Lack of uniform diagnostic criteria for cervical radiculopathy in conservative intervention studies: a systematic review

We used the search from that identified from the search of our review on the effectiveness of conservative interventions in CR [12]. The search strategy followed the recommendation by the Cochrane Handbook for Systematic Review of Interventions [13]. Electronic searches included Cochrane Controlled Trials Register, MEDLINE, EMBASE, and CINAHL from 1966 up to October 2010. Manual searches of review bibliographies and reference lists of primary studies were undertaken to obtain possible studies not captured by the electronic searches. Two librarians together with a review author (ET) performed the electronic search (“Appendix”). Studies were included that evaluated a conservative intervention in a randomised clinical trial (RCT) in patients with CR. Different from our previous review [12], this time the type of outcome measures or the type of comparison interventions used were not taken as an exclusion criterion.

From each included study, the diagnostic criteria used to define the diagnosis of CR (not criteria related to the intervention) and the definitions for CR were extracted.

Three reviewer authors (ET, RO, AdB) performed the data extraction independently, using a pre-determined data extraction form (available from the authors).

The criteria for patient selection were qualitatively and quantitatively evaluated on uniformity. We divided selection criteria into clinical symptoms, clinical testing, diagnostic imaging, and exclusion criteria. Clinical symptoms were subdivided into pain and sensory symptoms. Clinical tests were subdivided into pain provocation tests, changes in range of motion, and neurological examination, e.g. motor disturbances and reflex changes. We aimed to identify either corresponding or contradictory diagnostic tests and features of CR. Items were considered to correspond if they described the same test, cluster of tests, or feature in labelling “cervical radiculopathy” (e.g. “Spurling’s compression test” or “combined lateral flexion and extension”) provoking neck and/or arm pain. Our conclusion of consensus between criteria or definitions across the studies was (arbitrary) set at 75 %: if more than 75 % of studies set a certain criterion. Criteria were considered to be contradictory when the item was a reason for inclusion in one article and a reason for exclusion in another.

Figure 1 (PRISMA Flow Diagram), [14] identifies the study selection process. We included 17 articles reporting on 13 studies [15‐30]. Four articles reported on one single trial [23‐26] and two articles were identical, but one was published in Dutch [31] and one in English [22].

Two studies included patients with no other specific selection criteria besides having CR [15, 29]. Table 1 presents the results of data extraction.

Six studies used a definition of CR [17, 19, 20, 22, 27, 30]. Two studies [17, 20] used identical definitions. Four definitions [17, 19, 20, 27, 30] mentioned nerve root compression resulting in neck pain radiating to the arm. We concluded that there was no consensus on a definition in the literature.

Imaging methods were used in four studies for the identification of patients with CR [18‐20, 23].

All four studies used magnetic resonance imaging (MRI) in their selection criteria and two of these studies mentioned the use of MRI to reveal a herniated intervertebral disc [18, 19]. One study included patients with pathological lesions corresponding to the C7 nerve root detected by MRI, but the exact descriptions of the pathological lesions were unclear [20]. One study used MRI and radiographs (X-rays) as part of the neurological examination, but it was unclear when patients were regarded as eligible based on the imaging results [23].

In conclusion, imaging methods were not uniformly used in studies to label patients as having CR.

All but three studies explicitly stated exclusion criteria [15, 17, 29]. Eight studies excluded patients with specific pathology or medical ‘red flags’, although the examples mentioned differed [16, 18, 20, 21, 23, 27, 28, 30]. Planned imminent treatments (surgery or injection), previous injuries or surgery of the spine, use of medication, clinical signs of cord compression, abnormal neurological signs, or the inability to tolerate the planned intervention were also mentioned as exclusion criteria across the different studies.

Overall, many studies mentioned exclusion criteria, but we found no uniformity in the criteria used to exclude patients with symptoms of CR. No criteria were considered to be contradictory.

In the literature, we found no other SR on selection criteria or definitions of CR. A qualitative review on the diagnosis of CR mentioned conventional neurologic examination findings (testing of strength, muscle stretch reflexes, and sensation) and cervical ROM as a part of the clinical examination procedure [9]. This review suggested the use of a test item cluster for diagnosing CR. A systematic review of the diagnostic accuracy of provocative tests of the neck for diagnosing CR suggested that, when consistent with the history and other physical findings, a positive Spurling‘s, traction/neck distraction and positive Valsalva’s test might be indicative of a CR, while a negative Upper Limb Tension Test might be used to rule it out [10].

