Diagnostic blocks for chronic pain

2.1 Test blocks

Test blocks are ones that are performed simply to test if a particular nerve is involved in the patient’s symptoms. Typically, the nerve supplies different tissues that may be the source of pain, or can be itself the origin of a neuropathic pain condition. The block is not diagnostic, for the block cannot distinguish one source of pain from another; it serves only to test if the nerve blocked is responsible for mediating the pain. Examples include blocks of the median, ulnar or radial nerve for pain in the hand, and ilioinguinal and iliohypogastric nerve blocks for pain after herniotomy.

The nature of the information gained by test blocks and its utility have been poorly elaborated. At present, it is unclear whether these blocks add information that contributes to the diagnostic process and whether the results have therapeutic utility.

2.2 Prognostic blocks

A prognostic block is one undertaken to test if a treatment will be successful. The rationale is that if a nerve block with local anaesthetic relieves the pain, then a treatment capable of treating the primary lesion or interrupting conduction along the nerve should relieve the pain for a prolonged period.

For prognostic blocks to be valid, evidence is required to complete a table like that depicted in Table 1. There needs to be a strong association between a positive response to the block and a successful outcome from treatment, and between a negative response to the block and failure of treatment. Such data are hard to produce: they would require a study in which patients undergo treatment irrespective of their response to blocks. They also require a treatment that is dependably effective. Unless that is the case, failures may be due to the failure of treatment rather than an error in the response to blocks. Few such treatments exist. For these reasons, there is little evidence behind the practice of prognostic blocks.

2.3 Diagnostic blocks

Diagnostic blocks are used to identify the anatomical structures that are the source of pain. The rationale is that if a structure is the source of pain, then anaesthetizing it or its nerve supply should relieve the pain. If the suspected structure is not the source of pain, anaesthetizing it should not relieve the pain. The essential difference from “test blocks” is that the target nerve supplies only one defined anatomical structure that is suspected to be the pain generator. It is important to recognize that diagnostic blocks can reveal the source of pain, but are not suitable to identify the pathology that is the cause of the pain (e.g. inflammatory, neuropathic, etc.). Diagnostic blocks are the major focus of the present review.

3.1 Cervical medial branch blocks and third occipital nerve blocks

The medial branches of the cervical dorsal rami supply the zygapophysial joints C3-7 and their medial posterior muscles (multifidus, interspinales, semispinalis). The C2-3 zygapophysial joint is innervated by the third occipital nerve, which is the superficial medial branch of the C3 dorsal ramus. Blocks of these nerves would relieve pain from any of these structures. However, there are no known causes of chronic neck pain that lie in the posterior neck muscles. Therefore, in practice, cervical medial branch blocks are a test for pain stemming from the cervical zygapophysial joints.

Cervical medial branch blocks are the only means available by which either to implicate or to refute the cervical zygapophysial joints as the source of neck pain. Clinical examination [2] and imaging techniques [3] have no proven validity. Even assessing mechanical pain sensitivity using quantitative pressure algometry does not distinguish affected from asymptomatic joints, indicating that palpation of the joint is not diagnostic for this condition [4].

It has been shown that cervical medial branch blocks are target specific [5], provided that small volumes are injected [6]. Single medial branch blocks have false-positive rates of 27% [7]. Patients can report relief of pain for reasons other than the effect of a local anaesthetic injected during a diagnostic block, e.g. as the result of placebo effect. Therefore, in order to be valid, cervical medial branch blocks must be controlled in each patient. Many practitioners find limitations in the clinical applicability of placebocontrolled blocks. Comparative blocks constitute an alternative paradigm. These involve administering a particular agent on the occasion of the first block, but using a different agent on a second occasion. The agents advocated are lignocaine and bupivacaine. A positive response requires short-lasting relief when lignocaine is used, and long-lasting relief when bupivacaine is used.

