The value of any test depends upon the a priori certainty of the diagnosis in question, and this principle applies to electrodiagnostic testing. For a condition or diagnosis for which there is great certainty before additional testing, the results of the subsequent tests are of limited value. The concept of diminishing returns on the road to diagnostic certainty is an important one. For instance, an EMG test will be of limited value in confirming the diagnosis of radiculopathy in a patient with acute-onset sciatica while lifting, L5 muscle weakness, a positive straight-leg raise, and an MRI showing a large extruded L4-5 nucleus pulposus. This is because the a priori certainty of the diagnosis was very high. In contrast, an elderly diabetic patient with sciatica, limited physical examination findings, and equivocal or age-related MRI changes presents an unclear picture. In this latter case the electrodiagnostic testing is of high value, placing in perspective the imaging findings and also excluding diabetic polyneuropathy or plexopathy as confounding conditions.

The American Association of Neuromuscular and Electrodiagnostic Medicine (AANEM, formerly AAEM) guidelines recommend that for an optimal evaluation of a patient with suspected radiculopathy, a needle EMG screen of a sufficient number of muscles and at least one motor and one sensory nerve conduction study should be performed in the involved limb (7). The nerve conduction studies are necessary to exclude polyneuropathy. The sufficiency of the EMG screen and a recommended number of muscles is discussed in detail below. An EMG study is considered diagnostic for a radiculopathy if EMG abnormalities are found in two or more muscles innervated by the same nerve root and different peripheral nerves, yet muscles innervated by adjacent nerve roots are normal (8). This assumes, of course, that other generalized conditions such as polyneuropathy are not present.

EMG study of bilateral limbs is often necessary, particularly if a single limb shows EMG findings suggestive of radiculopathy and the patient has symptoms in both the studied and the contralateral limb. If bilateral limbs are involved, then the electrodiagnostician should proceed by studying selected muscles in an upper limb (if the lower limbs are abnormal on EMG) or a lower limb (if both upper limbs are abnormal) to exclude a generalized process such as polyneuropathy or motor neuron disease. Likewise, additional nerve conduction studies are appropriate to exclude other suspected conditions, and the electrodiagnostician should have a low threshold for expanding the study.

H reflexes have commonly been used to determine whether a radiculopathy demonstrates S1 involvement (8). This is a monosynaptic reflex that is an S1 nerve root-mediated response and can
differentiate to some extent L5 from S1 radiculopathy. Many researchers have evaluated the sensitivity and specificity of the H reflex with respect to lumbosacral radiculopathies and generally found a range of sensitivities from 32% to 88% (8,9,10,11,12,13). However, many of these studies suffered from lack of a control group, imprecise inclusion criteria, or small sample sizes.

Marin et al (13) prospectively examined the H reflex and the extensor digitorum brevis muscle stretch reflex in 53 normal subjects, 17 patients with L5, and 18 patients with S1 radiculopathy. Patients in the study had all of the following: low back pain radiating to the leg; reduced sensation, weakness, or positive straight leg raise test; and either EMG evidence of radiculopathy or structural changes on MRI or CT. The H reflex maximal side-to-side latency difference was 1.8 ms, as derived from the normal group. They analyzed the sensitivity of the H reflex for side-to-side differences greater than 1.8 ms or a unilaterally absent H reflex on the affected side. The H reflex demonstrated only 50% sensitivity for S1 radiculopathy and 6% for L5 radiculopathy but had 91% specificity. H reflex amplitudes were not assessed in this study. These results suggest that the H reflex has a low sensitivity for S1 root level involvement but may help differentiate L5 from S1 root involvement.

H reflexes may be useful to identify subtle S1 radiculopathy, yet there are a number of shortcomings related to these responses. They can be normal with radiculopathy (13), and because they are mediated over such a long physiologic pathway, they can be abnormal due to polyneuropathy, sciatic neuropathy, or plexopathy (8). They are most useful in the assessment for polyneuropathy.

To interpret a latency or amplitude value, and to render a judgment as to the probability that it is abnormal, precise population-based normative values encompassing a large range of ages of normal subjects must be available to which to compare these nerve conduction findings. Falco et al (14) demonstrated in a group of healthy elderly subjects (60 to 88 years old) that the tibial H reflex was present and recorded bilaterally in 92%. Most elderly persons are expected to have normal H reflex studies, and when abnormalities are found in these persons, the electrodiagnostician should critically evaluate these findings and the clinical scenario before attributing H reflex abnormalities to the aging process.

