Spinal Stenosis, Back Pain, or No Symptoms at All? A Masked Study Comparing Radiologic and Electrodiagnostic Diagnoses to the Clinical Impression

Participants and Testing 

Table 1 summarizes the subject recruitment and testing protocol. We screened serial lumbar MRI reports from the university scanner for persons age 55 to 80 years and for any exclusion criteria (eg, previous surgery, tumor). 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 no stenosis.” We then screened all potential subjects by telephone for exclusion criteria, including known polyneuropathy, diabetes, heavy alcohol use, previous lumbar surgery, or relative contraindications to MRI or electrodiagnostic testing. Subjects who had plans for surgery were also excluded, as one of the project’s long-term goals is to follow symptoms and signs over 18 months. We intended to recruit 90 persons with preliminary radiologic diagnosis of spinal stenosis regardless of symptoms and 30 persons who had a preliminary diagnosis of nonradiating back pain with no stenosis. Findings from this preliminary process were not revealed to the examining physiatrist who made the final clinical diagnosis.

Table 1. Subject Recruitment and Groupings

	Recruitment Stage	Asymptomatic	Back Pain	Stenosis	Total
	Telephone contact	80	100‖	142	322
	Tested	34	32	84	150
	Clinically disqualified⁎	3	0	6	9
	Missing key data†	1	2	12	15
	Preliminary groupings‡	30	30	66	126
	Clinical impression§	32	44	50	126

⁎ Includes patients subsequently found to have neuromuscular disease or cancer. † Includes absent or uninterpretable MRI (n=4), or unreliable paraspinal electromyographic data at 2 or more levels on any side (n=10), and 1 subject (not in the table) who dropped out before clinical status was established. ‡ Preliminary groupings are based on prerecruitment review of MRI and pain complaints. § The physiatrist’s impression after comprehensive spinal history, physical examination, and patient questionnaire. ‖ Number of potential subjects with back pain was not exactly recorded, but is estimated to be 100.

In addition, we recruited 30 people from the community with no back pain complaint, who had none of the exclusion criteria, but were within the same age range, via postings and advertisements. These asymptomatic volunteers underwent lumbar MRIs.

All subjects gave informed consent and were compensated. The university’s ethical review board approved the study.

Subjects filled out a 5-page clinical questionnaire, the Pain Disability Index,26 the Quebec Back Pain Disability Scale,27 the McGill Pain Questionnaire,28 a visual analog scale (VAS) for pain,29 and a pain drawing.30 Each performed a 15-minute ambulation velocity test and wore a pedometer during waking hours at home for a week.

The final “clinical impression” was made by 1 of 9 physiatrists who performed a comprehensive and codified, but unconstrained, spinal history and physical examination after reviewing the questionnaires and the 15-minute ambulation test results. (The 1-week ambulation test began on this day, so results were not available.) Four of the physiatrists were board certified in physical medicine and rehabilitation, pain medicine, and electrodiagnostic medicine, and 5 were in a 1-year clinical fellowship designed to qualify them for these 3 boards. Their primary residency training was at 8 different universities, suggesting a diverse background. The physiatrist recorded a clinical impression of LBP, spinal stenosis, or asymptomatic volunteer, along with any confounding factors such as polyarthritis or peripheral vascular disease. Severity of stenosis was subjectively rated as mild, moderate, or severe.

Electrodiagnostics 

Electrodiagnostics were performed with a Nicolet Viking IIa using a 50- to 75-mm monopolar needle. Skin temperature was monitored and corrected as needed with hydrocollator packs to maintain a temperature over 32°C. As recommended by Dillingham et al,31 5 muscles with overlapping root innervation were tested on the most symptomatic side or if symptoms were absent or symmetrical, a side chosen by the assistant via coin toss. These muscles were the tensor fascia lata, vastus medialis, tibialis anterior, extensor hallucis longus, and the medial gastrocnemius. In each muscle, after 6 insertions in 4 directions, spontaneous activity was scored using Daube’s definitions (range, 0−4+).32 Ten motor units were observed in each muscle and estimates of typical motor unit amplitude, number of polyphasic motor units, and motor unit recruitment were recorded. Sural sensory and peroneal motor nerve conduction studies, bilateral H-wave and peroneal F-wave tests were performed.

