M Sensitivity,Specificity,andVariabilityofNerveConductionVelocityMeasurementsinCarpalTunnelSyndrome

Sensitivity, Specificity, and Variability of Nerve Conduction Velocity Measurements in Carpal Tunnel Syndrome

Henry L. Lew, MD, PhD, Elaine S. Date, MD, Steven S. Pan, MD, Peter Wu, MD, Paul F. Ware, MD, Wade S. Kingery, MD

ABSTRACT. Lew HL, Date ES, Pan SS, Wu P, Ware PF, Kingery WS. Sensitivity, specificity, and variability of nerve conduction velocity measurements in carpal tunnel syndrome.

Arch Phys Med Rehabil 2005;86:12-6.

Objective: To explore the diagnostic values of 8 commonly used electrodiagnostic techniques for measuring median nerve conduction velocity (NCV) in carpal tunnel syndrome (CTS).

Design: Sensitivity and specificity analyses.

Setting: A hospital-based electrodiagnostic laboratory.

Participants: Forty-four normal hands and 136 symptom- atic hands.

Interventions: Not applicable.

Main Outcome Measures: (1) Long-segment studies: anti- dromic wrist-to-digit sensory NCV without subtraction, (2) short-segment studies: transcarpal palm-to-wrist mixed NCV without subtraction, and (3) 2 segment studies: antidromic transcarpal sensory NCV with subtraction (differential calcu- lation from wrist-to-digit and palm-to-digit segments). Both onset and peak latency values were obtained for calculating the NCV. Sensitivity, specificity, and coefficient of variance were calculated for each NCV study.

Results: The short-segment, onset latency– based transcar- pal mixed NCV yielded the highest sensitivity (75%).

Conclusions: Results from measurement of a single, short- nerve segment tended to be superior to results obtained by either long-segment studies or differential subtraction between 2 segments of the same nerve in the electrodiagnosis of CTS.

Explanations for our results are offered from both electrophysi- ologic and statistical perspectives.

Key Words: Carpal tunnel syndrome; Eletrodiagnosis; Re- habilitation; Sensitivity and specificity.

© 2005 by American Congress of Rehabilitation Medicine and the American Academy of Physical Medicine and Rehabilitation

M

ANY ELECTRODIAGNOSTIC TECHNIQUES have been developed to assist in the diagnosis of carpal tunnel syndrome (CTS), the most common nerve entrapment syn- drome.^1,2 However, there is no consensus as to which tech- nique is the most sensitive or specific.^3-11 The discrepancies between previous studies have been attributed to lack of con- sistency regarding (1) the techniques used (sensory vs motor, antidromic vs orthodromic), (2) use of parameters (peak la-

tency vs onset latency), (3) inclusion and exclusion criteria, and (4) statistical analysis. Moreover, previous data have rarely been discussed from the perspective of coefficient of variance (CV).

In 1991, the American Academy of Electrodiagnostic Med- icine’s (AAEM) Quality Assurance Committee performed an exhaustive critical analysis of the CTS electrodiagnostic liter- ature¹²and concluded that (1) median sensory nerve conduc- tion studies (NCSs) are more sensitive than median motor NCSs and (2) short-segment median sensory or mixed NCSs (wrist-to-palm) are more sensitive than long-segment sensory (wrist-to-digit) or mixed NCSs. By using standardized mea- surement techniques and statistical analysis, we sought to an- alyze the sensitivity, specificity, and CV of 8 commonly used electrodiagnostic techniques in the diagnosis of CTS. The techniques were (1) long-segment: antidromic digital (wrist-to- digits) sensory nerve conduction velocity (NCV), (2) short- segment: transcarpal (palm-to-wrist) mixed NCV, and (3) 2 segments: antidromic transcarpal sensory NCV (derived from the difference between wrist-to-digit and palm-to-digit seg- ments of the same nerve branch).

METHODS

The local institution’s human subjects subcommittee ap- proved the experimental protocol and informed consent was obtained for all testing.

References Group and Control Group

Subjects with no symptoms or signs of median mononeu- ropathy were recruited from among healthy volunteers. The exclusion criteria were (1) numbness, tingling, or pain in the hands or digits; (2) history of median, radial, or ulnar monon- europathy; (3) other peripheral neuropathy; and (4) diabetes.

Forty-four healthy subjects (24 men, 20 women; mean age, 44.0⫾12.9y; range, 25⫺80y) were recruited. Results of studies from only the right hands were included to minimize inherent biologic variability.¹³To establish the normative database, we randomly selected 20 of the 44 healthy subjects as our refer- ence group, and the remaining 24 subjects were assigned as the control group for later comparison with the CTS group (136 symptomatic hands).

