) shows a left-sided pseudoarticulation, also seen on the axial CT image ( arrows in [F] ). Narrow, parallel end plates are visible in the transitional interspace on the lateral radiograph (G) . This patient had axial pain attributed to the pseudoarticulation; this was injected (H) with relief of the index pain.

Correlation of pain syndromes with imaging findings can be confounded in patients with transitional anatomy or anomalous segmentation. It is best to consider that, in general, radicular innervation patterns remain relatively constant when counted caudally from the skull base, but the skeletal anatomy may change about them. For example, the 26th nerve (8 cervical nerves, 12 thoracic, 5 lumbar, 1st sacral nerve) from the skull base most commonly innervates the medial head of the gastrocnemius and the soleus muscles, the basis of the S1 radicular pain pattern. In a patient with 25 presacral vertebral segments, this may exit under the pedicle of the lowest lumbar type vertebral body, creating confusion for the unwary practitioner ( Fig. 15.4 ). There is also variation in typical innervation patterns in the presence of transitional anatomy, which may introduce further localization challenges. Only meticulous attention to spine enumeration, best provided by plain radiographs, sometimes supplemented by total spine MRI localizer images, will protect the spine pain patient and interventionalist from wrong segment invasive procedures ( Box 15.2 ).

The importance of segmental enumeration.

This 36-year-old male who has failed conservative care presents with a left-sided L5 radicular pain pattern involving the lateral thigh, lateral calf, and dorsum of the foot. His MRI (sagittal T2 [A] and axial T1 [B] at level of dotted line) shows a subtle disk extrusion in the left lateral recess ( arrow in [B] ) at the penultimate interspace. The lowest-most disk appears transitional. MRI scout (C) image demonstrates 24 presacral segments. Frontal and lateral radiographs ( D and E ) may suggest 5 lumbar vertebrae on cursory examination; careful counting at fluoroscopy noted 11 rib-bearing vertebrae. T12 has no ribs and L5 is transitional, Castellvi type 4. The transforaminal epidural steroid injection ( F and G ) was performed under the pedicle bearing the dysplastic left T5 transverse process and relieved the patient’s pain. Note (G) the typical appearance of a transitional disk space. Careful enumeration of every case is the only way to avoid wrong segment interventions.

When radiographs are obtained, they should be upright, weight-bearing images. Only in a physiologic posture can one assess sagittal and coronal balance ( Fig. 15.5 ). Upright radiographs demonstrate more thoracic kyphosis and lumbar lordosis than do recumbent images. Weight-bearing images may also demonstrate instability, most commonly L4-L5 degenerative spondylolisthesis, which would be occult on recumbent films. In spinal deformity patients, the exacerbation of spinal curves with weight bearing can be dramatic. Flexion-extension radiographs in the lumbar or cervical spine are not indicated as part of an initial imaging investigation. There is no role for oblique radiographs of the spine; in the lumbar region they double the gonadal dose and do not provide useful information that will affect clinical decision making. Cervical spine oblique views similarly serve only to irradiate sensitive tissue (thyroid, lens of the eye) without clinical benefit.

The effect of upright weight-bearing radiographs on balance and deformity.

Recumbent radiograph (A) shows lumbar scoliosis with a rotatory component, which is significantly exacerbated in a subsequent upright radiograph (B) . In another patient, the recumbent radiograph (C) underestimates the true scoliotic curve seen on an upright radiograph (D). Sagittal and coronal balance can only be assessed on upright, weight-bearing imaging.

Advanced Imaging Modalities

When radiographs are not explanatory of an unremitting pain syndrome or suggest underlying systemic disease, advanced imaging (CT, MRI, nuclear medicine) may be obtained. CT has undergone a revolution since the early 2000s with the advancement of multidetector technology. A data set for the lumbar spine can now be obtained in a few seconds, eliminating motion artifact and dramatically improving patient tolerance. This data set can then be reconstructed in any plane without a loss of spatial resolution or additional radiation exposure. CT provides superior imaging of cortical and trabecular bone when compared with MRI. For this reason, CT may be necessary to characterize primary bone tumors of the spine. CT also provides reasonable contrast resolution and can identify root compressive lesions such as disk herniations or characterize central canal, lateral recess, and foraminal compromise in the majority of cases. CT cannot identify intrathecal pathology and is less sensitive than MRI in the detection of early inflammatory or infectious processes, neoplasm, or paraspinal soft tissue lesions. Radiation dose must always be considered when employing CT, particularly in young patients or in serial studies. One by-product of the rapid technological advance of CT is that the literature contains no comparative studies between MRI and the latest generation of multidetector CT scanners in the detection and characterization of disk herniations.

