A prospective, masked 18 month minimum follow-up on neurophysiologic changes in persons with spinal stenosis, low back pain, and no symptoms

Abstract

Objective

To describe neurophysiological changes over time in persons with and without spinal complaints, and to assess whether paraspinal denervation predicts change in stenosis on MRI and clinical course.

Design

Prospective, controlled, masked trial.

Setting

University spine program.

Participants

Persons aged 55–80, screened for polyneuropathy and determined on clinical examination to have spinal stenosis, mechanical low back pain, or no spinal symptoms.

Interventions

Subjects underwent comprehensive codified history and physical examination, ambulation testing, masked electrodiagnostic testing including paraspinal mapping, and MRI; repeated at >18 months. This publication presents detailed technical information and additional analyses not reported previously.

Main Outcome Measurements

Change in electrodiagnostic findings. Among persons with clinical stenosis, relationship of change in paraspinal mapping scores to MRI findings and clinical changes.

Results

Of 149 initial subjects, 83 (79.3% of eligible subjects) repeated testing at 20 +/− 2 s.d.) months. No significant change in limb muscle spontaneous activity or motor unit pathology was noted in any group. In 23 persons with initial diagnosis of stenosis, paraspinal mapping EMG related to change in diagnosis over time (ANOVA F=3.77, p=0.037), but not to most initial MRI measurements or to change in spinal canal diameter.

Conclusions

Clinical spinal stenosis is neurophysiologically stable in most persons. Paraspinal EMG changes reflect large changes in clinical course, but neither neurophysiological nor clinical changes relate to change in spinal geometry over 20 months.

Introduction

Lumbar spinal stenosis is a commonly treated, but poorly understood problem. The hallmark clinical symptom of neurogenic claudication does not occur in all persons thought to have stenosis, and symptoms mimicking neurogenic claudication occur with vascular and even other spinal disorders.^{1, 15, 24} While the terminology suggests an anatomical lesion, anatomical stenosis is so common in persons without symptoms or with mechanical back pain that MRI or other anatomical tests are not definitive.^{5, 26, 31}

For persons with spinal stenosis, a serious consideration is the possible progression of neurological deficit. This fear may drive them and their doctors towards more invasive treatments. Electrodiagnostic testing (EDX) is commonly used to determine the neurological deficit associated with spinal stenosis.^{12, 30, 32, 36, 40, 43} There is a perception that changes in EDX relate to changes in symptoms; however to our knowledge there has never been a case report, let alone prospective study, to validate this belief for any spinal disorder.

In addition to the practical clinical question answered by following electrodiagnostic findings over time, this kind of inquiry can help solve some theoretical questions. We have proposed that denervation of paraspinal muscles can cause segmental hypermobility and loss of kinesthetic sense, leading to more wear on facet joints, joint hypertrophy, and eventually a stenotic spinal canal.²¹

The Michigan Spinal Stenosis Study was designed to answer these questions. It is a masked cohort study of EDX, MRI, and clinical examination which follows persons with clinically defined spinal stenosis, low back pain only, or no symptoms for over 18 months. Analysis of our data on the initial cohort has resulted in a number of publications including the first masked, controlled trial of electrodiagnosis to establish the validity of the test²³ and subsequent analysis demonstrating the relative benefits of electrodiagnosis over MRI scanning in diagnosing spinal stenosis.^{26, 27} More pertinent to the current question, we have shown that function and pain at 18 months are predicted by baseline function and sleep deficits, not MRI or EDX findings.²⁸ While previous reports used summarized EDX results, they did not provide sufficient technical detail for EDX specialists to understand the variation in individual tests. Furthermore, date regarding the relationship of paraspinal denervation with radiological changes over time has not been reported sufficiently to address the principle hypothesis of the research project.

The current paper has two purposes: 1. To determine if specific electrodiagnostic findings change over time in persons with clinical stenosis, low back pain, or asymptomatic volunteers. 2. To seek any relationship between paraspinal denervation or change in paraspinal denervation with change in clinical diagnosis or radiological measures over 18 months.

Methods

Subjects and testing

The study protocol has been described previously.²⁷ In summary, the study sought to recruit persons with no back pain, mechanical low back pain without radiological stenosis, and clinically evident spinal stenosis. Subjects were recruited based on preliminary screening criteria for these 3 categories, but final diagnosis as used in this paper is based on a comprehensive history and physical examination as noted below. All subjects underwent masked MRI or review of MRI, masked electrodiagnostic testing, ambulation testing and numerous questionnaires, as detailed below. All tests were repeated at more than 18 months after the initial testing. Figure 1 outlines these steps. Specific details are below:

Figure 1.