CTS Group

Patients referred to our electrodiagnostic laboratories with suspected CTS were recruited for this study. The clinical diagnosis of CTS was based on the presence of numbness, tingling, or pain in a median nerve distribution that had occurred at least 3 times a week for at least 3 months.

Patients with either a history of CTS release surgery, absent median sensory nerve action potential (SNAP), diminished ulnar SNAP amplitude (⬍12␮V), or prolonged ulnar sen- sory distal latency (⬎3.7ms)¹⁴were excluded. One hundred seventy hands were screened and 34 were eliminated, based on the exclusion criteria, leaving 136 hands for final anal- ysis. The excluded subjects were older than the recruited subjects (mean age, 51.5⫾18.2y vs 45.3⫾15.6y, P⬍.05),

From the Physical Medicine and Rehabilitation Service, VA Palo Alto Health Care System, Palo Alto, CA (Lew, Date, Pan, Ware, Kingery); Division of PM&R, Stanford University School of Medicine, Stanford, CA (Lew, Date, Pan, Kingery);

and Palo Alto Medical Foundation, Palo Alto, CA (Wu).

No commercial party having a direct financial interest in the results of the research supporting this article has or will confer a benefit upon the authors(s) or upon any organization with which the author(s) is/are associated.

Reprint requests to Henry L. Lew, MD, PhD, PM&R Service 117, VAPAHCS, 3801 Miranda Ave, Palo Alto, CA 94304, e-mail: henrylew@stanford.edu.

0003-9993/05/8601-8974$30.00/0 doi:10.1016/j.apmr.2004.03.023

and the excluded subjects were predominantly male (81.2%

vs 47.8%, P⬍.001). By using the normative values derived from our reference group, we compared CTS and control groups to determine the sensitivity and specificity of various electrodiagnostic techniques.

Electrodiagnostic Studies

Limb temperature (recorded over distal wrist crease) was maintained at or above 32°C. We used a standard electrodiag- nostic instrument^a with bandwidth setting of 20Hz to 3kHz, gain of 20␮V/division, and sweep speed of 1ms/division. The 8 electrodiagnostic techniques are described in the following sections.

Long-segment studies (antidromic wrist-to-digit sensory NCV without subtraction). Figure 1illustrates our standard-

ized electrode placements for stimulation and recording. The median nerve was stimulated at the wrist. For digits 2, 3, and 4 (D2, D3, D4), the active recording ring electrode was placed 14cm distal to the stimulation site. For digit 1 (D1), the active recording ring electrode was placed 10cm distal to stimulation sites, measured in a straight line with the thumb extended, and on the palmar plane of the hand.³ The active ring recording electrode was always separated from the reference electrode by 3cm, and the “just supramaximal” response was used for de- termination of latency. Both onset latency and peak latency were recorded.

Short-segment studies (transcarpal palm-to-wrist mixed NCV without subtraction). The median nerve was stimulated in the mid-palm (between the second and third metacarpals), and waveforms were recorded at the wrist (8cm proximal to the

Fig 1. Eight commonly used electrodiagnostic tests for CTS. Abbreviations: D1, digit 1 ; D2, digit 2; D3, digit 3; D4, digit 4; NA, not available; P, palm; SNCV, sensory nerve conduction velocity; W, wrist.

stimulation site) with bar electrodes. Both peak and onset latencies were measured.

Two segment studies (antidromic transcarpal sensory NCV with subtraction). The median nerve was stimulated at the wrist and at the palm. The distance between stimulation sites (wrist, palm) was 8cm. The active recording ring electrodes were placed at D2, D3, and D4. Again, both onset latency and peak latency were recorded.

Statistical Analysis

Data from 20 reference subjects were used to develop the normative means and standard deviations (SDs) for the previ- ously mentioned techniques. There were 8 tests but 16 param- eters because both onset and peak latency data were collected.

NCV values were calculated by dividing the distance traveled over the corresponding latency. Then, the 24 control subjects were compared with the CTS patients to derive sensitivity and specificity values. We used the Wilcoxon nonparametric signed-rank test to compare the CVs in each pair of reference values. Statistical analysis was performed by using SAS, ver- sions 8.1.^b

RESULTS

Table 1summarizes the mean NCV values in the reference, control, and CTS groups. The lower limits of reference values were derived by subtracting 2 SDs from the means (left half of table 1). The CV for each reference value was derived from the SD divided by the mean. For ease of visualization, NCV values based on onset latency and peak latency are listed in sequence.

