Discriminatory sonographic criteria for the diagnosis of carpal tunnel syndrome

Abstract

Objective

Sonographic examination of the median nerve has been suggested as a useful alternative to electrophysiologic study in the diagnosis of carpal tunnel syndrome. To determine its usefulness and the best diagnostic criterion, sonograms of patients with the disease were compared with sonograms of healthy subjects in a case–control study.

Methods

Patients with carpal tunnel syndrome and asymptomatic controls who were matched for age and sex were enrolled and underwent sonography of the wrists. Eight separate sonographic criteria were analyzed in each wrist. Data from the patient group and the control group were compared to establish optimal diagnostic criteria for carpal tunnel syndrome, using receiver operating characteristic analytic techniques.

Results

Thirty-five patients with carpal tunnel syndrome and 35 asymptomatic controls were examined. Increased cross-sectional area of the median nerve was found to be the most predictive measure of carpal tunnel syndrome, proximal to the tunnel inlet, at the tunnel inlet, and at the tunnel outlet, with significant differences between patients and controls. Using a receiver operating characteristic curve, a cut-off value >0.098 cm² at the tunnel inlet provided a diagnostic sensitivity of 89% and a specificity of 83%.

Conclusion

Sonographic measurement of the median nerve cross-sectional area is both sensitive and specific for the diagnosis of carpal tunnel syndrome.

A diagnosis of carpal tunnel syndrome (CTS) is usually based on symptom characteristics and confirmatory neurophysiologic evaluation (1). Sonographic examination of the median nerve in CTS has been proposed as a useful alternative to neurophysiologic study in the diagnosis of CTS, and proposals for different criteria have been presented (2–7), as summarized in Table 1. The lack of consensus on the optimal criteria for CTS may be attributable to differences in equipment, patient populations, and sonographic techniques. These discrepancies underscore the need for each center to validate internally the criteria it uses to diagnose CTS with ultrasound.

Table 1. Sonographic criteria for the diagnosis of carpal tunnel syndrome*

Source (ref.)	Number of patients/wrists	Significant positive criteria for cross-sectional area of median nerve	Sensitivity	Specificity
* Additional diagnostic criteria were as follows: bowing of retinaculum at tunnel outlet >2.5 mm (sensitivity 0.81, specificity 0.64) in Sarria et al (3); increased flattening ratio at tunnel outlet >4.6 (sensitivity and specificity both NA) and increased bowing of the flexor retinaculum at tunnel outlet (sensitivity and specificity both NA) in Buchberger et al (6); increased flattening ratio at tunnel outlet (sensitivity and specificity both NA) in Nakamichi et al (7). NA = not available.
Swen et al (2)	63/NA	At tunnel inlet >0.11 cm^2	0.7	0.63
Sarria et al (3)	40/64	Proximal to tunnel >0.11 cm^2;	0.73	0.57
		at middle tunnel >0.11 cm^2;	0.73	0.57
		at tunnel outlet >0.11 cm^2	0.74	0.57
Lee et al (4)	16/16	At tunnel inlet >0.15 cm^2	0.88	0.96
Duncan et al (5)	68/102	At tunnel inlet >0.09 cm^2	0.82	0.97
Buchberger et al (6)	18/20	Enlarged at tunnel inlet;	NA	NA
		enlarged at tunnel outlet	NA	NA
Nakamichi et al (7)	125/201	Enlarged proximal to tunnel;	NA	NA
		enlarged at tunnel outlet;	NA	NA
		enlarged beyond tunnel outlet

This study aimed to prospectively evaluate the characteristics of the median nerve in CTS patients against the findings in normal, healthy subjects, to elicit the optimal discriminatory sonographic criteria and relevant threshold values.

PATIENTS AND METHODS

Cases and controls.