Two types of positive response to comparative blocks can occur. A concordant response is one in which pain is completely relieved for a duration concordant with the duration of action of the local anaesthetic used. A discordant response is one in which pain is completely relieved but the duration of effect is longer when a short-acting agent is used. In the context of cervical medial branch blocks, both responses have been validated [8]. Concordant responses have a high specificity (88%) when compared with placebo-controlled blocks. This means that positive responses are very unlikely to be false. However, concordant responses have a low sensitivity (54%). This means that they do not detect all patients with cervical zygapophysial joint pain. Discordant responses have a high sensitivity (100%), i.e. they detect all patients who have zygapophysial joint pain. However, their specificity is limited (65%). That means that positive responses can include some placebo responses.

For practical purposes, concordant and discordant responses are equally valid for the diagnosis of cervical zygapophysial joint pain, because its prevalence is high [9,10]. For a prevalence of 60%, concordant responses yield a diagnostic confidence (positive predictive value) of 87%, and discordant responses yield a diagnostic confidence of 81% (Table 2). The lower specificity of discordant responses is offset by their higher sensitivity. Thus, if an operator accepts that a positive response is complete relief of pain on each of two occasions that the block is performed, irrespective of duration of response, the reward is that all legitimate cases are detected, and only about 1 in 5 will be false positive.

Cervical medial branch blocks have established utility: those patients who obtain complete relief from cervical medial branch blocks can expect a 70% chance of achieving complete relief of pain after cervical medial branch neurotomy [11,12,13,14].

3.2 Lumbar medial branch blocks

The medial branches of the lumbar dorsal rami supply the zygapophysial joints of the lumbar spine and their medial posterior muscles (multifidus, interspinales). If defined strictly, lumbar medial branch blocks would relieve pain from any of the structures innervated by these nerves. However, there is no known pathology that affects the muscles supplied by the lumbar medial branches, which can be a cause of chronic pain. Therefore, in practice, medial branch blocks are a test for pain from the zygapophysial joints.

No studies have identified the pathology responsible for zygapophysial joint pain. Changes as seen on computed tomography do not correlate with the joint being painful [3] or with low back pain [15]. There are preliminary findings that a form of synovitis, demonstrable by magnetic resonance imaging, might be the underlying pathology, but this pathology has still to be correlated with relief of pain following blocks [16].

The validity of lumbar medial branch blocks has been studied more than any other block in pain medicine. The blocks are target specific. Provided that the correct target points are used, and provided that small volumes are used, the injectate remains exclusively on the target nerve [17]. Face validity has been demonstrated. Medial branch blocks protect normal volunteers from experimentally induced zygapophysial joint pain [18].

Single blocks of the lumbar medial branches are not valid, because they have false-positive rates of 38% [19]. This means that if the prevalence of lumbar zygapophysial joint pain is 15% [20], for every three blocks that appear to be positive, two will be falsepositive. Such yields of false-positive results are unacceptable and indicate that a valid diagnosis cannot be made using a single diagnostic block. In order to reduce false-positive diagnoses, controlled blocks must be used in every case.

Comparative local anaesthetic blocks have been advocated as controls for lumbar medial branch blocks on the grounds that they have been validated for the cervical spine. In the cervical spine, comparative blocks have a specificity of either 65% or 88%, depending on the stringency of the criteria for a positive response [8]. These values are acceptable for cervical medial branch blocks, because cervical zygapophysial joint pain has a high prevalence [9,10,21]. However, in the lumbar spine, where zygapophysial joint pain is far lesscommon [20], comparative blocks result in inordinate numbers of false-positive results [22].

If the liberal criteria of a discordant response are applied, the specificity of comparative blocks will be 65%. If the prevalence of lumbar zygapophysial joint pain is 40%, the diagnostic confidence will be 66% (Table 3). This means that 1 in 3 positive responses will be false. In younger patients, and in injured workers, the prevalence of lumbar zygapophysial joint pain may be as high as 15% or as low as 5% or less, depending on whether 50% or complete pain relief after the block is considered as a criterion to define zygapophysial joint pain [23,24]. If the prevalence drops to 15%, the diagnostic confidence falls to 32%. If the prevalence is as low as 5%, the diagnostic confidence is only 13% [22] (Table 3).