Recruitment and preliminary screening

At a university hospital, serial lumbar MRI reports from the university scanner were screened for persons age 55–80 and for any exclusion criteria (previous surgery, tumor, etc.). Radiologist reports were supplemented with review of all films by a study physician to select persons with ‘preliminary diagnosis of stenosis’. Among those with no apparent stenosis on MRI, further review of the university computerized medical record excluded persons with report of pain radiating below the knee. The subsequent group was labeled ‘preliminary diagnosis of mechanical back pain’. People from the community with no back pain complaint, who had none of the exclusion criteria, but were within the same age range, were recruited via postings and advertisements.

All potential subjects were then screened by phone for exclusion criteria, including known polyneuropathy, diabetes, heavy alcohol use, previous lumbar surgery, or relative contraindications to MRI or EDX. Subjects who had plans for surgery were also excluded due to the project’s long term goals is to follow the natural history of the disorders over 18 months.

Findings from this preliminary process were not revealed to the examining physiatrist who made the final clinical diagnosis. All subjects were given informed consent and were compensated. The university’s ethical review board approved the study.

Clinical evaluation

All subjects filled out an extensive patient questionnaire, including the Pain Disability Index, the Quebec Back Pain Disability Index, and the McGill pain scale, a visual analog pain scale, and a pain drawing, along with a 5 page clinical spine questionnaire that encompassed medical history, review of systems, family history, and social history.^{4, 37, 38, 39, 44} Each performed an ambulation test in which they were instructed to walk at a comfortable speed for 15 minutes, and wore a pedometer during waking hours at home for a week.

Physiatrists reviewed the questionnaires and performed a comprehensive and codified spinal history and physical examination. Four of the physiatrists were board certified in all three specialties of Physical Medicine and Rehabilitation, Pain Medicine, and Electrodiagnostic Medicine, and five others were in a clinical fellowship designed to qualify them for all three boards. Their primary residency training was at 8 different universities, suggesting a diverse background. The physiatrist’s impression as to whether the subject had low back pain, had spinal stenosis, or was asymptomatic is termed the “clinical diagnosis” throughout this paper.

The diagnostic standard for the current study was the conclusion of the examining physician, who was not restricted by any a-priori criteria. To establish face validity of the clinical diagnosis, we examined a number of potential associations with findings often thought to relate to spinal stenosis. Portions of the clinical examination that are thought to be consistent with the clinical syndrome of spinal stenosis were found to relate to the clinician’s diagnosis. These included pain below the knee, pain severity, difficulty with walking, 15 minute ambulation velocity, spine tenderness, strength deficits, reflex deficits, straight leg raise test and femoral stretch test; was performed to demonstrate face validity of the clinical diagnosis. Also a senior academic spine surgeon who was masked from the radiologist’s and physiatrist’s diagnosis reviewed the MRI films, history and physical examination data, and patient questionnaire to independently arrive at a neurosurgical diagnosis. This was compared to the clinical diagnosis. Both the spine surgeon’s impression and the examination details supported the reasonableness of the clinical diagnosis.²⁶

MRI scanning

The asymptomatic volunteers underwent lumbar MRI scans, and MRI scans of the others, performed within 6 months of the study, were all reviewed. All scans were non-contrast lumbosacral spinal MRI performed on a GE Signa Horizon LX (GE Medical Systems, Milwaukee, Wisconsin). They including sagittal T2-weighted (FOV: 30; scan thickness (ST): 3.0mm, interscan spacing (IS): 0.5mm, Matrix 384 X 192; TR: 3000 AND TE: 102 Pulse: FSE), sagittal T1-weighted (FOV: 30; ST: 3mm, IS: 0.5mm, Matrix: 256 × 192, TR: 400–700, TE: MIN FULL Pulse: SE) and axial T2-weighted scans (FOV: 20, ST: 4mm, IS 5mm, 5 slices through each disc space T12-L1 through L5-S1; Matrix: 256× 256, Tr: 3000–5000, TE: 102 Pulse: FSE).

All images were reviewed by a masked neuroradiologist at a workstation (Windows Advantage Workstation, General Electric Medical Systems, Milwaukee, Wisconsin). Anatomic measurements were made using an electronic cursor at each lumbar intervertebral disc level. Measurements included: midline anterior-posterior osseous spinal canal diameter, midline anterior-posterior thecal sac diameter, osseous spinal canal cross-sectional area, thecal sac cross-sectional area, osseous interfacet distance (measured between the medial osseous margins of each facet joint), distance between the medial joint capsular margins of each facet joint, and the anterior-posterior lateral recess diameter on the right and left side. Previous work suggested symptoms occur because vascular compromise occurs when the vasa nervorum are compressed at two levels.^{20, 42} Therefore composite scores of the average of the smallest two canal diameters, smallest two thecal sac diameters, etc. were developed. Test-retest validity of the radiologists was excellent, as documented elsewhere.⁴⁵

Electrodiagnostic testing

A masked EDX specialist performed a detailed electrodiagnostic study. The adequacy of masking is described elsewhere, but in summary, at most, unmasking could have affected 6% or fewer of the examinations.²⁵ Electrodiagnostics were performed with a Nicolet Viking II (Nicolet Biomedical, Madison Wisconsin) using a 50- or 75- mm monopolar needle. Skin temperature was monitored and heat was applied when necessary to keep skin temperature above 32 degrees centigrade.