In the long-segment wrist-to-digit NCV without subtraction, the calculated wrist-to-digit NCV values based on onset latency were all faster than those derived from peak latency because onset latency was always shorter.

The CV of each reference value reflects the standardized variability of each parameter: larger CV suggests greater vari-

ability of the measured value. Among the long-segment and 2-segment nerve conduction parameters, the CVs based on onset latency were all larger than those estimated from peak latency (signed-rank test, P⬍.05;table 1, middle column). This trend was not observed in the short-segment studies. Because the SNAP amplitudes obtained with recording electrodes over the median nerve at the wrist were consistently larger, it was relatively easy to specify the onset latencies in short-segment studies.

The sensitivity and specificity values are listed in the last 2 columns on the right side of table 1. Among the 8 tests (16 parameters), test 5 (onset latency– based NCV) yielded the highest sensitivity (75%). The sensitivities of other tests were relatively low (between 33%– 46%). All tests had very high specificities, ranging from 83% to 100%.

A post hoc analysis was added to compare median and ulnar sensory peak latencies from digits 4 and 5. The analysis of latency difference between D4 and D5 showed a mean value of .07⫾.22ms (upper limit⫽.51ms) in the reference group, as shown intable 2. By using the previously mentioned reference values, very low sensitivity (25%) and very high specificity (100%) were derived. In this scenario, in which the mean centers on zero, it is not appropriate to use CV to evaluate data variability.

DISCUSSION

In this study, we attempted to explore the diagnostic values of 8 commonly used electrodiagnostic techniques for CTS by comparing their sensitivity, specificity, and CV. Our results showed that the short-segment, onset latency– based transcar- pal mixed NCSs had the highest sensitivity (75%). Although our normative values were similar to data reported by other authors,^3,4,15-20 the sensitivity values in our study were rela- tively low. Perhaps the inclusion and exclusion criteria caused this decline in sensitivity. Because we excluded patients with

Table 1: Reference Values, Means, Sensitivities, and Specificities of 8 Electrodiagnostic Tests

Reference Group Control Group CTS Group

Sensitivity (%)

Specificity (%) Mean⫾ SD

(m/s)

Lower Limit

CV (%)

Mean (m/s)

Mean (m/s)

Long segment: wrist-to-digit sensory NCV (antidromic wrist-to-digit without subtraction)

1 OL D1 48.7⫾5.0 38.7 10.2 47.5 41.8 38 100

PL D1 38.1⫾3.2 31.7 8.4 37.3 33.1 39 96

2 OL D2 55.6⫾4.4 46.8 8.0 54.8 49.7 42 100

PL D2 44.1⫾2.7 38.7 6.1 43.3 39.5 39 92

3 OL D3 56.1⫾4.3 47.4 7.7 54.5 48.2 44 96

PL D3 44.8⫾3.4 38.0 7.6 43.6 39.0 42 100

4 OL D4 53.8⫾5.0 43.7 9.4 52.5 47.0 35 100

PL D4 43.8⫾3.7 36.3 8.5 42.4 38.7 37

(avg⫽39.5)

96

Short segment: transcarpal mixed NCV (palm-to-wrist without subtraction)

5 OL 51.6⫾3.4 44.7 6.7 51.1 46.2 75 83

PL 40.8⫾4.2 32.4 10.3 39.7 35.7 37

(avg⫽56)

96

Two segments: transcarpal sensory NCV (antidromic wrist-to-digit and palm-to-digit difference)

6 OL D2 55.7⫾7.5 40.8 13.4 54.6 48.3 33 96

PL D2 52.6⫾5.2 42.1 10.0 50.0 42.8 46 88

7 OL D3 57.0⫾7.6 41.8 13.3 55.1 46.8 40 100

PL D3 53.3⫾6.4 40.5 12.0 50.7 43.1 41 100

8 OL D4 54.8⫾6.6 41.6 12.1 51.8 45.5 39 96

PL D4 51.7⫾5.7 40.4 11.0 49.0 41.6 44

(avg⫽40.5)

88

Abbreviations: avg, average; OL, onset latency based, PL, peak latency based.

either unobtainable SNAP or history of CTS release surgery, our chosen patient population could be shifted toward milder cases of CTS. Although we purposefully divided the healthy subjects into a reference group and a control group to decrease the possibility of falsely enhancing the specificity, the resultant specificities were still high (83%–100%). It may suggest that our healthy subjects were genuinely free from CTS. Atroshi et al²¹estimated an overall positive rate of 18% (ie, 82% speci- ficity) by electrophysiologic tests in an age- and gender-strat- ified asymptomatic sample that was randomly selected from the general population. The specificity values in our study are higher than those reported in several population-based stud- ies,^21-23 perhaps because only healthy adults were included in the control group. Therefore, the generalizability of our results is limited because the control group did not precisely represent the general population.

