CHAPTER 67 Cervical Discography: Diagnostic Value and Complications
INTRODUCTION
Diagnostic procedures can confirm specific diagnoses. Establishing a correct diagnosis should allow prediction of outcome following appropriate treatment. To evaluate the diagnostic value of a procedure one must determine how well the test can discriminate between the diseased and nondiseased state. A diagnostic test is accepted or rejected after evaluating evidence, based on an accepted methodology. One must ask, is there historical need for the test, are there studies that validate acceptable specificity and sensitivity, are there other comparable, less invasive procedures, is the test clinically useful, is the incidence of significant morbidity low, and finally are serious complications preventable?1
Historical need
Ralph Cloward acknowledged that George Smith performed the first cervical discography in1952, but both independently described the technique and use of cervical discography in the late 1950s.2–4 These surgeons were performing anterior cervical fusions for chronic axial pain and were searching for a procedure to identify the painful cervical discs. Both were frustrated by the insensitivity of available diagnostic tests. Plain film could diagnose disc degeneration but there was poor correlation with pain. Myelography with Pantopaque could identify central and paracentral disc herniation but the false-positive incidence was high, and myelograms were insensitive to lateral lesions.
Both surgeons felt that reproduction of a patient’s usual pain during disc injection was evidence that the disc was symptomatic. Cervical fusions were performed based on this information. They recognized that most discs were anatomically disrupted,3,5 but not all were painful. Controversy arose because many surgeons felt that cervical spine surgery should not be performed for neck pain without radiculopathy. Cervical discography was berated because of the high prevalence of structurally abnormal cervical discs and because injection into most cervical discs provoked pain.2–16 None of the early reports mentioned the imprecise methods used to perform and evaluate discography. Many surgeons perceived cervical discography as a useless procedure.9,12,16 On the other hand, there were others who believed in discogenic pain, used discography to identify painful discs, and fused cervical segments found to be positive.3,5,7,13
Sensitivity and specificity
We can indirectly evaluate the specificity of cervical discography. Specificity is the measure of how well cervical discography can rule out a disc as a pain generator when the pain is caused by some other structure. Specificity equals true negative divided by (false positive + true negative). Ideally, one would perform discography on patients with chronic axial neck pain and patients with neck pain referred from other body regions. A less robust design would be to perform discograms on asymptomatic volunteers. This design, however, might give a higher false-positive rate because concordance of pain cannot be evaluated in asymptomatic volunteers. As with lumbar discography,17–19 a significant proportion of asymptomatic individuals with degenerated discs will have low-level pain during disc injection. A positive response should include a limiting verbal pain score (e.g. > 6–7/10 pain). Provoked pain below that level does not constitute a positive response. Alternatively, one could identify patients with painful Z-joints as the source of neck symptoms and perform discography on them. Patients with a positive discogram response who also had complete relief of their pain following placebo-controlled analgesic medial branch blocks are discogram false-positive responders.
Soon after the separate introduction of cervical discography by Smith and Cloward in the late 1950s, Holt reported on the ‘fallacy of cervical discography.’ He performed discography on 50 ‘normal’ subjects who were inmates at the Missouri State Penitentiary.12 A total of 148 discs in 50 subjects were studied using 22-gauge needles placed in the neck blindly, using X-rays to document needle placement in the discs. Each ‘extremely co-operative’ inmate was given either Demerol or Nembutal 20 minutes before the procedure. Each disc was injected with 50% sodium diatrizoate (Hypaque sodium) and ‘… even in 0.2 cc amounts, great pain was produced in every subject at every space.’ Holt reported that every disc hurt when injected (i.e. the specificity was 0%). He observed that the pain lasted 5 minutes then subsided. Pain was variously described as ‘being hit in the back of the neck with a maul,’ ‘like sticking a knife into the base of my neck,’ or ‘a hot poker between the shoulder blades.’ He described almost 100% incidence of contrast extravasation despite a volumes of only 0.2–0.3 mL. According to Holt, only one inmate reported to sick call and all others resumed normal activity the next day without apparent problems. He did conclude that cervical discography might be made relevant by a radical change in contrast media and ‘techniques of performance.’ He politely conceded that further studies should be done to establish objective findings, including control patterns in normal individuals.