This was a prospective, age-matched case–control study of CTS in a cohort comprised solely of women, since women account for the majority of CTS patients in the local population (8). The study ethics were reviewed and approved, and patients gave their written informed consent. Consecutive CTS patients from our neurology and rheumatology clinics were invited to join the study. Patients were enrolled into the study after electrophysiologic examination, if the following criteria were fulfilled: 1) presence of sensory symptoms over the median nerve distribution, and 2) confirmatory neurophysiologic results (prolonged median nerve distal motor latencies >4 ms or a median-ulnar palmer sensory latency difference >0.5 ms).

Motor and sensory nerve conduction studies were performed on the median and ulnar nerves using standard techniques of supramaximal stimulation and surface electrodes, in accordance with the American Academy of Neurology summary statement on the desired nerve conduction and electromyographic studies in patients with suspected CTS (9). Compound muscle action potentials, distal motor latencies, nerve conduction velocities, and sensory nerve action potentials were measured, and needle electromyographic recordings of the abductor pollicis brevis were performed.

The following exclusion criteria were applied: 1) history of wrist surgery (including carpal tunnel injection) or fracture; 2) clinical or electrophysiologic evidence of accompanying conditions that mimic CTS or interfere with its evaluation, such as proximal median neuropathy, cervical radiculopathy, or polyneuropathy; 3) history of underlying disorders associated with CTS such as diabetes mellitus, rheumatoid arthritis, pregnancy, acromegaly, or hypothyroidism; and 4) anatomic variation of the median nerve such as bifurcation proximal to the tunnel inlet.

In addition, sonographic examination was performed on asymptomatic control subjects who had no disorder of the upper limbs or conditions predisposing to CTS (such as those listed in the exclusion criteria above). Controls were recruited from among healthy volunteers whose occupation base was similar to that of the patients, to minimize the impact of occupation. The occupation profile mostly consisted of housewives, with hospital staff, clerical workers, saleswomen, and teachers making up the rest.

To ensure unbiased examination, a random identification number generated by computer identified each participant, and a single radiologist (JFG) performed the sonographic examinations. The examiner was requested not to enquire about symptoms and the patients were asked not to divulge their symptoms during the examination. Immediately prior to sonography, the height and weight of patients and controls were measured, with the subjects attired in light indoor clothing without shoes. A single physician (SMW) obtained each subject's clinical history and performed a physical examination to exclude polyneuropathy or any other coexisting secondary cause for CTS.

Sonographic technique.

Sonographic examinations were performed within 1 week of the electrophysiologic study. Examinations were performed with a 13-5 MHz Multi-D linear array transducer (Siemens Sonoline Elegra Advanced; Siemens, Munich, Germany). Subjects were seated facing the examiner, with their wrists resting on a hard surface in a prone neutral position (10) and the fingers semiextended. Transverse images of the median nerve were obtained at 4 levels, namely 1) the distal forearm, 2) immediately proximal to the carpal tunnel inlet, 3) at the carpal tunnel inlet, and 4) at the carpal tunnel outlet. Locations 2, 3, and 4 were identified by direct visualization of the flexor retinaculum and its proximal and distal margins on sonography.

A total of 8 separate measurements were obtained on each wrist. At each level, the cross-sectional area of the median nerve was measured by tracing with electronic calipers around the margin of the nerve at the time of sonography (direct tracing) (5). The margin of the nerve was defined as the margin outside the hypoechoic nerve fascicles and inside the hyperechoic nerve sheath (4). The flattening ratio (defined as the ratio of the width of the nerve at its midpoint to the depth of the nerve at its midpoint) was measured at the tunnel inlet and outlet. Bowing of the flexor retinaculum was measured at the distal margin of the pisiform bone, equating to the midpoint of the carpal tunnel. Bowing was defined as the maximum height of the retinaculum above a line subtended between its radial and ulnar carpal attachments. The thickness of the flexor retinaculum was measured as close to the midline as possible in the midportion of the carpal tunnel.

In order to assess reliability, every eighth subject was asked to return within 24 hours for a repeat ultrasound. A total of 9 subjects, consisting of 7 controls and 2 CTS patients, were assessed for this purpose.