We tested the paraspinal muscles using the MiniPM technique, an abbreviation of the original paraspinal mapping technique, bilaterally.33, 34 This technique is described in detail elsewhere.33 It includes palpation of the inferior border of the 3 lowest spinous processes and the midpoint between the posterior superior iliac spines, measuring 2.5cm laterally, and (for the L3, L4, and L5 spinous processes) 1cm cranial. From each of these 4 locations, a monopolar electromyography needle is inserted at a 45° to 60° angle. It is advanced in 5-mm increments in 3 different directions (mediocranial, 45°; medial; mediocaudal, 45°). Instability (positive waves or fibrillations) is not scored unless it lasts more than 1 second and is reproducible on up to 2 repeated passes through the area. The most medial 1cm, anatomically specific to a particular root innervation, is scored separately from the more lateral mixed components of the insertion. Scores range from 0 to 4+ in any of 24 total locations. A total score is determined by summing 0 to 4+ scores noted in each of 24 locations, resulting in a potential score range of 0 to 96. The range of normal, interrater and test-retest reliability have been established, and limited evidence suggests that MiniPM localizes the root level of a lesion.33, 35, 36

The techniques for masking the electrodiagnostician and the adequacy of masking in this study have been described in detail elsewhere.37 At most, 6% of subjects might have been compromised by unmasking, but the actual influence of potential unmasking was likely much less. The electrodiagnostician was asked for an impression of the presence or absence of spinal stenosis, and for the severity of the lesion. This impression was not required to relate to any particular piece of data or set of norms, but was based on the electrodiagnostician’s judgment.

Magnetic Resonance Imaging 

Subjects underwent a noncontrast lumbosacral spinal MRI scan (Signa Horizon LX)b including sagittal T2-weighted (field of view [FOV]: 30cm; scan thickness [ST], 3.0mm; interscan spacing [IS], 0.5mm; matrix, 384×192; repeat time [TR], 3000ms; echo time [TE], 102ms; pulse: fast spin echo [FSE]), sagittal T1-weighted (FOV, 30cm; ST, 3mm; IS, 0.5mm; matrix, 256×192; TR, 400−700ms; TE, minimum full; pulse, spin echo), and axial T2-weighted scans (FOV, 20cm; ST, 4mm; IS, 5mm; 5 slices through each disk space T12-L1 through L5-S1; matrix, 256×256; TR, 3000−5000ms; TE, 102ms; pulse: FSE). Images were reviewed at a workstation (Windows Advantage Workstation).b

A masked neuroradiologist reviewed the lumbosacral images and rated the degree of spinal stenosis at each lumbar vertebral interspace. Any anatomic abnormalities contributing to the stenosis, including articular facet degenerative changes, ligamenta flava thickening, anterolisthesis, retrolisthesis, disk bulging, disk protrusion and extrusion, or mass lesion was noted. The neuroradiologist made an overall impression of the presence or absence of stenosis in a way similar to that of Boden14 and Jensen15 and colleagues. In addition, subjective overall severity of the stenosis was recorded as mild, moderate, or severe.

The construct validity of the clinician’s diagnosis was assessed by comparing it to components of the history, physical examination, and ambulation tests that might be associated with spinal stenosis. Also, a senior academic neurosurgeon was asked to review the MRI films, paper report of the history and physical examination data, and patient questionnaire, but not the electrodiagnostic testing findings or the physiatrist’s impression, to arrive at a diagnosis. His impression was compared to the clinical diagnosis.

To ensure that no subjects with exclusion criteria such as polyneuropathy were unintentionally included in the final cohort, we reviewed electrodiagnostic data and available clinical records on all 150 subjects at least 1 year after testing. Neuromuscular disorders were determined by characteristic needle examination findings (eg, brief, short, early recruited polyphasic motor units in proximal muscles for myopathy) and nerve conduction findings (eg, sural nerve or peroneal nerve slowing). We also excluded subjects in whom diagnostic tests were not completed or were technically deficient. For paraspinal mapping, persons who had poor relaxation at 2 or more levels on either side of the paraspinal mapping grid were defined as having inadequate test results and eliminated.