If the stricter criteria of a concordant response are applied, the specificity is 88%. For a prevalence of 40%, the diagnostic confidence is 85% (Table 3). For a prevalence of 15%, the confidence falls to 60%; and for a prevalence of 5% is falls to 30% (Table 3). In such populations, the false-positive rates, even of concordant responses, are prohibitively high. Between 4 in 10 and 7 in 10 responses will be false.

Under these conditions, discordant responses are not valid. Concordant responses are valid if the prevalence is 40%, as observed in elderly patients [20], but in other populations concordant responses are not valid. In those populations placebo controls become the only means of securing the diagnosis [22].

Establishing a diagnosis of lumbar zygapophysial joint pain allows treatment to be offered in the form of lumbar radiofrequency medial branch neurotomy [25]. According to an observational study, if they are selected on the basis of positive responses to controlled blocks, 60% of patients so treated can expect to obtain at least 80% relief of pain, lasting 12 months; and 80% can expect at least 60% relief [26]. If pain recurs, neurotomy can be repeated in order to reinstate relief [27].

3.3 Spinal nerve blocks

In some patients with radicular pain, medical imaging is not diagnostic. Computed tomography or magnetic resonance imaging may show more than one spinal nerve affected by a disc herniation or by foraminal stenosis, and the clinical features do not help to determine which nerve is the source of symptoms. Spinal nerve blocks were developed to assist surgeons to determine the symptomatic level in such cases.

Partial data on the validity of spinal nerve blocks are available from studies that report the proportion of patients with positive responses to blocks in whom pathology is found at surgery. This proportion is not the sensitivity of the test, but is its positive predictive value in the specific sample tested.

For lumbar spinal nerve blocks, various studies have reported positive predictive values that ranged between 80% and 100% [28,29]. These high values imply that the false-positive rate of lumbar spinal nerve blocks is low and, therefore, that their specificity is likely to be high. In one study, a small number of patients with negative responses to blocks nevertheless underwent surgery [29]. Pathology was found in all cases, but not at the level tested; in no case did surgery find pathology at the level tested when the block had been negative. These patients proved to have multilevel disease or anomalous nerve roots. This implies that the false-negative rate of lumbar spinal nerve blocks is low and, therefore, that their sensitivity is high.

One study measured sensitivity by performing lumbar spinal nerve blocks in 46 patients with clinical and radiological evidence of nerve root compression, subsequently confirmed at surgery [30]. The sensitivity was 100%, with 95% confidence intervals of 88-100%. That same study measured specificity by performing blocks in 23 patients at levels known by radiology not to be symptomatic. No false-positive responses were encountered, and the specificity of lumbar spinal nerve blocks was estimated as “around 90%”.

The above data are limited by the fact that single, rather than controlled blocks were performed, thereby exposing the results to possible false-positive responses. The only study that used controlled blocks enlisted 18 patients with cervical radiculopathy and 83 with lumbar radiculopathy, but did not stratify the results according to region investigated [31]. Using a criterion standard of good outcome from surgery, it found a sensitivity of 93% but a specificity of only 33%. Spinal nerve blocks were no better than magnetic resonance imaging in predicting good outcome, but the particular advantage of spinal nerve blocks, demonstrated in this study, was the ability of negative responses to blocks to predict poor outcome, particularly when MRI was negative or ambiguous.

The summary of the above data suggest that spinal nerve root blocks may have diagnostic utility for radicular pain and perhaps prognostic value for the outcome of surgery.