As recommended by Dillingham et al., the testing included exploration of 5 muscles with overlapping root innervation-- the tensor fascia lata (L4, L5, and S1 innervated), vastus medialis (L2, L3, and L4), tibialis anterior (L4 and L5), extensor hallucis longus (L5 and S1), and the medial gastrocnemius (S1 and S2) on either the most symptomatic side or if symptoms were absent or symmetrical, a leg chosen by the assistant by coin toss. ⁸ In each muscle the presence of fibrillation potentials (a sign of denervation) was scored after 6 insertions in 4 different directions as 0–4+ using Daube’s definitions.⁶ Ten motor units were sampled in each muscle. Because it is not the usual clinical practice to actually capture and measure each motor unit, examiners were asked to make informal estimates of typical motor unit amplitude, number of polyphasic motor units, and motor unit recruitment (firing rate of the first motor unit when a second motor unit was recruited). The number of polyphasic motor units seen per muscle was recoded as none, 1–2/10 polyphasic motor units (“poly’s”), and >2/10 poly’s. Sural sensory response and peroneal motor responses from the ankle, fibular head, and popliteal space were performed. Bilateral H-waves and peroneal F-waves were performed by a technician and interpreted by the electromyodiagnostician.

This study used the MiniPM abbreviated version of the original Paraspinal Mapping (PM) to study the paraspinal muscles bilaterally.^{16, 17, 18, 22} Convention has lead others to use the term ‘Paraspinal Mapping’ for this abbreviated version, so we will use that term throughout this document. Paraspinal Mapping is an anatomically validated, quantified scoring system for the paraspinal muscles. The range of normals for PM has been defined, good inter-rater reliability has been established (r=.830, p=.041), and limited clinical evidence suggests that it can localize the root level of a lesion.^{19, 45, 46}

The PM technique is described in detail elsewhere.^{18, 22} Briefly it includes palpation of the inferior border of the three lowest lumbar spinous processes and the midpoint between the posterior superior iliac spines, measuring 2.5 cm laterally, and (for the L3, 4, and 5 spinous processes) 1 cm cranial. From each of these 4 locations, a 50 to 75 mm monopolar EMG needle is inserted at a 45–60 degree angle to the surface in three different directions—cranial 45 degrees, directly across to the spinous process, and caudal 45 degrees—and advanced through the muscle in 5mm movements to detect abnormal muscle membrane instability. Any fibrillations must last more than 1 second and be reproducible. Scores for the medial most 1 cm are scored separately from the more lateral components of the insertion. Depending on the severity of findings, scores ranged from 0–4+ in any of 24 total locations.

A total score for the side (number of +’s) is determined, resulting in a potential range of 0–96, but 95% of asymptomatic younger persons score ≤ 2. A cutoff of >4 for abnormal on the PM score was established prior to this analysis based on data from the entire asymptomatic population presented elsewhere, which shows that the range of normal in older asymptomatic persons is higher than the 95% cutoff of >2 that we had earlier reported in younger persons.^{18, 46}. PM was performed bilaterally, but the current paper used only the data for the side on which limb electrodiagnostic testing was performed.

Follow-up and disqualifications

Subjects were invited back for retesting at 18 months after initial testing. This time frame was chosen primarily from a pragmatic standpoint. We did not think that the federal funding source would support a longer term study. Those who had a newly diagnosed exclusion criteria, including surgery, were eliminated. Initial testing occurred late enough in the funding cycle that a number of subjects were not included in the follow-up cohort. A few others declined participation in the 18 month follow up.

At more than a year after testing, a quality check on electrodiagnostic data on all subjects, in conjunction with review of available medical records when needed, determined if there was evidence of polyneuropathy or myopathy which would have excluded them from recruitment, if it had been known. Neuromuscular disorders were determined by characteristic needle examination findings not seen in radiculopathy (e.g. brief, short, early recruited polyphasic motor units in proximal muscles for myopathy) and nerve conduction findings (e.g. sural nerve or peroneal nerve slowing for polyneuropathy). These subjects were removed from further analysis.