MRI has been the dominant spine imaging modality since the 1990s, despite modest technological advancement in the realm of spine imaging in that time span. MRI has superior contrast resolution and thus the ability to distinguish between soft tissue types, allowing it to detect intrathecal pathology and identify subtle root compressive lesions. MRI has superior sensitivity in the detection of neoplasm and infection. With the use of gadolinium contrast, or heavily T2-weighted imaging sequences (short-tau inversion recovery [STIR] or fast spin echo T2 sequences with fat saturation), MRI can detect the physiologic parameters of edema, hyperemia, and inflammatory change. It has greater specificity than CT in characterizing the chronicity of fractures. With gadolinium enhancement, MRI can distinguish between recurrent disk herniation and scarring in the postoperative patient. MRI does not evaluate cortical bone well. Patient acceptance remains problematic because of prolonged imaging times and up to 10% examination failures caused by claustrophobia. Open magnets have improved patient acceptance, but at the cost of image quality. A small percentage of patients are MRI incompatible due to pacemakers, spinal cord stimulators, or other implanted devices. MRI remains costly. The sensitivity challenges imposed by recumbent MRI imaging were discussed earlier; the hope is that the engineering challenges of high field strength, weight-bearing MRI imaging will be met in the near future.

CT myelography retains a problem-solving role in the lumbar spine; it will substitute for MRI in the incompatible patient. CT myelography has superior spatial resolution when compared with MRI but lacks its soft tissue contrast resolution. It can provide an exquisite demonstration of root compressive lesions and central canal, lateral recess, and foraminal compromise. In the cervical spine, the superior spatial resolution of CT myelography and its ability to discriminate between bone and soft tissue compressive lesions give it a continuing role. CT myelography is minimally invasive, expensive, operator dependent to a degree, and also requires current CT technology to be maximally useful.

Cone-beam CT, weight-bearing myelography holds promise for assessing position-dependent pain syndromes, at least in part defeating the sensitivity fault of recumbent advanced imaging. In this technology, arising from rotational angiography roots, the C-arm rapidly rotates about the standing patient who has had contrast introduced into the thecal sac. A flat panel fluoroscopy detector gathers a data set that can be reconstructed in any plane. Soft tissue contrast does not approach that of true CT, and the longitudinal field of view is limited to the length of the flat panel detector. Despite these limitations, the high inherent contrast among intrathecal contrast, the soft tissue structures of the spinal column, and bone allows for a very good depiction of central canal, lateral recess, and foraminal compromise. This technology will only improve ( Fig. 15.6 ).

Ligamentum flavum redundancy (buckling) on upright myelography (cone beam CT).

A 66-year-old man imaged for neurogenic intermittent claudication. Conventional myelogram in the lateral plane with the patient in the prone position (A) and postmyelogram sagittal (B) and axial CT at L4-5 interspace (C) with the patient in the prone position show minimal redundancy of the ligamentum flavum (black arrows) and mild central canal compromise at the L4-5 level. Cone beam CT myelography with the patient in an upright position demonstrates a marked increase in ligamentous buckling (white arrow) on the sagittal reconstruction (D) and complete effacement of the thecal sac on the axial reconstruction (E) .

(Courtesy of Kent Thielen, MD, Mayo Clinic, Rochester, MN.)

Nuclear medicine studies are growing in importance in spine imaging. Technetium pyrophosphate bone scans detect increased blood flow and accelerated bone metabolic activity. With the addition of SPECT and SPECT/CT image fusion, significant additional spatial localization of hyperemia and increased metabolic activity are possible. This imaging is traditionally useful in assessing the burden of metastatic disease but can also be valuable in assessing non-neoplastic inflammatory states such as spondylolysis. SPECT/CT can potentially identify inflammatory synovitis in the facet and sacroiliac joints, which might guide interventions. However, validation studies of these techniques against accepted reference standards such as comparative blocks in the facet joints or intra-articular sacroiliac blocks have not yet been done. When MRI is not technically feasible, technetium bone scanning can be used to characterize the chronicity of vertebral fractures in selecting patients for bone augmentation. The technetium bone scan, in combination with gallium scan, offers sensitivity equal to MRI in the detection of spondylodiscitis. However, these techniques provide less anatomic information and MRI may ultimately be necessary to characterize the degree of central canal compromise that may influence surgical decision making. PET or PET/CT scans have an increasing role in assessing the burden of metastatic disease and in selecting lesions for percutaneous biopsy.