A note of caution needs to be made. Transforaminal injections have caused severe complications, such as paraplegia or death, probably due to spinal cord infarction [32,33,34,35]. A likely mechanism is obliteration of radicular arteries by the injection of particulate steroids [36]. We are not aware of the occurrence of such events after diagnostic blocks, which are performed only with local anaesthetics. Nevertheless, the use of methods to minimize the risk of these complications [36] is imperative.

3.4 Sympathetic blocks

Blocks of various elements of the sympathetic nervous system are used to test if the patient’s clinical features are mediated by sympathetic nerves. The use of fluoroscopy or ultrasound establishes the face validity of these blocks, i.e. the sympathetic trunk is accurately and selectively infiltrated. Face validity is confirmed physiologically by observing a rise in skin temperature in the limb of the target side, which shows that the sympathetic trunk has, indeed, been anaesthetized.

What has not been shown is that sympathetic blocks have construct validity, i.e. that the physiological response is due to the effects of local anaesthetic and not to non-specific factors (i.e. placebo). This requires the execution of controlled blocks, but sympathetic blocks have mostly been exempt the requirement for controls. Interestingly, a study using a controlled paradigm in complex regional pain syndrome showed that stellate ganglion blocks with normal saline were as effective as blocks with local anaesthetic [37]. Statistically, a difference arose in that patients who received local anaesthetic tended to have relief of pain lasting into the next day. Accordingly, the authors recommended that diagnostic decisions be based, not on the immediate response, but on the response evident on the next day. However, this is not a secure operational criterion, because some patients who received normal saline also had prolonged responses. The only way to secure the validity of stellate ganglion blocks is to perform controls in every patient. Unless this is done, the investigator cannot distinguish a true-positive response from a placebo response.

Failing to use controlled blocks in the past may have led to overestimates of so-called sympathetic-mediated pain. Unless and until studies are conducted using controlled blocks, its true prevalence will not be known.

To our knowledge, the prognostic validity of sympathetic blocks in relation to outcomes of sympathectomy has never been investigated according to the principles illustrated in Table 1.

3.5 Intra-articular blocks

Intra-articular injections are performed on joints of the appendicular skeleton for a variety of reasons. However, diagnostic blocks, using local anaesthetic agents, are not common practice. The diagnosis of peripheral joint pain is established largely on the basis of physical examination and medical imaging. Some practitioners might use joint blocks in patients in whom the source of pain is ambiguous, such as hip joint blocks to distinguish hip joint pain from referred pain to the groin or buttock; but such practices are not widely used, and their validity and utility have still to be established.

Diagnostic joint blocks have a greater application in the diagnosis of spinal pain, because the joints of the spine are less accessible to physical examination, and medical imaging is typically unhelpful in pinpointing a particular joint as the source of pain. Intra-articular blocks of the zygapophysial joints are used as a means of diagnosing zygapophysial joint pain. However, intra-articular blocks of the spine have never been validated and have not been shown to have therapeutic utility. They have largely been supplanted by medial branch blocks, which have been validated, and whichdohave therapeutic utility. It is unclear whether joint pain is the result of intra- or periarticular pathology, or both. If periarticular lesions are involved, intra-articular blocks would not be sensitive for the diagnosis of joint pain. Periarticular infiltrations are performed in clinical practice, but their validity and therapeutic utility is unknown.

There are two spinal joints for which nerve blocks are not practical. They are the lateral atlanto-axial joint and the sacroiliac joint. There are few prevalence studies of lateral atlanto-axial joint pain. One study found that at least 16% of patients presenting with headache responded to lateral atlanto-axial joint blocks [38]. Another study found a prevalence of at least 13% [39]. Each of these may be underestimates because not all eligible patients underwent diagnostic blocks. Conversely, the figures might be lower because controlled blocks were not used.