A few MRI scans were not completed or had technical errors. All but one subject completed the electrodiagnostic testing, but persons who had poor relaxation at two or more of the four levels explored on either side of the paraspinal mapping grid were eliminated. A total of 15 subjects were eliminated from analysis because of these testing issues. Three subjects had some unreliable paraspinal EMG data at only one spinal level. These levels were assumed to be normal.

Statistical Methods

Data was initially entered in a Microsoft Excel database where errors were checked and cleaned. SPSS Version 12.0 was used for statistical analysis. Significance was accepted at p<.05.

Results

Subjects

Figure 1 shows the process of subject recruitment, selection, and follow-up. After initial testing certain subjects were disqualified, including 9 due to spinal surgery, 1 deceased, 12 new contraindications (anticoagulation, cancer, diabetes, neuropathy), and 2 who moved out of state. Seven did not respond to requests for testing and 16 were not interested in repeat testing. Unfortunately study funding ran out before 20 other subjects became eligible. Eighty-three persons (79% of subjects who were eligible for repeated testing), including 32 persons with clinically defined stenosis, underwent follow-up testing. Average time to follow-up was 20.0 ± 2.0 months (range 18.3 – 29.2 months).

Compared to the subjects who did not complete follow up testing, when stratified by clinical diagnosis the final cohort was in general not any different in terms of EMG findings—limb fibrillations, limb polyphasic motor units, and PM scores. Two exceptions include the Extensor Hallicus Longus in the asymptomatic group, where a significantly higher percentage of dropouts (58.3%) had >2/10 polyphasic motor units (58% vs. 13%, Chi-square=7.926, p = 0.005) and ‘any fibrillations in the whole subject’ among the stenosis group, which approached significance (dropouts 44.8% vs. follow-up subjects 21.9%, Chi-square=3.637, p=0.057).

Limb EMG findings

Abnormal spontaneous activity is considered a fairly concrete finding of pathology. Changes in limb muscle spontaneous activity in the three diagnostic groups over time are reported in Table 1. A large percentage of the subjects across all 3 groups, ranging from 67.9% to 100.0%, had unchanged scores in all muscles. None of the asymptomatic subjects had abnormal spontaneous activity on initial evaluation, but at follow-up 1 had 2+ fibrillations in the tibialis anterior and another had 3+ positive waves in the extensor hallucis longus. Among persons with low back pain but no stenosis, twenty-two (78.6%) were initially completely normal. The vastus medialis remained normal in all subjects, but each of the other muscles was abnormal in about 10% of instances. As a majority (> 50%) of the cells have expected count less than 5, Chi-square analyses were not performed to assess the differences in the distribution (0 – 4+ according to Daube’s definitions) of the initial vs. follow-up scores.⁶ There was no significant increase in spontaneous activity in any muscle or in the subjects as a whole over time. Although a few individuals changed in severity (0–4 rating), most were unchanged. The limb exam for fibrillations was abnormal in 7 (21.9%), but essentially unchanged over time, with an equal number of subjects (15.6%) worsening or improving somewhat electrophysiologically.

Table 1.

Spontaneous activity changes in the limb muscles in persons with no symptoms, low back pain, and clinical spinal stenosis. For each diagnosis and muscle, the number of subjects abnormal on initial testing is indicated in the first row. The number of muscles with an increase in spontaneous activity over 18 months is in the next row, and the number of muscles with a decrease in spontaneous activity is in the third row. Increase and decrease are based on any change in the 0–4+ scoring of spontaneous activity. Thus a person whose spontaneous activity changed from 1+ to 2+ in a muscle would have an increase, while a person whose spontaneous activity changed from 2+ to normal would have a decrease.

	Tensor Fascia Lata	Vastis Medialis	Tibialis Anterior	Extensor Hallucis Longus	Medial Gastroc nemus	Whole subject
Asymptomatic N=23
Abnormal	0	0	0	0	0	0 (0%)
Increased	0	0	1	1	0	2 (8.7%)
Decreased	0	0	0	0	0	0 (0%)
Low back pain N=28
Abnormal	3	0	2	3	2	6 (21.4%)
Increased	2	0	1	2	3	6 (21.4%)
Decreased	2	0	1	0	2	3 (10.7%)
Clinical spinal stenosis N=32
Abnormal	4	2	2	3	4	7 (21.9%)
Increased	0	0	2	2	4	5 (15.6%)
Decreased	4	2	2	3	4	5 (15.6%)

Motor unit configuration changes over time are presented in Table 2. Among the asymptomatic group, 26.1% had >2/10 polyphasic motor units. The tensor fascia lata and the extensor hallucis longus, which share L-5 innervation, both had an abnormal proportion of polyphasic motor units in over 10% of asymptomatic subjects. Almost 40% of persons thought to have mechanical low back pain had >2/10 polyphasic motor units, again with the two L5 muscles most involved. A similar number (43.8%) of stenosis subjects had abnormal polyphasics with a similar distribution. The majority of all three groups increased or decreased somewhat in the extent of polyphasicity, but there was no trend towards worsening or improving.