Short-segment (palm-to-wrist, 8cm) NCSs theoretically have better diagnostic sensitivity than long-segment (wrist-to-digit, 14cm) studies because the effect of the focal pathology on NCV may be diluted in long-segment studies. This phenome- non was reflected in the sensitivity section of our results, which is also consistent with the AAEM report.¹² Although short- segment (palm-to-wrist) NCSs showed a similar “transcarpal”

concept with the 2-segment NCSs (derived from the difference between wrist-to-digit and palm-to-digit results), the latter had low sensitivity values (average, 40.5%), whereas the former yielded a higher sensitivity (average, 56%). It implies that direct measurement from a single nerve segment may be su- perior to indirect differential calculation from 2 related nerve segments. From the perspective of data variability, the variance increases as a result of doubling the measurement error. Not surprisingly, we found that all the 2-segment transcarpal NCV (derived from 2 sets of latency and distance measurements) had larger CVs. Because larger CVs also correspond with wider ranges of normal limits in the reference values, the sensitivities of these tests would therefore be lower. We have yet to explain why the onset latency– based transcarpal mixed NCV showed much higher sensitivity than its peak latency counterpart (75%

vs 37%). A possibility may come from a pathophysiologic standpoint: in the earlier phases of CTS, the fast conducting, larger myelinated fibers may be affected to a larger degree. If this is true, it will have a more pronounced effect on the onset-phase of the SNAP waveform morphology.

In the 2002 summary statement of AAEM, the American Academy of Neurology, and the American Academy of Phys- ical Medicine and Rehabilitation,²⁴and a previous study on combined sensory index,¹ the validity and reliability of com- paring electrophysiologic measures have been established. Our present study focused on the sensitivity, specificity, and vari- ability of single nerve parameters derived from median NCS.

We acknowledge that these values are influenced by biologic determinants such as the subjects’ age, concurrent disease, gender, body mass index, or other anthropometric factors.²² Theoretically, it would be best to calculate median-ulnar peak latency difference from a single finger such as D4. Neverthe- less, we did a retrospective analysis of the median-ulnar peak latency difference by comparing median and ulnar latency

values obtained from D4 and D5, respectively. The result was a very low sensitivity (25%) and very high specificity (100%).

Because the latency differences were obtained from 2 separate measurements (wrist-to-digit 4, wrist-to-digit 5), the measure- ment error naturally increased 2-fold. The increased variability widened the range of the normal limits in the reference values and contributed partially to the low sensitivity. Another likely explanation for the low sensitivity lies with the inclusion and exclusion criteria. Because we excluded patients with unob- tainable SNAP and previous CTS release surgery, the CTS group was probably milder in disease severity than one would expect in a typical CTS population. This was reflected by the small difference between the mean latency comparison values in CTS group and the upper limit derived from reference group (.53ms vs .51ms).

CONCLUSIONS

Our study showed that among the 8 median NCV tests, the short-segment, onset latency– based transcarpal NCV was most sensitive in diagnosing CTS. The study also suggests that direct measurement of a single nerve segment is superior to either long-segment studies or differential subtraction between 2 seg- ments of the same nerve. The data presented here were derived from a single electrodiagnostic laboratory; therefore, more research into the validity and generalizability of our results may be warranted.

References

1. Lew HL, Wang L, Robinson LR. Test-retest reliability of com- bined sensory index: implications for diagnosing carpal tunnel syndrome. Muscle Nerve 2000;23:1261-4.

2. Stevens JC, Sun S, Beard CM, O’Fallon WM, Kurland LT. Carpal tunnel syndrome in Rochester, Minnesota, 1961 to 1980. Neurol- ogy 1988;38:134-8.

3. Kothari MJ, Rutkove SB, Caress JB, Hinchey J, Logigian EL, Preston DC. Comparison of digital sensory studies in patients with carpal tunnel syndrome. Muscle Nerve 1995;18:1272-6.

4. Macdonell RA, Schwartz MS, Swash M. Carpal tunnel syndrome:

which finger should be tested? An analysis of sensory conduction in digital branches of the median nerve. Muscle Nerve 1990;13:

601-6.

5. Versino M, Romani A. Test-retest reliability of combined sensory index. Muscle Nerve 2001;24:1097-8. Comment in: Muscle Nerve 2000;23:1261-4.

6. Robinson LR, Micklesen PJ, Wang L. Optimizing the number of tests for carpal tunnel syndrome. Muscle Nerve 2000;23:1880-2.