A decade later in 1975, Holt reflected on the value of cervical discography by rehashing his previous observations and commenting on Struck’s study of 1200 patients undergoing cervical discography followed by about 800 surgical fusions.20 Holt comments, ‘Colorado has had statehood slightly less than 100 years. It is hard to conceive of so many ruptured cervical discs in all its history.’ He called for a ‘hard and fast moratorium on cervical discography.’21
Other early observational studies also noted a high incidence of painful disc injections.
Meyer discounted the use of cervical discography because almost all the radiographic patterns were abnormal and he felt ‘there was no good correlation of pain radiation with the patient’s clinical symptoms or the roentogenographic findings.’10 Klafta and Collis discounted the 89% of positive pain provocation in patients because injection of contrast into cervical discs shown to have disc herniation on myelography did not usually reproduce the patient’s radicular pain.16
Despite negative bias, suboptimal imaging, and the use of irritating contrast, cervical discography was criticized but not challenged with new data for many years. In 1993, Shinomiya et al. in Japan partially reconfirmed Holt’s findings.22 These authors performed discography on 128 patients with spondylotic mylelopathy and radiculopathy. In the 72 patients with neck pain, 65% had reproduction of their usual pain. However, in the control group who had only neurologic symptoms, 50% reported pain provocation during injection.
In the same year, Aprill and Bogduk found a potential source of the high false-positive rate of cervical discography while studying the prevalence of Z-joint and discogenic pain in patients with chronic axial neck pain.23 The authors performed both diagnostic medial branch blocks using a double-block protocol and cervical discography on a series of patients. Forty-one percent of patients had a positive discogram, but also had full relief of their axial neck pain for the duration of the local anesthetic following medial branch block with 0.5 mL of bupivacaine. These were false-positive discograms.
In 1996, Kurt Schellhas, an experienced interventional radiologist, reported a 0% false-positive rate by using more precise methods and strict criteria for defining a positive response.24 He published a prospective observational study in which discography was performed on 40 discs in 10 asymptomatic volunteers and compared the results to 10 symptomatic patients. He used modern discography technique including 25-gauge needles, fluoroscopic guidance, and nonirritating contrast material. He also precisely graded pain intensity on a 10-point verbal pain scale, recorded the exact pain referral pattern, disc morphology, and pain concordance in patients. A minimum of four levels were studied in each patient. The mean age of the volunteers was 30 years. The disc morphology was abnormal in 35 of the 45 discs studied. While most abnormal discs did produce pain, the average pain was 2.42/10 (SD = 1.5). The most frequent response (14 discs) was 2–2.5/10 intensity. Pain intensity was 0–3/10 in 27 discs and 3–6/10 in 9 others. An intensity of 6/10 pain was reported in only one disc. By comparison, there were only two morphologically normal discs in the symptomatic group, but the average pain intensity and standard deviation was not calculated. The patient raw data show that there were 12 discs with minimal (0–3/10) pain responses, 11 with moderate (4–6/10) pain, and 17 discs with severe pain (7–10/10). In those moderately painful discs which might be most likely to be false-positive responses less than 50% (5/11) had concordant pain.
Competing diagnostic tests
MRI
Structural correlation
Patients with predominately axial neck pain, who have failed conservative care, and have unacceptable persistent symptoms may be surgical candidates. A surgeon must determine which level or levels are causing the pain prior to any operation. Because of the high sensitivity of MRI, some authors suggest that MRI can replace discography.25,26 MRI does demonstrate disc morphology, disc degeneration, and the contour of the outer anulus. But the clinical significance of these abnormalities in the absence of neural compression and radiculopathy is unclear. Are MRI findings of disc degeneration or minor protrusion of clinical importance?27,28
The presence of annular tears and loss of hydration are non-specific findings and many asymptomatic individuals have significant degenerative MRI structural abnormalities.29 Normal discs are uncommon after the mid-twenties and the development of posterior lateral fissures through the uncinate region are present in the majority of mature cervical discs.30 Since the early use of cervical discography, authors have reported the high incidence of annular tears in most cervical discs,2–12,31 but the sensitivity of an MRI scan compared to discography for detecting degeneration and annular tears in cervical discs is still debated.