Table 2.

Motor unit morphology changes in persons with no symptoms, low back pain, and clinical spinal stenosis. For each muscle and diagnosis the number of subjects with more than 2/10 polyphasic motor units and the percentage of motor units that were polyphasic are noted at initial assessment and at 18 month follow-up. The number of muscles with an increase or decrease in polyphasia over 18 months is also noted.

	Tensor Fascia Lata	Vastis Medialis	Tibialis Anterior	Extensor Hallicus Longus	Medial Gastrocnemus	Whole subject
Asymptomatic N=23
>2/10 polyphasic Initial	3	1	1	3	0	6 (26.1%)
Follow-up	4	1	4	3	0	7 (30.4%)
Polyphasic motor units Initial (s.d.)	0.86 (1.13)	0.43 (0.95)	0.65 (0.78)	1.26 (1.14)	0.35 (0.57)	1.74 (1.25)
Follow-up (s.d.)	0.55 (2.74)	0.26 (0.69)	0.78 (1.54)	0.87 (1.22)	0.35 (0.71)	1.43 (1.78)
Increase in %polyphasic	6	3	5	4	4	6 (26.1%)
Decrease in % polyphasic	8	5	10	11	6	11 (47.8%)
Low Back Pain N=28
>2/10 polyphasicInitial	5	1	3	7	0	11 (39.3%)
Follow-up	5	3	5	8	1	10 (35.7%)
Polyphasic motor units Initial (s.d.)	1.07 (1.49)	0.36 (0.78)	1.00 (1.39)	1.33 (1.59)	0.21 (0.50)	2.07 (1.51)
Follow-up (s.d.)	0.89 (1.58)	0.50 (1.11)	0.79 (1.57)	1.63 (2.12)	0.36 (0.78)	1.82 (2.09)
Increase in % polyphasic	6	4	4	7	5	7 (25.0%)
Decrease in % polyphasic	9	4	11	7	4	14 (50.0%)
Clinical Spinal Stenosis N=32
>2/10 polyphasic Initial	4	2	5	10	4	14 (43.8%)
Follow-up	5	2	6	8	3	13 (40.6%)
Polyphasic motor units Initial (s.d.)	1.16 (1.22)	0.81 (1.06)	1.22 (1.34)	1.88 (1.72)	0.94 (1.16)	2.25 (1.67)
Follow-up (s.d.)	0.72 (1.11)	0.31 (0.82)	0.94 (1.44)	1.31 (1.64)	0.56 (1.37)	1.94 (1.80)
Increase in % polyphasic	8	3	7	9	6	12 (37.5%)
Decrease in % polyphasic	12	13	11	14	14	14 (43.8%)

There was a significant change in the mean number of polyphasic motor units in the vastus medialis, but none of the other muscles. Chi-square analyses were not performed to assess the differences in the distributions of the initial vs. follow-up scores broken down by three groups (0/10 polyphasic, 1–2/10 polyphasic, and >2/10 polyphasic), as a majority of the cells have expected count less than 5.

Paraspinal denervation

As expected the PM scores of patients with spinal stenosis averaged higher than those with back pain and these were higher than the asymptomatic group. (Table 3). There were no statistically significant differences in PM score among subjects who underwent follow-up and those who did not. (independent samples t-test p>.20 all cases). The paired-samples t-test did not find a significant mean difference between the initial PM score and the follow-up PM score. Non-significant Kappa results indicate lack of agreement between two categorical measures of normal vs. abnormal PM score and initial vs. follow-up clinical finding. There was substantial variability (though generally within our ‘normal’ range) in paraspinal mapping scores from initial to follow-up in the asymptomatic population.

Table 3.

Initial and follow-up paraspinal mapping (PM) scores in the stenosis, back pain, and asymptomatic groups. There was no significant change in the PM scores over time (paired t-test <1.3, p>.2 all cases,) or in the category of normal vs. abnormal over time, although there was a trend for the stenosis subjects to have similar scores at follow-up.