Statistical analysis.

According to the results of previous studies (2–7) (Table 1), the cross-sectional area of the median nerve was found to be the best discriminatory criterion to identify patients with CTS. The sample size of 35 selected for both the patient and the control group allowed detection of a difference in cross-sectional area of 0.03 cm², assuming α = 0.05 and a power of 90% with a SD of 0.04 cm² (GraphPad Statmate software; GraphPad, San Diego, CA).

Separate analysis of sonographic measurements of the right and left wrists was made for CTS patients and controls. Paired t-test was used after initial analysis confirmed Gaussian distribution of the data. The reliability of the sonographic measurements was calculated (11). Pearson correlation was performed to assess the strength of the association between sonographic measurements and electrophysiologic measurements. The sensitivity, specificity, positive-predictive value, and negative-predictive value together with the likelihood ratio were calculated, and receiver operating characteristic (ROC) curves were used to illustrate optimal threshold values. A 2-sided P value of less than 0.05 was considered statistically significant. SPSS, version 10.1 (SPSS, Chicago, IL) was used in the performance of all statistical analyses.

RESULTS

Thirty-six consecutive patients with CTS were invited to join the study together with 36 controls. One patient had a bifid median nerve found by sonographic examination and was excluded from the final analysis, along with the corresponding control, leaving 70 subjects (35 pairs of CTS patients and controls). The mean ages of the CTS patients and controls were 44.4 years (SD 8.8) and 44.0 years (SD 8.9), respectively. There was a statistically significant difference between the body mass index (BMI) of the CTS patients (mean of 24.9 [SD 4.3]) and that of the controls (22.3 [SD 3.7]) (P = 0.01). Among the CTS patients, 12 had symptoms in the right hand, 4 had symptoms in the left hand, and the remaining 19 had bilateral symptoms. All participants had both hands examined sonographically, but we considered each wrist separately in clinical diagnosis; thus, 70 wrists of the controls and 54 symptomatic wrists of the CTS patients were analyzed.

Comparison results for the 8 separate sonographic measurements analyzed, along with the reliability of the measurements, are summarized in Table 2. Statistically significant differences between CTS patients and controls were found in the median nerve cross-sectional area proximal to the carpal tunnel, at the tunnel inlet, and at the tunnel outlet, in both the right and the left wrists (P < 0.01). There were also significant differences found for the flattening ratio on the left side and retinacular bowing on the right side. Consistent reliability on both sides was found in measurements proximal to the tunnel and at the tunnel inlet. A positive correlation with the latency delay was found (data not shown), but this was only significant for increased cross-sectional area at the tunnel inlet (r = 0.72, P < 0.01) and retinacular bowing (r = 0.36, P = 0.03) in the right wrist, and for increased cross-sectional area at the tunnel outlet (r = 0.39, P = 0.02) in the left wrist.

Table 2. Sonographic measurements of the median nerve*

Measure, wrist	CTS patients	Controls	P	Reliability
* Except where otherwise indicated, values are the mean ± SD. CTS = carpal tunnel syndrome.
1. Cross-sectional area in forearm, cm^2
Right	0.06 ± 0.01	0.06 ± 0.01	0.48	0.33
Left	0.07 ± 0.01	0.07 ± 0.01	0.69	0.74
2. Cross-sectional area proximal to tunnel, cm^2
Right	0.11 ± 0.04	0.08 ± 0.02	<0.01	0.79
Left	0.10 ± 0.03	0.08 ± 0.01	<0.01	0.65
3. Cross-sectional area at tunnel inlet, cm^2
Right	0.13 ± 0.05	0.09 ± 0.02	<0.01	0.82
Left	0.12 ± 0.03	0.08 ± 0.01	<0.01	0.81
4. Flattening ratio at tunnel inlet
Right	2.88 ± 0.58	3.20 ± 0.57	0.04	0.32
Left	2.99 ± 0.58	3.32 ± 0.84	0.05	−0.53
5. Cross-sectional area at tunnel outlet, cm^2
Right	0.12 ± 0.05	0.08 ± 0.02	<0.01	0.54
Left	0.12 ± 0.04	0.08 ± 0.02	<0.01	0.37
6. Flattening ratio at tunnel outlet
Right	3.35 ± 0.68	3.48 ± 0.86	0.44	0.47
Left	3.10 ± 0.65	3.83 ± 0.92	<0.01	0.00
7. Retinacular bowing, mm
Right	2.75 ± 0.71	2.33 ± 0.54	<0.01	0.57
Left	2.68 ± 0.64	2.43 ± 0.54	0.08	−0.02
8. Retinacular thickness, mm
Right	1.02 ± 0.14	0.99 ± 0.21	0.36	0.46
Left	1.00 ± 0.23	0.97 ± 0.19	0.61	0.03