Injection of contrast medium in order to establish that the injectate enters the joint cavity and does not escape from it establishes the face validity of the injection for the diagnosis of C1-2 cervicogenic headache. The construct validity of lateral atlanto-axial joints has not been tested. Single diagnostic blocks may have a falsepositive rate, which has not been measured. In principle, controlled blocks are required in order to maximize validity. Controls using comparative local anaesthetic blocks might be used, but the validity of comparative blocks inside joints, where blood flow is low, has not been established. Anatomical controls might be an alternative, in which the control is to block an adjacent joint (e.g. C2-3), but the results of such controls have not yet been reported.

Lateral atlanto-axial joint blocks have therapeutic utility because patients can be treated by atlanto-axial arthrodesis. Follow-up studies have reported that arthrodesis succeeds in providing sustained relief of pain [40,41,42]. Intra-articular injections of corticosteroids are a more conservative alternative, but provide long-term benefit only in about one in eight patients [39].

Sacroiliac joint blocks are a test for back pain stemming from a sacroiliac joint. Two types of controls have been used in order to maximize the validity of sacroiliac joint blocks. One study used anatomical controls, by previously anaesthetizing lumbar zygapophysial joints [43]. Another study used comparative local anaesthetic blocks [44]. Respectively, these studies found a prevalence (95% confidence intervals) of 13% (6-20%) and 19% (9-29%). Medical history and clinical examination did not correlate with the results of sacroiliac joint blocks, indicating lack of validity [44,45]. However, a recent systematic review found that different types of physical examination have discriminative value for sacroiliac joint pain, but warned that the gold standard, i.e. sacroiliac joint injection, has no proven validity [46].

The therapeutic utility of sacroiliac joint blocks is still evolving. Various studies have explored the efficacy of percutaneous radiofrequency denervation of the joint, in patients who are positive to blocks [47]. The magnitude and duration of relief from these procedures has still to be established by large trials.

3.6 Blocks for discogenic pain

There is no validated method for the diagnosis of discogenic pain. Provocative discography is widely used, but opinions on its validity are polarized [48].While supporters of this method underscore the unique feature of associating an objective morphological study with the subjective experience of pain, opponents stress the risk of false-positive responses [49] and the danger of producing long-term damage to the tested discs [50].

Local anaesthetic blocks are possible alternatives to provocation discography, but research on this topic is sparse. Twomethods have been explored: intradiscal blocks, and sinuvertebral nerve blocks.

Intradiscal blocks involve injecting local anaesthetic agents into the putatively painful disc [51]. Relief of pain would establish the disc as the source of pain. The few studies that have been published do not provide compelling evidence of validity. Responses have been assessed only immediately after injection; prolonged responses have not been assessed [52]. Few patients report complete relief of pain. The criterion for a positive response has been 50% relief of pain, but such responses are open to question. They imply that blocks are as much ineffective as they are effective. Whether the remaining pain has an organic or psychological basis is not known. No studies have subjected the diagnostic blocks to controls. In essence, intradiscal blocks have yet to be subjected to the same rigour as other spinal blocks. A potential problem is that, after intradiscal injection, local anaesthetics may fail to permeate the disc and reach the innervated outer third of the anulus, resulting in false-negative responses. Notwithstanding these limitations, there is some evidence that relief of pain following intradiscal blocks predicts better outcome after surgery than does provocation discography [51].

Sinuvertebral nerve blocks are an attempt to block the nerves that innervate the disc. In essence they constitute a field block, which overcomes the problem of having the local anaesthetic reach the nerve endings responsible for the pain. The one study that has appeared, to date, showed that sinuvertebral nerve blocks had a sensitivity of 73% [53]. The specificity was not evaluated, because in that exploratory study the block was performed only on patients with a positive diagnosis of discogenic pain. The block had a target specificity of only 40%, when strict criteria were used. The latter arised largely because the injectate frequently reached the nearby spinal nerve, as assessed by computed tomography. Refinements in technique and evaluation of the diagnostic specificity are required before sinuvertebral nerve blocks are ready for greater use.