In patients with upper limb symptoms suggestive of cervical radiculopathy, H reflexes and F waves are not useful in diagnosis but rather help exclude polyneuropathy as an underlying cause of symptoms. One study by Miller et al (15) examined the H reflexes in the upper limb in a set of patients defined by a combination of clinical criteria (no imaging or EMG studies, however) as having definite or probable cervical radiculopathy. They tested the H reflex for the flexor carpi radialis, extensor carpi radialis, and abductor pollicis brevis muscles and also evaluated the biceps brachii heteronymous reflex. The later reflex is derived by stimulating the median nerve in the cubital fossa and recording over the biceps brachii muscle, averaging 40 to 100 trials. These reflex studies had a 72% sensitivity overall for the group, with 100% for the subset of patients with definite cervical radiculopathy. In contrast, needle EMG demonstrated 90% sensitivity for the definite group. Although these findings suggest a possible role for these upper limb H reflexes, they are highly specialized, time consuming, and difficult to consistently elicit. They may have a role in sensory radiculopathies where needle EMG will not be positive and imaging findings are equivocal. Further studies are necessary to clarify whether these findings of Miller et al (15) can be duplicated in other centers.

F waves are late responses involving the motor axons and axonal pool at the spinal cord level. They can be assessed and classified by using the minimal latency, mean latency, and chronodispersion or scatter (8). As in the case of H reflexes, F waves demonstrate low sensitivities and are not specific for radiculopathy; rather, they are a better test to screen for polyneuropathy. Published sensitivities in radiculopathies range from 13% to 69%, but these studies suffer from many of the shortcomings that are found in the H reflex studies (9,16,17).

London and England (18) reported two cases of persons with neurogenic claudication from lumbosacral spinal stenosis. They demonstrated that the F wave responses could be reversibly
changed after 15 minutes of ambulation that provoked symptoms. This suggested an ischemia-induced conduction block in proximal motor neurons. A larger-scale study of this type might find a use for F waves in the identification of lumbosacral spinal stenosis and assist with the delineation of neurogenic from vascular claudication.

Standard motor and sensory nerve conduction studies are not helpful in identifying a cervical or lumbosacral radiculopathy; however, they should be performed to screen for polyneuropathy and to exclude common entrapment neuropathies if the patient’s symptoms could be explained by a focal entrapment. It is important to remember that based upon the anatomy of the dorsal root ganglion, sensory responses should be normal in most radiculopathies. If they are found to be absent, this should increase suspicion for another diagnosis, such as polyneuropathy or plexopathy.

Plexopathies often pose a diagnostic challenge, as they are similar to radiculopathies in symptoms and signs. To distinguish plexopathy from radiculopathy, sensory nerve conduction test responses that are accessible in a limb should be tested. In plexopathy, they are likely to be reduced in amplitude, whereas in radiculopathy they are generally normal. If substantial axonal loss has occurred at the root level, the compound muscle action potential recorded in muscles innervated by that root may be reduced in both plexopathies and radiculopathies. This is usually when severe axonal loss has occurred, such as with cauda equina lesions or penetrating trauma that severely injures a nerve root. The distal motor latencies and conduction velocities are usually preserved as they reflect the fastest-conducting nerve fibers (8).

The AANEM guidelines recently examined the literature and concluded that somatosensory evoked potentials (SEPs) may be useful for cervical spondylosis with cord compression. Likewise, in lumbosacral spinal stenosis, dermatomal SEPs (DSEPs) may be useful in defining levels of deficits (7). These tests are not necessary for electrodiagnostic testing for persons with suspected radiculopathies, and their usefulness is limited to special circumstances.

DSEPs can document physiologic evidence of multiple- or single-root involvement in lumbosacral spinal stenosis and may be useful in the case where spinal canal narrowing is minimal and the patient has symptoms. This testing also complements standard needle EMG. Snowden et al (19) found that for single-level and multilevel lumbosacral spinal stenosis, DSEPs revealed 78% sensitivity relative to spinal imaging. In this well-designed prospective study, DSEP criteria as well as inclusion criteria were precisely defined. The predictive value was 93%. Yiannikas (20) demonstrated that SEPs may be useful for cervical myelopathy. In a study of patients with clinical signs of myelopathy, all 10 had abnormal peroneal SEPs, and 7 had abnormal median SEPs.