Imaging of Axial Pain Generators

Axial pain that may have an imaging correlate derives primarily from stimulation of nociception of the spinal articulations: the intervertebral disk, the zygapophysial or facet joints, and the sacroiliac joint. More broadly, it may include pain originating from the muscular or ligamentous structures in the supporting architecture of the spine. Axial pain is clinically characterized as constant, dull, deep, poorly localized, and aching, located primarily in the paraspinous region with inconstant referral to the proximal extremities. This is in distinction to the neuropathic pain of radicular character, which is typically sharp, electric, lancinating, and experienced in a bandlike distribution into the more distal extremities. The prevalence of axial spine pain generators has been well described by Depalma ( Table 15.4 ). Intervertebral disk disruption (IDD) was the most common axial pain generator in this series, followed by the facet joint, the sacroiliac joint, and insufficiency fractures of the spine or pelvis. This work also emphasized the age dependence of these pain sources. Patients with diskogenic pain (IDD) were significantly younger than patients with facet or sacroiliac joint pain. As age increased, the probability of IDD as a pain source decreased and the probability of facet or sacroiliac pain increased, up to approximately age 70. A later multivariable analysis also showed a gender relationship, with IDD more prevalent in young men. This echoes the earlier work of Schwarzer, who demonstrated that lumbar facet–mediated pain was uncommon in a population of young workers (∼15%) but more highly prevalent in an aged population (∼32%).

Imaging Correlates of DisKogenic Pain

The imaging diagnosis of diskogenic pain is made challenging by the lack of a pathoanatomic gold standard against which to assess imaging parameters. It is not possible to evaluate a disk, either at surgery or upon histologic examination, and deem it painful. The current most stringent reference standard for the diagnosis of diskogenic pain is manometrically controlled provocation diskography with normal control levels as documented in the practice guidelines of the International Spine Intervention Society (ISIS).

It is important to observe, however, that examination of the same body of evidence regarding the validity of diskography as the reference standard has resulted in diametrically opposed recommendations regarding its use by different physician societies. The ISIS, the North American Spine Society, and the International Association for the Study of Pain accept diskography as a useful diagnostic tool in back pain patients and recommend its use. The American Pain Society rejects diskography as a diagnostically useful test. A comprehensive review of diskography in the journal of the American Society of Regional Anesthesia and Pain Medicine notes that whereas CT diskography is the gold standard for the assessment of structural disk alteration, there is no convincing evidence that the use of diskography as a selection tool improves surgical outcomes. Any analysis of imaging findings in diskogenic pain patients thus remains based on a reference standard (provocation diskography) that is ultimately unproved against a pathoanatomic gold standard. This is further confounded by evolution of the criteria for a positive diskogram since the 1990s. For the purposes of this discussion, only concordant pain responses were considered to represent a positive diskogram. A significant concordant pain response without a specification of pain intensity or the use of manometry is defined as a Walsh criterion. Inclusion of the requirement for a normal control disk elevates the criteria to that of the International Association for the Study of Pain (IASP). None of the studies reviewed here meaningfully used manometric control or met the ISIS criteria for positive diskography.

Although challenging, there is motivation to make the diagnosis of diskogenic pain via noninvasive imaging. Diskography has until recently been considered a minimally invasive and nondestructive test. There is now in vitro and in vivo evidence suggesting that disk puncture or diskography may contribute to disk dysfunction. Korecki and colleagues noted that in a bovine disk model, single punctures with a 25-gauge needle resulted in biomechanical degradation of disk function with cyclic loading. Carragee and coworkers demonstrated on 10-year follow-up MR imaging that asymptomatic subjects who had undergone investigational diskography showed more degenerative phenomena than did matched control subjects. Although the clinical significance of these observations remains uncertain, noninvasive diagnosis is desirable.