	No symptoms	Back pain	Stenosis
Number of subjects followed	22	24	29
PM score Initial (X, s.d.)	1.68 (2.61)	2.83 (5.21)	3.93 (5.11)
PM score Follow-up	1.36 (2.06)	4.42 (6.57)	3.62 (3.84)
Initial vs. follow up t-value (paired-samples test)	−0.471	1.278	− 0.299
Certainty (p-value)	0.642	0.214	0.767
PM Abnormal (>4) initial (N, %)	3 (13.6%)	4 (16.7%)	11 (37.9%)
PM Abnormal (>4) follow-up (N, %)	2 (9.1%)	7 (29.2%)	10 (34.5%)
Initial vs. follow-up chi-square (Kappa statistics)	− 0.122	− 0.038	0.329
Certainty (p-value)	0.556	0.841	0.076

The ability of initial PM score to predict change in clinical diagnosis is reported in Table 4. The Chi-square test shows a significant agreement between the physiatrist initial diagnosis and final diagnosis (Kappa = 0.396, p < 0.001). We did not find significant mean differences with respect to change in diagnosis (for example, asymptomatic = asymptomatic or asymptomatic→ back pain or asymptomatic→ stenosis) within each respective population. The very high scores in the 2 stenosis subjects who were later asymptomatic, and the higher scores in the two asymptomatic subjects who went on to present as stenosis are of interest. Other disorders which denervate the paraspinal muscles can mimic spinal stenosis.

Table 4.

The relationship between initial PM scores and change in clinician diagnosis over 18 months.

	Initial diagnosis*
	Asymptomatic		Back Pain		Stenosis
Final		PM1 X (s.d.)		PM2		PM3
Asymptomatic	15	1.87 (2.67)	1	3.00 (−)	2	10.50 (0.71)
Back Pain	5	0.00 (0.00)	14	2.21 (3.89)	11	4.09 (6.27)
Stenosis	2	4.50 (3.54)	9	3.78 (7.19)	16	3.00 (3.97)
Total	22	1.68 (2.61)	24	2.83 (5.21)	29	3.93 (5.11)

ANOVA

¹F = 2.585, p = 0.102

²F = 0.231, p = 0.796

³F = 2.075, p = 0.146

Not shown here, among the persons who had stenosis, there was also no significant relationship between PM scores and numerous clinical factors, including the clinical severity (mild, moderate, or severe), ambulation velocity, Pain Disability Index, Visual Analog Pain Scale, the Quebec Back Pain Disability Scale, the McGill, or the presence of pain below the knee on pain drawing.

The relationship between the numerous radiological measures and PM scores was explored among the persons with clinical spinal stenosis, as shown in Table 5. There were significant relationships between PM score and measures of the thecal sac at L4–5 (antero-posterior dimension, r=−0.298, p=.038, area of the sac r=−0.295, p=.040). The measure of the ‘average of the two smallest anteroposterior sac measures’ trended towards significance (r=− 0.261, p=.070). Also the area of the spinal canal at L1–2 trended towards a significant relationship with PM (r=− 0.257, p=.081). All other measures, including anterior-posterior measures of, and area of the spinal canal at each of the 5 levels, smallest sac and canal measurements, and averages of the two smallest sac and canal measures, a total of 24 of the 28 measures on the symptomatic side, showed no trend. There were no significant relationships between PM score on the asymptomatic side regarding any of these findings.

Table 5.

The relationship between change in diagnosis over 18 months and changes in test results (MRI antero-posterior spinal canal diameter (AP canal) and paraspinal mapping scores).

	Initial diagnosis
	Asymptomatic			Back Pain			Stenosis
		PM1 X (sd)	AP canal2		PM3	AP canal4		PM5	AP canal6
Asymptomatic	15	−0.20 (3.36)	−0.19 (1.87)	1	0.00 (−)	1.70 (−)	2	−9.50 (2.12)	2.65 (3.75)
Back Pain	5	0.60 (0.89)	−0.62 (2.50)	14	2.93 (7.47)	− 0.03 (2.80)	11	− 0.55 (6.19)	− 0.21 (2.31)
Stenosis	2	−3.50 (4.95)	− 0.60 (0.28)	9	− 0.33 (2.78)	0.17 (0.66)	16	1.00 (4.41)	0.69 (1.81)
Total	22	−0.32 (3.17	−0.33 (1.89	24	1.58 (6.07)	0.11 (2.26)	29	− 0.31 (5.59)	0.51 (2.16)

ANOVA

¹F = 1.259, p = 0.306

²F = 0.109, p = 0.898

³F = 0.813, p = 0457

⁴F = 0.257, p = 0.776

⁵F = 3.770, p = 0.037 (significant)

⁶F = 1.676, p = 0207

For persons with spinal stenosis the change in diagnostic category did relate significantly to the change in PM score (ANOVA, F = 3.770, p = 0.037), but not to change in spinal canal diameter. (see table 6). The post hoc analysis using Tukey’s honest significant difference test in the stenosis subjects identified a significant mean difference between the stenosis → asymptomatic and stenosis = stenosis groups (X = −9.50 (2.12) vs. X = 1.00 (4.41), p = 0.029). No significant relationship between change in PM or MRI measures and change in clinical diagnosis was found for persons with back pain or no symptoms.