Maertens de Noordhout et al (21) examined motor evoked potentials (MEPs) and SEPs in 55 persons with unequivocal signs and symptoms of cervical spinal cord myelopathy. In this group, 87% showed gait disturbances and 82% showed hyperreflexia. MRI was not the diagnostic standard, as these authors felt that MRI was prone to overdiagnosis; rather, metrizamide myelography showed unequivocal signs of cervical cord compression for all of these patients. Magnetic stimulation of the cortex was performed and the responses were measured with surface electrodes over limb muscles. In these subjects, 89% demonstrated abnormalities in MEP to the first dorsal interosseus muscle, and 93% had one MEP abnormality. At least one SEP abnormality was noted in 73%.

Tavy et al (22) examined whether MEPs or SEPs assisted in identifying persons with radiologic evidence of cervical cord compression but who were without clinical markers for myelopathy. All patients had clinical symptoms of cervical radiculopathy, but not myelopathy. In this group, MEPs were normal in 92% and SEPs were normal in 96%. These investigators concluded that MEPs and SEPs are normal in most cases of persons with asymptomatic cervical stenosis. This indicates that abnormal MEPs and SEPs are likely to be true-positive findings and not false-positives related to mild cord compression that does not cause symptoms. It is important to remember that cervical
spondylosis is a process that causes a continuum of problems, including both radiculopathy and myelopathy.

The inherent variability in SEP recordings and difficulty in determinations as to what constitutes “normal” prompted investigation. Dumitru et al (23) examined the variations in latencies with SEPs. In 29 normal subjects, they examined the ipsilateral intertrial variations, arithmetic mean side-to-side differences, and maximum potential side-to-side differences with stimulation of the superficial peroneal sensory nerve, sural nerve, and the L5 and S1 dermatomes with respect to P1 and N1 latencies and peak-to-peak amplitudes. Considerable ipsilateral intertrial variation was observed, and side-to-side comparisons revealed a further increase in this inherent variation regarding the above-measured parameters. They suggested an additional parameter with which to evaluate SEPs: the maximum side-to-side latency difference.

Dumitru et al (24), in a later study involving persons with unilateral and unilevel L5 and S1 radiculopathies, evaluated dermatomal and segmental SEPs. History, physical examination, imaging studies, and electrodiagnostic medicine evaluations clearly defined patients with isolated L5 or S1 nerve root compromise. Regression equation analysis for cortical P1 latencies evaluating age and height based on comparable patient and control reference populations revealed segmental and dermatomal sensitivities for L5 radiculopathies to be 70% and 50%, respectively, at 90% confidence intervals. Similar sensitivities were obtained for cortical P1 latencies using the mean ± 2 standard deviations (SD) method. Side-to-side cortical P1 latency difference data revealed SEP and DSEP sensitivities for S1 radiculopathies to be 50% and 10%, respectively, also at 2 SD. These investigators questioned the clinical utility of both segmental and dermatomal SEPs in the evaluation of patients with suspected focal L5 or S1 nerve root compromise, finding little utility for these tests in persons with single-level lumbosacral radiculopathy.

In summary, SEPs, DSEPs, and MEPs are not necessary or recommended for the vast majority of patients referred for possible cervical or lumbosacral radiculopathy (7). They can be helpful if performed in special laboratories skilled in such testing, with adequate normal values, and in certain clinical situations that can be clarified by such results.

The need for EMG, particularly in relationship to imaging of the spine, has been recently highlighted (25). Needle EMG is particularly helpful in view of the fact that the false-positive rates for MRI of the lumbar spine are high, with 27% of healthy subjects having a disc protrusion (26). For the cervical spine the false-positive rate for MRI is much lower, with 19% of subjects demonstrating an abnormality, but only 10% showing a herniated or bulging disc (27). Radiculopathies can occur without structural findings on MRI, and likewise without EMG findings. The EMG evaluates only motor axonal loss or motor axon conduction block, and for these reasons a radiculopathy affecting the sensory root will not yield abnormalities by EMG. If the rate of denervation is balanced by reinnervation in the muscle, then spontaneous activity is less likely to occur and be identified with needle EMG.