The specificity fault inherent in spine imaging was previously discussed: manifestations of disk “degeneration” are ubiquitous, usually asymptomatic, and primarily represent normal age change. In a population of symptomatic patients with suspected diskogenic pain, however, are there imaging findings that correlate with positive provocation diskography? The findings evaluated in the literature include (1) loss of disk space height, (2) generalized alterations in T2 signal within the disk, (3) alterations of disk contour, (4) Modic end plate marrow changes, and (5) the presence of high intensity zones (HIZ) or fissures within the posterior disk annulus. These imaging features are examined initially as independent variables with subsequent discussion of the more limited literature in which they are combined in a multivariate analysis. A significant portion of the presented data were drawn from a systematic review of imaging and clinical markers of axial pain generators in the lumbar spine performed by Hancock and colleagues. Additional studies not included in that report or published subsequent to it have been added. A common set of measures was compiled from the many studies: sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), and likelihood ratios (LRs). When imaging features were quantified (T2 signal loss in the disk was reported as normal, moderate, or severe), a threshold was used. Original data were combined and recalculated to reflect setting a detection threshold as moderate (including moderate and severe cases) or severe only. Because the diagnosis of diskogenic pain may provoke therapeutic interventions (most of which carry risk and have unproved efficacy), emphasis was placed on those measurements that inform about false-positive results: specificity (true negatives/ true negatives + false positives) and PPV (true positives/true positives + false positives).

Diskogenic Pain

Nuclear T2 Signal Loss

Studies addressing the correlation between MR imaging evidence of disk “degeneration,” primarily nuclear T2 signal loss, and diskogenic pain reach back to the 1990s. The challenges in this analysis are well illustrated in Table 15.5 . Although disk “degeneration” definitions are inconsistent, original data have been recalculated into threshold values to provide a reasonable basis of comparison. The changes in diskography criteria over time are also noted. Despite these shortcomings, conclusions can be reasonably drawn. The NPV of disks of normal nuclear signal is uniformly high and –LRs are highly informative; disks of normal nuclear signal are rarely painful. Severe, uniform loss of T2 signal, with or without loss of disk space height, is a finding of high specificity (88% to 96% in studies using a three-part classification system) with strongly informative + LR. Disks with severe T2 signal loss are rarely nonpainful. The utility of this finding is reduced by its low prevalence (it is found in 13% to 25% of disks undergoing diskography) and low sensitivity (23%, 24%, 37%, and 70% in the studies with a three-part classification system). Disks with intermediate signal loss may be painful but with less certainty ( Fig. 15.7 ).

Table 15.5

T2 Signal Loss as a Predictor of Positive Provocation Diskography

Author (ref), Date	Diskogram Criteria	T2 Signal Criteria	Prevalence	Sensitivity	Specificity	PPV	NPV	+LR (CI)	−LR (CI)
Osti ( ) 1992	Walsh	Moderate + Severe	47%	70%	64%	50%	80%	1.9 (1.3-2.7)	0.49 (0.3-0.8)
		Severe only	13%	23%	92%	60%	70%	2.8 (1.1-7)	0.83 (0.7-1)
Horton( ) 1992	Walsh	Moderate + Severe	69%	95%	43%	44%	94%	1.6 (1.2-2.2)	0.18 (0.04-0.9)
		Severe only	20%	37%	88%	58%	74%	2.8 (1.1-7.3)	0.72 (0.5-1)
Ito ( ) 1998	Walsh	Moderate + Severe	63%	96%	46%	34%	97%	1.7 (1.4-2.2)	0.14 (0.03-0.6)
		Severe only	25%	70%	89%	64%	91%	5.7 (3-11)	0.36 (0.2-0.7)
Weishaupt ( ) 2001	IASP	3-5 of 5-grade Pearce	65%	98%	59%	64%	98%	2.3 (1.8-3.1)	0.05 (0.01-0.3)
Lim ( ) 2005	Walsh	4 and 5 of 5-grade Pearce	62%	88%	52%	50%	89%	1.8 (1.4-2.4)	0.25 (0.1-0.6)
Lei ( ) 2008	IASP	3 and 4 of 4-point Woodward	57%	94%	77%	78%	94%	4 (2.5-6.4)	0.07 (0.02-0.2)
O’Neill ( ) 2008	IASP	Moderate + Severe	62%	90%	67%	75%	86%	2.7 (2.2-3.3)	0.16 (0.1-0.2)
		Severe only	15%	24%	96%	87%	54%	6 (3-11.7)	0.79 (0.7-0.9)
Kang ( ) 2009	IASP	3, 4, and 5 on Pfirrmann 5-point scale	70%	95%	39%	34%	96%	1.6 (1.3-1.8)	0.12 (0.03-0.5)

PPV, positive predictive value; NPV, negative predictive value; LR, likelihood ratio; IASP, International Association for the Study of Pain.