Discussion

This is the first study to systematically look at a needle EMG diagnostic protocol for lumbar radiculopathy over time in an asymptomatic population or in a group with spinal disorders. The findings suggest that denervation as tested by EMG remains generally stable over 18 months. There are a number of clinical and scientific implications.

The study methodology has much to offer. This is also the first adequately masked study of needle EMG and the first to prospectively evaluate spinal stenosis alongside two different control groups—asymptomatic persons and those with a more clinically relevant alternative presentation of mechanical back pain. The codified, quantified procedures and the diversity of testing physicians suggest that the findings are both reproducible and representative of typical practice.

Some limitations exist, however. It is legitimate to question the ability of needle EMG to measure change over time. Neither the 0–4+ scale for abnormal insertional activity proposed by Daube nor the informal motor unit analysis used here (and in most clinical settings) have been subjected to inter- or intra- rater reliability studies. With repair and regeneration, it is possible that the primary manifestation of denervation over time is motor unit amplitude or duration changes, which were not measured precisely in this study. Our study points out the need to demonstrate the reliability of these core measures in order to validate the use of electrodiagnostic testing in general.

We circumvented some of these issues by using paraspinal mapping. Tong has performed inter-rater reliability testing of PM showing good reproducibility (r=.830, p=.041) with monopolar needles. ⁴⁵ No valid method yet exists to quantify motor unit abnormalities in the paraspinal muscles. Since paraspinal muscle denervation was highly specific in our earlier analyses ²³, it is possible that the consequences of denervation—motor unit changes—would show a difference if we could only reproducibly measure these changes.

Because the purpose of the study was to follow changes over 18 month, we excluded people who were planning on surgery. Thus the population thus does not include many persons with severe disease or disability.

The time between examinations was relatively short in the context of a lifetime. It is always possible that more obvious changes would occur over longer periods of time. The majority of subjects who were not tested at 18 months dropped due to budgetary or scientific issues, not a decision by themselves. Although the 9 subjects who underwent surgery did have more severe disease as measured by EMG and MRI (subgroup analysis not presented here), no statistically important bias was found between the dropouts as a group and the study completers. The dropout of 20 subjects due to funding did not likely bias the results, but may have resulted in underpowering of the study to detect a change. At a minimum the final population represents an important subgroup of persons with spinal stenosis, but we believe that it is representative of the natural history of the disease in general.

In Tong’s population of primarily younger persons including disk herniation and polyneuropathy, a change of +7 or − 9 was beyond the 95% confidence interval for repeatability.⁴⁵ A tighter standard is likely more appropriate in the population presented here, who tended to have lower scores.

Signs of denervation in asymptomatic persons

Aside from the foot, it is thought that limb fibrillations are not commonly found in asymptomatic persons.^{9, 14} This was generally true in our initial cohort, but it is interesting to note that two went on to have some fibrillations. It is accepted that a certain percentage of motor units in limb muscles are polyphasic.^{10, 35} Twenty six percent of our asymptomatic group had more than 2/10 polyphasic motor units in a muscle—primarily the extensor hallicus longus and tensor fascia lata. These two muscles share the L5 nerve root, which is the most common root involved in radiculopathy. This pattern suggests that many asymptomatic older persons may have had an L5 radiculopathy (symptomatic or asymptomatic) in the past. Alternatively an examiner bias due to expectation could be implicated. Regardless, since these people are without symptoms, clinicians are cautioned that the presence of polyphasic motor units in an L5 pattern does not necessarily mean that there is a clinically relevant disease

It is now well established that asymptomatic persons—young and old--have ‘abnormal’ spontaneous activity in the paraspinal muscles.^{7, 9, 17} The current study confirms this. Three (13%) subjects had initial denervation sufficient to fall outside of the norms we had established a priori. The data suggest that a cutoff PM score of 6 is would more likely fit the traditional need for a 95% confidence interval. On the other hand, as reported elsewhere, the subgroup in which the physiatrist, radiologist, and spine surgeon all agreed independently were asymptomatic had PM scores of 0.67 +/− 1.07 (s.d.), which compare well with the norms established for younger persons.^{18, 46}

It has been hypothesized that the spontaneous paraspinal denervation that occurs in asymptomatic persons results in transient weakness and kinesthetic deficits, which in turn destabilize the spine, causing more ligamentous laxity and degenerative changes in the facet joints—perhaps leading to spinal stenosis, and also putting the posterior primary ramus at further risk. ²¹ No significant change was detected in the PM scores over time, and PM score did not relate to changes in the spinal measurements we made. This study does not dismiss these theories. First, the radiological measures did not include measures of facet joint hypertrophy or the nerve foramen. Second, the rate of change in bony structures may not allow us to measure a difference in 18 months.