Statistical Analyses 

Data were entered in an Excel database,c checked for errors, and cleaned. SPSSd was used for statistical analysis.38 The Cohen κ was used to measure the agreement between the diagnosis of physiatrist and spine surgeon. A large κ, which is statistically significant, indicates high agreement between the 2 categoric measures. Chi-square tests were performed to examine the relation between categoric variables. One-way analysis of variance was used for 3 or more group comparisons with a Tukey honestly significant difference adjustment for multiple pairwise comparisons of the group means and with a modified Bonferroni adjustment39 applied for the number of tests performed. A P value of .05 or less was considered as statistically significant in all the analyses, except for the multiple pairwise comparisons, where a P value of .018 or less was considered as statistically significant given the Bonferroni adjustment.39 The F test, which tests whether the ratio of the 2 variance estimates is significantly greater than 1, was used to evaluate the statistical significance of between-groups differences for clinical variables and the 3 diagnostic categories.

Participants 

Table 1 demonstrates the selection, dropout, and reallocation of subjects within categories. Forty-seven percent of subjects contacted by telephone underwent testing. One subject dropped out before performing any tests. Another 14 were eliminated due to missing or invalid diagnostic tests, leaving 126 subjects. Despite preliminary screening for neuropathy and risk factors for neuropathy, 8 (5%) of subjects were still found to have a neuropathy or myopathy on electrodiagnostic testing, and another was subsequently found to have sacral cancer which in retrospect could not be seen on the initial MRI. Among the subjects thought by the clinician to have spinal stenosis, the clinician rated the stenosis as severe in 4 subjects, moderate in 22 subjects, and mild in 24 subjects.

Radiologist and Electrodiagnostician Diagnosis 

The radiologist sensitivity for the clinical diagnosis of lumbar stenosis was 59.2% and specificity ranged from 40.9% versus the back pain group to 43.8% versus the asymptomatic group. Combining the control groups, there was no significant relation between the radiologic and the clinical diagnosis of stenosis (χ12 test=0.0, P=.89). The sensitivity of the electrodiagnostician was 70.0% and the specificity ranged from 46.9% versus back pain to 47.4% versus the asymptomatic group. For the combined control groups the electrodiagnostic impression was significant in its relation with clinical stenosis (χ2 test=3.8, P=.05).

To examine the ability of the diagnoses established using electrodiagnostic testing and radiologic examination to predict the clinical diagnosis of spinal stenosis, logistic regression analyses were utilized. Comparing persons diagnosed with stenosis to those diagnosed as being asymptomatic, the radiologist’s diagnosis did not significantly predict the clinician’s diagnosis (χ12 test=0.1, P=.80). The diagnosis obtained through electrodiagnostic testing also did not significantly predict the diagnosis given by the clinician (χ12 test=2.2, P=.14). Also, both variables combined did not significantly predict clinical diagnosis (χ12 test=2.2, P=.33). Comparing subjects with stenosis to those diagnosed with back pain, the findings were similar. Radiologic diagnosis did not significantly predict clinical diagnosis (χ12 test=0.0, P=.99), and electrodiagnostic findings were not significantly predicted of clinical diagnosis, although these results approached significance (χ12 test=2.9, P=.09). The model combining both radiologic and electrodiagnostic findings was not statistically significant (χ22 test=2.9, P=.24).

Data from this study have changed or refined our range of normal for older persons, including higher normative values for paraspinal mapping (>4 in older persons, while >2 was our previous upper limit of normal) and tolerance for polyphasic motor units, which do not differentiate symptomatic from asymptomatic persons.24, 25 In retrospect among the 13 asymptomatic persons labeled by the electrodiagnostician as having spinal stenosis, the only “abnormalities” in 2 were reproducible paraspinal fibrillations that did not meet our current age-related criteria of a score greater than 4. In 5, the only “abnormality” was greater than 2 in 10 polyphasic motor units, a finding that we now disregard in this population. These 2 groups along with combinations of the 3 who were diagnosed as having stenosis with no discernable electromyographic abnormalities and 3 with “abnormal” motor units and paraspinal mapping score of 4 or less totaled 13 (76%) of the 17 asymptomatic persons who were finally given an electrodiagnostic impression of spinal stenosis.

Construct Validity of the Clinical Examination 

We did 2 analyses to examine the construct validity of the clinical diagnosis. First, clinical information was compared with the clinical diagnosis. These include pain severity, ambulation difficulty, tenderness, reflexes, strength, and nerve tension signs. Table 2 shows that there were significant differences between the stenosis, back pain, and asymptomatic groups in all areas except reflex deficit, femoral stretch test, and any abnormality. The Tukey honestly significant difference adjustment for multiple comparisons revealed that stenosis subjects were significantly older than persons in the back pain group (67.66y vs 62.27y, P=.001), had significantly greater pain severity VAS (4.42cm vs 0.16cm, P<.001), and walked significantly slower in a 15-minute walk test (3.76km/h vs 4.37km/h, P=.007) compared with the asymptomatic subjects.