From Maus TP, Martin DP. Imaging for discogenic pain. In: Kapural L, Kim P, Deer T, eds. Diagnosis, Management, and Treatment of Discogenic Pain. Philadelphia: Elsevier/Saunders; 2012, p 38, Table 3-5.)

Disk height and nuclear T2 signal loss.

Lateral radiograph (A) of a 50-year-old male with intractable axial pain. Note loss of height of lumbosacral disk, which contains gas. There is slight retrolisthesis of L4 on L5 and L5 on S1. Sagittal fat-saturated T2-weighted MRI (B) shows loss of T2 signal in the L4-5 and L5-S1 disks; normal upper lumbar disks. Axial T2-weighted images at L3-4 (C) , L4-5 (D) , and L5-S1 (E) demonstrate normal L3-4 disk, loss of T2 signal in L4-5 with a small central herniation, and a broad bulge at L5-S1. Sagittal CT diskogram (F) and axial images at L3-4 (G) , L4-5 ( H, and L5-S1 (I) shows normal L3-4 disk, Grade IV annular disruption at L4-5 and L5-S1 with leak of contrast from the right posterolateral annulus at L5-S1. Patient had concordant axial pain at L4-5 and L5-S1 with a normal control disk at L3-4.

(From Maus TP, Aprill CN. Lumbar discogenic pain, provocation diskography, and imaging correlates. Radiol Clin North Am. 2012;50:681-704.)

Modic End Plate Changes

The functional unity of the disk and the cartilaginous end plate is manifest in signal changes within the end plate and adjacent subchondral marrow that accompany nuclear matrix degradation. End plate marrow changes were originally classified by Modic in 1988 ( Fig. 15.8 ). Type I change represents ingrowth of vascularized granulation tissue into sub–end plate marrow; it exhibits a hypointense T1 and hyperintense T2 signal on MRI and may enhance with gadolinium. Type II change exhibits elevated T1 and T2 signal and reflects fatty infiltration of sub–end plate marrow. Type III change is hypointense on T1 and T2; it correlates with bony sclerosis. Type I change is thought to represent an active inflammatory state, with type II more quiescent and type III post-inflammatory. Ohtori and colleagues noted elevated levels of protein gene product 9.5–immunoreactive nerve fibers and TNF-α immunoreactive cells in the cartilaginous end plates of patients with Modic changes. The immunoreactive nerve ingrowth was noted exclusively in patients with diskogenic low back pain. TNF-α immunoreactive cells were more common in type I end plate changes.

Modic end-plate changes.

Modic I Sagittal T2 (A) and T1 (B) images demonstrate elevated T2 and diminished T1 signal involving the superior half of the L5 vertebral body. The histologic correlate of Modic I is vascularized granulation tissue. There are disk herniations at the L3-4 and L4-5 disks.

Modic II Sagittal T2 (C) and T1 (D) images reveal elevated T1 and T2 signal in the sub–end plate marrow involving L2-L5. The histologic correlate of Modic II is fatty infiltration.

Modic III Sagittal T2 (E) and T1 (F) images show diminished T2 and T1 signal in the sub–end plate marrow about the L5 interspace. The histologic correlate of Modic III is sclerotic bone.

Modic end plate changes do carry an association with low back pain, particularly type I change. Toyone and colleagues found that 73% of patients with type I change had low back pain as opposed to 11% of type II patients. Likewise, Albert and Manniche reported low back pain in 60% of patients with Modic changes but in only 20% of those without Modic changes. Type I change was more strongly associated with low back pain than type II change. Modic type I change may also be associated with segmental instability. Persistent type I change after fusion surgery raises concern for pseudarthrosis; patients with solid fusions more likely have either persistent type II change or resolution of all marrow abnormality.

The data in Table 15.6 describe Modic-type end plate changes as predictors of diskogenic pain. The studies discussed previously suggested that type I Modic change would be more strongly correlated with positive provocation diskography than type II; that conclusion is not borne out by the data, which suggest that either type I or type II Modic change correlates with positive diskography, although not uniformly. The studies of Braithwaite and colleagues, Weishaupt and colleagues, Lei and colleagues, and O’Neill and colleagues had few false-positive results (i.e., disks with adjacent type I or type II end plate changes that were nonpainful). The specificity, PPV, and + LRs in these studies were high. The usefulness of the MR imaging findings were only hampered by their infrequency. Other studies were less supportive but may be confounded by technical flaws. The compelling results presented by Weishaupt and colleagues suggest a significant threshold for marrow change; with a threshold at 25% of vertebral body height, there were no false-positive results in this series. A reasonable conclusion is that Modic type I or II marrow change of this severity (25% of vertebral height) is an infrequent but highly specific finding, with a high PPV for diskogenic pain.