Because there were no significant differences in side-to-side variation in the median nerve caliber between the right and left wrists in controls, we took the average values from the control group to produce the ROC curve. ROC curves for the left and right wrists of the CTS patients were then created (Figures 1A and B). Since the ROC curves of the left and right hand of the CTS patients were comparable against the average value measurement of the control wrists, an averaged ROC curve was produced (Figure 1C). The cross-sectional areas of the median nerve proximal to the carpal tunnel, at the tunnel inlet, and at the tunnel outlet were found to be the most useful discriminatory criteria, with areas under the curve of 0.77 (P < 0.01), 0.91 (P < 0.01), and 0.81 (P < 0.01), respectively. The following optimal discriminatory threshold values were chosen: 0.088 cm² proximal to the carpal tunnel, 0.098 cm² at the tunnel inlet, and 0.085 cm² at the tunnel outlet. These yielded a sensitivity, specificity, positive-predictive value, and negative-predictive value proximal to the tunnel of 0.74, 0.63, 0.67, and 0.71, respectively, at the tunnel inlet 0.89, 0.83, 0.84, and 0.88, respectively, and at the tunnel outlet 0.80, 0.51, 0.64, and 0.73, respectively (see Figure 1).

Figure 1. A, Receiver operating characteristic curves for sonographic measurements (cm²) in A, the right hand of the carpal tunnel syndrome (CTS) patients and B, the left hand of the CTS patients, and C, all CTS sonographic measurements averaged against the controls, in the median nerve proximal to the tunnel (red line), at the tunnel inlet (green line), and at the tunnel outlet (blue line).

Data were further analyzed comparing the left and right hands of the CTS patients separately with the corresponding right and left hands of the control subjects, so that there were as many observations as there were people. For median nerve caliber, almost identical results (data not shown) were obtained. Applying this analysis, retinacular bowing and thickness were found to be statistically significantly greater in CTS patients than in controls. However, at determined threshold levels, the sensitivity/ specificity for both retinacular bowing (0.74/0.59 for right hand, 0.61/0.45 for left hand at a threshold of 2.5 mm) and retinacular thickness (0.64/0.36 for right hand, 0.65/ 0.38 for left hand at a threshold of 0.95 mm) was only moderate.

DISCUSSION

This study shows that the most useful discriminatory criterion in diagnosing CTS is the cross-sectional area of the median nerve, proximal to the carpal tunnel, at the tunnel inlet, and at the tunnel outlet, with those measurements at the tunnel inlet being the most discerning (Figures 1A–C, green line). The ROC curve illustrates that, by choosing a cross-sectional area of 0.098 cm² at the tunnel inlet to distinguish patients from controls, the diagnostic value of sonography approaches that of electrophysiologic study (12). The findings are consistent with those found in other studies (2–7), as summarized in Figures 2A–C.

Figure 2. Summary of sonographic study findings of the cross-sectional area of the median nerve in carpal tunnel syndrome patients compared with corresponding controls (bars show the means and ranges, in cm²) proximal to the tunnel (A), at the tunnel inlet (B), and at the tunnel outlet (C). Values for each respective control group are represented in gray, the present study in red, Sarria et al (3) in black, Swen et al (2) in green, and Buchberger et al (6) in blue.