Table 2. Differences in Clinical Factors Between the Clinically Diagnosed Stenosis Subjects and the 2 Different Control Groups, Separately and Combined

	Factors	Stenosis (n=50)	Back Pain (n=44)	Asymptomatic (n=32)	Statistic (χ2 or F test)	P
	Age (y)	67.66±7.27	62.27±6.70	65.84±8.12	F=6.469†	.002
	Pain and disability
	Pain below the knee	31(62.0)	14(31.8)	1(3.1)	χ2test=29.820‡	<.001
	Pain severity VAS	4.42±2.31	3.79±2.43	0.16±0.48	F=39.835‡	<.001
	Ambulation
	Difficulty with walking	33(66.0)	20(45.5)	0(0.0)	χ2test=35.196‡	<.001
	15-min ambulation velocity (km/h)	3.76±0.90	4.10±0.83	4.37±0.83	F=5.125†	.007
	Physical examination
	Tender lumbar or sacral spine	20(40.0)	20(45.5)	1(3.1)	χ2test=17.224	<.001
	Any strength deficit	16(32.0)	8(18.2)	3(9.4)	χ2test=6.356⁎	.042
	Any reflex deficit	19(38.0)	8(18.2)	8(25.0)	χ2test=4.747	.093
	Straight-leg raise: leg pain	13(26.0)	4(9.1)	0(0.0)	χ2test=12.423†	.002
	Femoral stretch test: leg pain	12(24.0)	6(13.6)	2(6.3)	χ2test=4.857	.088
	Clinical summary
	Any abnormalities	35(70.0)	23(52.3)	17(53.1)	χ2test=3.782	.151

NOTE. Values are mean ± standard deviation or n (%).

⁎ P<.05 † P<.01 ‡ P<.001.

The clinical diagnosis was compared to the neurosurgeon’s diagnosis, as shown in table 3. The surgeon’s opinion was influenced by the MRI. For example, both clinicians were asked the likelihood (low, medium, high) that surgery would help the subject. The surgeon’s opinion related to minimal canal diameter on MRI (regression coefficient B=−1.198, P<.05 for high likelihood of a surgical recommendation; B=−.398, P<.01 for medium likelihood), while the clinician’s decision was related to pain severity (B=.329, P<.05 for medium likelihood). Given these differences, a κ of .437 (P<.001) suggests that the clinician took an approach to diagnosis not dissimilar to the neurosurgeon.

Table 3. Agreement Between Clinical Impression and Spine Surgeon Record Review

		Spine Surgeon				κ	P
		Asymptomatic	Back Pain	Stenosis	Total
	Physiatrist					.437	< .001
	Asymptomatic	28	3	1	32
	Back pain	6	25	13	44
	Stenosis	2	22	26	50
	Total	36	50	40	126

NOTE. The spine surgeon reviewed MRI films and reports, physiatrist record of examination, and patient questionnaires, but not electromyographic information or the actual clinical impression.

Because the population included persons with varying levels of clinical and radiologic stenosis, one might wonder whether the relation between clinical diagnosis and radiologic diagnosis improved with more severe radiologic findings. Table 4 shows that this did not occur (χ62 test=3.9, P=.69). Indeed, 2 (25%) persons whom the radiologist diagnosed with severe stenosis had only LBP and 1 (12.5%) had no pain whatsoever. There was also no relation between the presence of a disk herniation and the clinical diagnosis.

Table 4. The Lack of Relation Between Radiologist Rated Severity and the Clinician Diagnosis

		Clinical Diagnosis			χ2	P
		Stenosis, n (%)	Back Pain, n (%)	Asymptomatic, n (%)
	Radiologist diagnosis				3.900	.690
	Severe stenosis	5(10.2)	2(4.5)	1(3.1)
	Moderate stenosis	17(34.7)	13(29.5)	9(28.1)
	Mild stenosis	7(14.3)	11(25.0)	8(25.0)
	No stenosis	20(40.8)	18(40.9)	14(43.8)
	Total	49(100.0)	44(100.0)	32(100.0)
	Disk herniation any level	8(16.3)	6(13.6)	6(18.8)	0.367	.832