Table 15.6

End Plate (Modic) Change as a Predictor of Positive Provocation Diskography

Author (reference), Date	Diskogram Criteria	Modic Type	Prevalence per Disk	Sensitivity	Specificity	PPV	NPV	+LR	−LR
Braithwaite (105) 1998	Walsh	I + II	25% imaged 15% tested	24%	96%	91%	47%	6 (1.7-21.2)	0.80 (0.7-0.9)
		I	4% tested	5%	100%	100%	42%	7.4 (0.4-131)	0.95 (0.9-1)
		II	12% tested	18%	96%	89%	48%	4.4 (1.2-16.1)	0.86 (0.8-0.9)
Ito (91) 1998	Walsh	I + II + III	9%	23%	94%	56%	80%	4 (1.3-12.8)	0.82 (0.7-1)
Weishaupt (97) 2001	IASP	I	14%	29%	97%	88%	66%	9.9 (2.4-41.6)	0.73 (0.6-0.9)
		II	9%	19%	99%	90%	63%	12.75 (1.7-97.3)	0.83 (0.7-0.9)
		I + II	22%	48%	96%	88%	72%	10.86 (3.5-34.1)	0.55 (0.4-0.7)
		I + II Mod + Severe	16%	38%	100%	100%	69%	52.1 (3.2-844)	0.63 (0.5-0.8)
Kokkonen (106) 2002	Walsh	I	17%	19%	85%	41%	65%	1.25 (0.5-3)	0.96 (0.8-1.2
		II	19%	19%	80%	35%	64%	0.96 (0.4-2.2)	1 (0.8-1.2)
		I + II	36%	38%	65%	38%	65%	1.1 (0.6-1.8)	0.95 (0.7-1.3)
Lim (92) 2005	Walsh	I + II	14%	9%	83%	21%	62%	0.6 (0.2-1.7)	1.1 (0.9-1.3)
Lei (98) 2008	IASP	I + II	14%	32%	98%	94%	62%	19.25 (2.7-140)	0.69 (0.6-0.8)
O’Neill (93) 2008	IASP	I	4%	6%	99%	88%	49%	6.94 (1.6-29.9)	0.95 (0.9-1)
		II	4%	7%	99%	90%	50%	8.32 (1.9-35.5)	0.93 (0.9-1)
		I + II	8%	14%	98%	89%	51%	7.63 (2.8-21.2)	0.88 (0.8-0.9)
Kang (94) 2009	IASP	I + II	13%	14%	87%	26%	76%	1.08 (0.5-2.6)	0.99 (0.9-1.1)

PPV, positive predictive value; NPV, negative predictive value; LR, likelihood ratio; IASP, International Association for the Study of Pain.

From Maus TP, Martin DP. Imaging for discogenic pain. In: Kapural L, Kim P, Deer T, eds. Diagnosis, Management, and Treatment of Discogenic Pain. Philadelphia: Elsevier/Saunders; 2012.

High Intensity Zone (HIZ)

In 1992, Aprill and Bogduk described the HIZ as an imaging marker of a painful disk at provocation diskography. Their definition of the HIZ is a

high-intensity signal (bright white) located in the substance of the posterior annulus fibrosis, clearly dissociated from the signal of the nucleus pulposis in that it is surrounded superiorly, inferiorly, posteriorly and anteriorly by the low intensity (black) signal of the annulus fibrosis and is appreciably brighter than that of the nucleus pulposis. Page 362

This finding was identified on a midsagittal T2-weighted image; it may occur centrally in an otherwise normal annulus, in a bulging annulus, or be located superiorly or inferiorly behind the edge of the vertebral body in a severely bulging annulus. In a series of 500 consecutive patients, the per-patient prevalence was 29%; the per-disk prevalence of an HIZ was 6%. The majority of HIZs were present at the L4 and L5 disk levels, confirmed on later studies.