Since the multifidus muscle is only a few centimeters from the spinal canal, one would expect that individuals with posterior primary ramus damage would have regrowth and repair resulting in a decrease in spontaneous activity over 18 months. The lack of a positive or negative trend in PM scores suggests an ongoing process of denervation and reinnervation, rather than a single injury in time. We can only speculate as to whether the substantial changes in a few individuals were related to an event or a symptom.

Is mechanical back pain a neurological disorder?

“Mechanical” back pain without leg pain is generally thought to spare the nerve roots. But this study shows evidence of denervation outside of the range of normal established in asymptomatic populations in a large minority of our subjects with back pain. Limb fibrillations were found in 21.4%, 39.3% had >2/10 polyphasic motor units and 16.7% had PM scores >4. In fact there is scant literature on needle EMG findings in persons with back pain alone, and past research has been hampered by the lack of masking and the lack of established norms.¹¹

These findings suggest that some people thought clinically to have mechanical back pain actually have radicular involvement. However the relationships between paraspinal denervation, radiculopathy, back pain, and spinal stenosis is not straightforward. In addition to radiculopathy, myopathies, polyneuropathies, and a difficulty-to-prove disorder involving entrapment of the posterior primary ramus¹³ are among the many causes of isolated paraspinal denervation. Since 8 persons in the initial cohort of 150 were found to have unsuspected neuromuscular diseases, and the EDX protocol here did not exhaustively examine for neuromuscular disease, an alternative explanation is the presence of a coincidentally occurring sub-clinical polyneuropathy or myopathy. Clinically we have noted that the diffuse neuromuscular diseases have diffuse paraspinal denervation, but this remains to be proven.

Even in the absence of diffuse neuromuscular disease, the presence of denervation does not always imply pain. Even when the radiculopathy is from spinal stenosis, the pain and disability that a patient complains of may come from some other source, such as the facet joint. Research from this study presented elsewhere shows a lack of any good relationship between the clinical syndrome recognized as spinal stenosis and radiological measures of spinal canal size.^{26, 27} At this point we do not believe there is a ‘gold standard’ for the diagnosis of the clinical syndrome commonly called spinal stenosis.

On the other hand the clinical syndrome did relate to PM scores, which were 100% specific, though only moderately sensitive. Thus we have proposed that the term ‘spinal stenosis’ be reserved for anatomists until such time as a clinically relevant anatomical definition can be found.²⁷ However this proposal challenges a century of medical tradition and textbooks, as well as the billing codes that justify surgical intervention.

The neurophysiological natural history of the paraspinal denervation syndrome (a.k.a. clinical spinal stenosis)

As expected the people with clinical evidence for spinal stenosis (a term we will use reluctantly for the purpose of clarity) had more denervation than the others—21.9% had limb fibrillations, 43.8% had >2 polyphasic motor units in one or more muscles, and 37.9% had paraspinal denervation outside of the range of normal. There was no significant progression of disease in this group over time. Instead, some subjects improved and some declined neurophysiologically. In fact, prior to this work there has never to our knowledge ever been a single case report that prospectively follows electrodiagnostic findings in a person with any spinal disorder over time. We can question the dogma regarding other spinal disorders, as well.

Clinical implications

Previous work from the Michigan Spinal Stenosis Study has shown MRI measures and the radiologist’s clinical impression do not differentiate persons with symptoms of clinical stenosis from age-matched people who have no symptoms whatsoever. The electrodiagnostic examination is highly specific and moderately sensitive for the disorder, but only if paraspinal mapping is used and norms established in this study are used. We have shown elsewhere that the EMG is not predictive of outcome. The current paper provides specific details of that conclusion, especially the relationships between paraspinal denervation and outcome. Previous limited information available on natural history also suggests that some subjects get better or worse over time clinically, with a general trend towards improvement.^{2, 3, 19, 30, 33, 34, 41, 42}

This slow or absent progression means that there need not be any rush to surgical intervention. Delays related to numerous conservative treatments, functional rehabilitation, and lifestyle modifications are not going to put most patients at neurological risk. Electrodiagnosticians and clinicians who interpret their findings should be slow to use a positive EMG as evidence that the patient’s problem is progressing. Instead the factors that help a clinician best decide whether leave symptoms alone, treat conservatively, or invoke surgical intervention remain poorly understood.

Conclusions

There is no significant trend towards neurological worsening or in persons without symptoms, with back pain, or with clinical spinal stenosis over 18 months. The presence of paraspinal denervation does not predict clinical, radiological, or electrdiagnostical decline in people with spinal stenosis over this period of time.