Gibson et al. examined 22 patients with both MRI and discography and on the basis of gross degeneration reached the conclusion that MRI scan was better for detecting gross degeneration.25 On the other hand, Schellhas et al. prospectively correlated MRI imaging and discography in 10 asymptomatic subjects and 10 patients with chronic axial pain.24 The authors found that various degrees of annular disruption were common in both symptomatic and asymptomatic subjects. The ruptures were usually posterolateral and full-thickness tears and leaks into the epidural space occurred in about 50% of the discs. Annular tears were identified during discography in 23 of 24 discs judged normal on MRI scans.
Pain provocation correlation
If one could predict pain provocation during discography based on the degree of signal changes, annular tears, or minor posterior annular contour abnormalities, discography is not necessary. Although prior studies correlating MRI findings with pain provocation during lumbar discography suggest that dark discs are often associated with positive pain provocation,32,33 this relationship in the cervical spine is less clear.
Parfenchuk and Janssen correlated the relationship between the MRI and CT scan obtained after cervical discography.34 They found no correlation between the MRI scan and the post-discogram CT scan to the pain response during discography. Although there was a significant correlation between the patterns seen on the MRI scan and discography, MRI had a high false-positive and false-negative rate compared to discography. They found that unlike lumbar discs, 63% of the ‘white’ cervical discs (normal) studied by discography had a positive pain response. The authors concluded that dark discs need not be studied by discography, but their study included few dark discs. Their findings were not supported by Zheng et al., who found only 63% of the dark discs were symptomatic.35 Furthermore, Schellhas et al. found that in their symptomatic patient group, 16 discs that were clearly abnormal on MR were negative to discographic testing, and 8 of 10 discs with normal MRI scans were shown to have fissures and provoke concordant pain during discography. In the asymptomatic group, 21 of 40 discs had abnormal morphology and none was symptomatic during discography.24
Surgical predictive value based on MRI compared to discography
There are no prospective randomized studies, but Zheng et al. indirectly arrived at the possible false-negative and false-positive rate for surgery based on MRI scans.35 This study retrospectively reviewed surgical outcome of 55 patients, between 1990 and 1995, who had clinically diagnosed cervical discogenic pain for at least 6 months and underwent an anterior cervical discectomy and fusion using a Simmons keystone technique and were available for follow-up at a minimum of 2 years. All had a preoperative MRI scan, discography, and a post-discogram CT scan. MRI disc morphology was classified as white, speckled, or dark and the integrity of the posterior anulus described as flat, bulging, torn, or small herniation. Discs that were thought to be symptomatic included dark herniated, dark torn, dark bulging, speckled herniated, speckled torn, and white herniated discs. Discography was performed by expert interventionists at an average of 2.9 levels using standard modern techniques. Discography was positive if injection of contrast provoked moderate or severe familiar pain. The MRI findings in 13 of 55 (24%) patients and 103 of 161 (64%) injected discs correlated completely with the results of discography. In the 79 discs with a normal MRI scan 58 (73%) had painful (positive) disc injections. Discography spared 21 asymptomatic discs from being fused and identified 21 discs that were deemed symptomatic but had a normal MRI scan. Since 58 of 79 levels with positive discography were abnormal on MRI, the MRI scan sensitivity was 73.4% (or a false-negative rate of 26.6%). There were 82 levels with negative discography and 40 of these levels had normal MRI scans resulting in an MRI specificity of only 49% when compared to positive discogram responses. Fifty-five patients had a total of 79 levels fused with overall clinical results of excellent in 49%, good in 27%, fair in 18%, and poor in 6%. The results showed that positive discography occurred in 59% of small herniated and torn discs, 35% of bulging discs, and 29% of flat discs. Dark discs with small herniations were the most commonly fused levels and 67% of these discs had a positive discogram.