The relationship of the HIZ to pain production was evaluated in a subset of 41 patients, selected for the presence of an HIZ on prediskography MR imaging. Diskography was performed with the requirement of a nonpainful control disk for a diagnosis of diskogenic pain. Pain responses were tabulated both as “exact” reproduction of pain as well as “similar” pain ( Table 15.7 ). In all, 118 disks were tested in 41 patients; the per-disk prevalence of the HIZ was 34%, reflecting the selection bias. In detecting exact pain, the HIZ had a sensitivity of 82% and specificity of 89%, a 70% PPV, and a + LR of 7.3. When the diskographic criteria were relaxed to exact or similar pain, the specificity rose to 97% with a PPV of 95%; there were only two false-positive cases where a disk bearing an HIZ was nonpainful. The investigators postulated that the HIZ represents a complex grade 4 fissure where the nuclear material has been trapped within the lamellae of the annulus fibrosis and become inflamed, accounting for the T2 hyperintensity, brighter than that of the parent nucleus. They advanced the HIZ finding as pathognomonic of a symptomatic disk. The publication of these findings elicited considerable interest and many subsequent studies attempting to verify or refute its conclusions.

Table 15.7

High Intensity Zone (HIZ) as a Predictor of Positive Provocation Diskography

Author (reference), Date	Diskogram Criteria	HIZ Criteria	Prevalence per Disk	Sensitivity	Specificity	PPV	NPV	+LR (CI)	−LR (CI)
Aprill ( ) 1992	IASP Exact pain	Aprill	34% ∗	82%	89%	78%	91%	7.3 (3.9-13.7)	0.21 (0.1-0.4)
	Exact or similar pain	Aprill		63%	97%	95%	72%	18.4 (4.6-72.7)	0.38 (0.3-0.5)
Schellhas ( ) 1996	IASP	Schellhas †	60% ∗	97%	83%	87%	97%	5.7 (3.5-9.3)	0.03 (0.01-0.11)
Ricketson ( ) 1996	Walsh	Aprill	9%	12%	92%	57%	54%	1.5 (0.4-5.6)	0.96 (0.8-1.1)
Saifuddin ( ) 1998	Walsh	Aprill	18%	27%	94%	89%	47%	4.8 (1.7-14.2)	0.77 (0.7-0.9)
Ito ( ) 1998	Walsh	Aprill	20%	52%	89%	60%	87%	4.8 (2.3-10.2)	0.54 (0.4-0.8)
Smith ( ) 1998	Walsh	Aprill	13%	27%	90%	40%	80%	2.6 (1.2-5.6)	0.82 (0.7-1)
Carragee ( ) 2000	Walsh	Carragee ‡	30%	45%	84%	73%	62%	2.8 (1.5-5.5)	0.7 (0.5-0.9)
Weishaupt ( ) 2001	IASP	Aprill	20%	27%	85%	56%	62%	1.8 (0.8-3.7)	0.86 (0.7-1)
Peng ( ) 2006	Walsh	Aprill	12%	NC	NC	100%	NC	NC	NC
Lei ( ) 2008	Walsh	Aprill	19%	25%	87%	62%	57%	1.8 (0.8-4.1)	0.87 (0.7-1.1)
O’Neill (93) 2008	IASP	O’Neill § 1+2+3 Intensity grades	28%	44%	89%	82%	60%	4.1 (2.7-6.1)	0.62 (0.5-0.7)
		2+3	16%	26%	95%	86%	54%	5.7 (3-10.9)	0.78 (0.7-0.8)
		3	9%	15%	98%	86%	52%	6.8 (2.7-17.1)	0.87 (0.8-0.9)
Kang ( ) 2009	IASP	Aprill	26%	57%	84%	53%	86%	3.46 (2.2-5.5)	0.52 (0.4-0.7)

PPV, positive predictive value; NPV, negative predictive value; LR, likelihood ratio; IASP, International Association for the Study of Pain.

From Maus TP, Martin DP. Imaging for discogenic pain. In: Kapural L, Kim P, Deer T, eds. Diagnosis, Management, and Treatment of Discogenic Pain. Philadelphia: Elsevier Saunders; 2012:43, Table 3-7.

∗ Sensitivity and prevalence values are not meaningful due to this selection bias.