Recently, a work group consensus statement from the North American Spine Society suggested defining CR from degenerative disorders as “pain in a radicular pattern in one or both upper extremities related to compression and/or irritation of one or more cervical nerve roots” [32]. This review found six studies mentioning definitions, all comprising a cluster of likely or possible symptoms including neck and arm pain and varying degrees of sensory, motor, and reflex changes. It is likely that the lack of a consensus definition for CR has contributed to the different findings with regards to the effectiveness of treatments for CR reported in the literature [12, 32, 33].

With regard to criteria for the inclusion of patients, many authors stress the value of elaborate history taking [9, 34]. Except on pain, however, we found no specific mentioning of history taking being included in the diagnostic criteria of studies.

Neck pain radiating into the arm or cervicobrachial pain is a common feature in CR, but not a distinguishable one per se, as it could well have other causes such as a thoracic outlet syndrome [35‐37] or referred pain [37]. Nerve root compression is supposed to result in both neuropathic and nociceptive pain [38, 39]. Possibly, future selection criteria should also take these different natures of pain into account. A more specific definition of CR should therefore include an ability to distinguish the nerve root pain present in CR from other musculoskeletal disorders and other neurological conditions, such as brachial plexus pathology, pseudo-radicular pain or a peripheral nerve entrapment, or even non-specific neck pain [1, 6].

We did not find any mention of the use of validated questionnaires for distinguishing neuropathic from nociceptive pain such as the Neuropathic Pain Questionnaire (NPQ) [40], ID Pain [41], Leeds Assessment of Neuropathic Symptoms and Signs (LANSS) [42], Douleur Neuropathique 4 (DN-4) [43], or painDETECT [44]. Because neuropathic pain has a worse prognosis than nociceptive pain and there are treatments that specifically target neuropathic pain, future studies could consider using the questionnaires to select and tailor participants to different treatment strategies [45].

Diagnostic labelling could well include results from physical examination. Only five studies mentioned selection of participants based on clinical (provocation) tests [18, 20, 22, 27, 30]. However, there is a lack of primary studies investigating the accuracy of these tests. The reported heterogeneity between the various studies, as well as numerous methodological problems, precludes any strong recommendations for the use of these tests, especially in the primary care setting. In the absence of other clinical information or corroborating evidence, the value of these tests should therefore be interpreted with caution.

Diagnostic imaging is used to confirm the presence of a clinically suspected CR. The diagnostic accuracy of imaging is thought to be limited, because asymptomatic radiological abnormalities are commonly seen with advanced imaging studies [46]. This also holds true for plain X-ray studies that exhibit degenerative changes increasing with age unrelated to clinical signs and symptoms [1]. Most often MRI was used, although data concerning the sensitivity and specificity for the diagnosis of CR for MRI are sparse and questionable [46, 47]. A more recent study on the interobserver reliability of MRI evaluation in patients with CR has shown it to be substantial for nerve root compression, with or without previous clinical information [48]. In our study, only RCTs in a conservative treatment group were included, implying that we selected a group of patients often seen in primary health-care settings. We suppose that diagnostic imaging is much less often used in this setting, as the access to it is limited. This is supported by the North American spine surgeons’ clinical guidelines for the diagnosis and treatment of CR from degenerative disorders. They recommended that MRI was suggested only for the confirmation of correlative compressive lesions in cervical spine patients who failed a course of conservative therapy and who may be candidates for interventional or surgical treatment [32].

This is the first SR on definitions and selection criteria for CR. Even though CR is often mentioned as a separate entity in the assessment of patients with neck pain, most studies use their own criteria to sub-classify patients with CR, making comparisons difficult.

One of our limitations was that we only included published RCTs. Relevant unpublished trials were not included, thus potentially leading to publication bias. However, if they exist, these studies are likely to be small and to increase heterogeneity and therefore are unlikely to change our results [49].

For the purpose of this review, we have chosen to exclude studies that only included surgical interventions, as surgical studies may possibly recruit a different group of patients to justify the need for an invasive treatment.

We recommend to aim for consensus on a definition of CR and selection criteria among experts in this area, preferably with criteria that are quantifiable with validated clinical tests. Data from two recent RCTs support the clinical finding that patients usually report experiencing more arm pain than neck pain [19, 22]. We therefore suggest defining CR as: “Radiating pain in the arm with motor, reflex and/or sensory changes (such as paraesthesiae or numbness), provoked by neck posture(s) and/or movement(s)”. It is the researchers’ intent to conduct a Delphi study on this definition among different international researchers and practitioners. We would suggest the following selection criteria: pain radiating into the arm and motor, reflex, and/or sensory changes in the upper limb such as paraesthesia or numbness.

Future research should aim at the validity of selection criteria.