7. Monga TN, Shanks GL, Poole BJ. Sensory palmar stimulation in the diagnosis of carpal tunnel syndrome. Arch Phys Med Rehabil 1985;66:598-600.

8. Di Benedetto M, Mitz M, Klingbeil GE, Davidoff D. New criteria for sensory nerve conduction especially useful in diagnosing car- pal tunnel syndrome. Arch Phys Med Rehabil 1986;67:586-9.

9. Kingery WS, Park KS, Wu PB, Date ES. Electromyographic motor Tinel’s sign in ulnar mononeuropathies at the elbow. Am J Phys Med Rehabil 1995;74:419-26.

Table 2: Reference Values, Means, Sensitivity, and Specificity of Latency Comparison Between D4 (median) and D5 (ulnar)

Reference Group Control Group CTS Patients

Sensitivity (%)

Specificity (%) Mean⫾ SD

(m/s)

Upper Limit

Mean⫾ SD (m/s)

Mean⫾ SD (m/s)

.07⫾.22 .51 .01⫾.21 .53⫾.92 25 100

10. Seror P. Comparative diagnostic sensitivities of orthodromic or antidromic sensory inching test in mild carpal tunnel syndrome.

Arch Phys Med Rehabil 2000;81:442-6.

11. Trojaborg W, Grewal RP, Weimer LH, Sheriff P. Value of latency measurements to the small palm muscles compared to other con- duction parameters in the carpal tunnel syndrome. Muscle Nerve 1996;19:243-5.

12. Jablecki CK, Andary MT, So YT, Wilkins DE, Williams FH.

Literature review of the usefulness of nerve conduction studies and electromyography for the evaluation of patients with carpal tunnel syndrome. AAEM Quality Assurance Committee. Muscle Nerve 1993;16:1392-414.

13. Dorfman LJ, Robinson LR. AAEM minimonograph #47: norma- tive data in electrodiagnostic medicine. Muscle Nerve 1997;20:4- 14.

14. Kimura J. The carpal tunnel syndrome: localization of conduction abnormalities within the distal segment of the median nerve. Brain 1979;102:619-35.

15. Buchthal F, Rosenfalck A. Sensory conduction from digit to palm and from palm to wrist in the carpal tunnel syndrome. J Neurol Neurosurg Psychiatry 1971;34:243-52.

16. Cioni R, Passero S, Paradiso C, Giannini F, Battistini N, Rush- worth G. Diagnostic specificity of sensory and motor nerve con- duction variables in early detection of carpal tunnel syndrome.

J Neurol 1989;236:208-13.

17. Goyal V, Bhatia M, Padma MV, Jain S, Maheshwari MC. Elec- trophysiological evaluation of 140 hands with carpal tunnel syn- drome. J Assoc Physicians India 2001;49:1070-3.

18. Jackson DA, Clifford JC. Electrodiagnosis of mild carpal tunnel syndrome. Arch Phys Med Rehabil 1989;70:199-204.

19. Kuntzer T. Carpal tunnel syndrome in 100 patients: sensitivity, specificity of multi-neurophysiological procedures and estimation of axonal loss of motor, sensory and sympathetic median nerve fibers. J Neurol Sci 1994;127:221-9.

20. Tackmann W, Kaeser HE, Magun HG. Comparison of ortho- dromic and antidromic sensory nerve conduction velocity mea- surements in the carpal tunnel syndrome. J Neurol 1981;224:257- 66.

21. Atroshi I, Gummesson C, Johnsson R, Ornstein E, Ranstam J, Rosen I. Prevalence of carpal tunnel syndrome in a general pop- ulation. JAMA 1999;282:153-8.

22. Salerno DF, Franzblau A, Werner RA, Bromberg MB, Armstrong TJ, Albers JW. Median and ulnar nerve conduction studies among workers: normative values. Muscle Nerve 1998;21:999-1005.

23. Bingham RC, Rosecrance JC, Cook TM. Prevalence of abnormal median nerve conduction in applicants for industrial jobs. Am J Ind Med 1996;30:355-61.

24. American Association of Electrodiagnostic Medicine, American Academy of Neurology, American Academy of Physical Medi- cine and Rehabilitation. Practice parameter for electrodiagnostic studies in carpal tunnel syndrome: summary statement. Muscle Nerve 2002;25:918-22.

Suppliers

a. Nicolet Viking IV; Nicolet Biomedical, PO Box 44451, Madison, WI 53744-4451.

b. SAS Institute, 100 SAS Campus Dr, Cary, NC 27513.