† Includes lesions with thin line of T2 hyperintensity within annulus or connecting nucleus to HIZ

‡ Includes posterolateral lesions, HIZ signal intensity within 10% of CSF T2 signal

§ Schellhas criteria plus posterolateral and lateral lesions

Table 15.7 demonstrates informative + LR and high specificity in most of the studies. There are dissenting voices. Carragee and colleagues evaluated the correlation of the HIZ and painful disks at provocation diskography in both symptomatic and asymptomatic subjects. In symptomatic patients, 30% of disks had an HIZ; only 9% of disks in asymptomatic subjects contained an HIZ (significant, P < 0.0001). The statistics for the symptomatic group are presented in Table 15.7 ; an HIZ disk had an 84% specificity, a 73% PPV, and a + LR of 2.8 for diskogenic pain. These data are supportive of the HIZ as a useful marker for the painful disk. The investigators, however, point out that the presence of an HIZ disk was strongly predictive of a positive pain response at diskography in both the symptomatic (73%) and asymptomatic (69%) groups. The asymptomatic group had also been stratified by psychometric evaluation; in participants with either chronic pain or abnormal psychometric studies, all HIZ disks produced pain with pressurization. The investigators contend that the similar painful response rate of HIZ disks in symptomatic and asymptomatic subjects devalues the HIZ as a useful finding, because the total weight of diagnosis depends on concordance versus nonconcordance of pain response.

The study by O’Neill and associates stratified HIZ lesions by relative signal intensity, and expanded the definition of the HIZ to include posterolateral and lateral lesions and those that demonstrated a connection to the nuclear compartment. Original data were recalculated to establish a three-part threshold of HIZ intensity: markedly intense cases, markedly and moderately intense cases, and a combination of all three. As the threshold tightened, the specificity and +LR rose; the PPV remained high for all three threshold levels. For only markedly hyperintense HIZs, the specificity was 98%, the PPV was 86%, and the +LR was 6.8. This would support Bogduk’s comments that “low intensity zones may well occur in asymptomatic volunteers, but that when activated (ostensibly inflamed), these fissures become painful and assume a higher signal intensity” ( Fig. 15.9 ) page 1260.

High Intensity Zone (HIZ).

Middle-aged male has previously undergone L4-5 decompressive procedure, without diskectomy, for back and leg pain. Axial back pain was unrelieved. A, Sagittal T2-weighted MRI shows loss of T2 signal in the L4 disk with an HIZ in the posterior annulus. D, Axial T2 MRI image at L4-5 interspace demonstrates the HIZ (arrow) in the posterior annulus. E, Axial-enhanced T1 MRI image showing enhancement in the HIZ, also demonstrated in the sagittal fat-saturated T1 image (B) . Sagittal postdiskogram CT (C) demonstrates annular fissure at L4-5 leading to HIZ. Pressurization of the L4-5 disk produced concordant axial pain at < 20 pounds per square inch (PSI) above opening pressure.

(From Maus TP, Aprill CN. Lumbar discogenic pain, provocation diskography, and imaging correlates. Radiol Clin North Am. 2012;50:681-704.)

Imaging Correlates: Conclusion

Imaging identification of diskogenic pain is challenging for a variety of reasons. (1) There is no pathologic or surgical gold standard. (2) The existing standard of comparison, diskography, is ultimately unproven, subjective in its interpretation, and has evolved over time in its criteria for a positive test. None of the studies reviewed earlier use the most current and restrictive criteria, those of ISIS. (3) The imaging findings likely have threshold effects, where only a significant expression of the finding (intensity of an HIZ, extent of marrow change) is a useful predictor of diskogenic pain. Most studies do not account for this factor. (4) Imaging findings are likely technique dependent to an unknown degree, and imaging techniques are evolving.

From the tangle of data, useful information can emerge; the imaging correlates of diskogenic pain are summarized in Box 15.3 . These imaging predictors of diskogenic pain, when applied to a select population of patients with axial back pain, in whom other pain generators have been excluded, could perhaps be used to initiate a proven therapy possessing a good safety profile. No such therapy exists. Diskography remains the reference standard for the diagnosis of diskogenic pain.

Box 15.3

Only gold members can continue reading. Log In or Register to continue

Share this:

• Click to share on Twitter (Opens in new window)
• Click to share on Facebook (Opens in new window)

Related

Related posts:

The History of Pain Medicine A Review of Pain-Processing Pharmacology Prediction and Prevention of Persistent Post-Surgical Pain Central Post-stroke Pain Syndrome Managing Pain during Pregnancy and Lactation Minor and Short-Acting Analgesics, Including Opioid Combination Products