This and previous studies (Table 1) indicate that sonography is likely to play an increasingly important role in establishing the diagnosis of CTS. In this respect, objective criteria that include an increase in the median nerve cross-sectional area are being proven to have greater discriminatory power than that of subjective criteria such as nerve echogenicity, shape, or position. Since the median nerve changes in caliber and shape as it passes from the distal forearm to the palm, it is imperative that reliable objective criteria for measurement technique and location are established at the outset, thereby facilitating reliable correlation between the findings in future studies.

Some previous studies (2, 5) have analyzed the median nerve at a single location only (i.e., the tunnel inlet). This is not ideal, because occasionally the nerve caliber may be within normal limits at the inlet and yet unduly swollen proximal to the inlet or at the outlet of the carpal tunnel. We propose that the cross-sectional area of the median nerve be measured immediately proximal to the tunnel inlet, at the tunnel inlet, and at the tunnel outlet. The nerve should also be observed beyond the carpal tunnel to ensure that it is not unduly swollen, although branching of the median nerve beyond the tunnel outlet (4) hinders comparative measurements of the cross-sectional area of the nerve at this location.

Measurements of the median nerve are probably best obtained within, rather than outside, the echogenic perineurium, for 2 reasons. First, the delineation between the hypoechoic nerve fascicles and the hyperechoic perineurium is more clearly defined than is the delineation between the echogenic perineurium and the similarly echogenic perineural fat. Second, it is our impression that the nerve fascicles, rather than the perineurium, become swollen in CTS. The perineurium becomes less apparent when the median nerve swells, probably as a result of the intrinsic compression (Figures 3A and B). Measuring the cross-sectional area of the nerve fascicles alone should therefore be a potentially better discriminator than measuring the nerve fascicles and perineurium in unison. Tracing the margin of the nerve at the time of examination is more accurate than applying the ellipsoid formula (0.25 × height × width × 3.14), since the median nerve is frequently nonellipsoid in outline (5) (Figure 4).

Figure 3. A, Transverse ultrasound examination of the wrist showing an enlarged median nerve (arrows). The perineurium is not visible. The calipers indicate the borders of flexor retinaculum. B, Transverse ultrasound examination of the contralateral wrist showing a nonenlarged median nerve. The echogenic perineurium is seen surrounding the hypoechoic median nerve (arrows). The calipers indicate the borders of flexor retinaculum.

Figure 4. Transverse ultrasound examination of the wrist, showing a nonenlarged median nerve (arrows) with an irregular geographic outline deep to the flexor retinaculum. The perineurium is not visible on this examination.

We analyzed the averaged results of the right and left hand with respect to median nerve cross-sectional area. Almost identical results for the predictive capacity of the median nerve cross-sectional area (which has good reliability; see Table 1) were found when the right and left hands were individually analyzed. Analyzing the right and left hands independently, we revealed statistically significant differences for retinacular bowing and thickness (data not shown). However, because the sensitivity and specificity at the designated thresholds for these measures were only moderate at best and the reliability (Table 2) was low, these criteria are likely to be of doubtful practical significance on an individual patient basis. These findings may help to explain some of the discrepancies found in other studies (Table 1).

We found retinacular bowing at the midtunnel region to be an insensitive discriminatory criterion for CTS, although other investigators have indicated that this is not the case at the tunnel outlet (3, 6). Similarly, flattening of the median nerve was not found to be a useful discriminatory criterion. Instead of being a rigid structure, the median nerve is quite malleable, visibly changing in shape on real-time sonography with movement of the flexor tendons. This feature is reflected in the poor reliability of these measurements. Table 3 summarizes the useful and nonuseful criteria found in this and other studies to date. This should help to improve standardization of sonographic protocols in future CTS studies. CTS patients in this study were found to have a greater BMI than that of the controls. This is in keeping with previous reports showing that an increased BMI predisposes to CTS (13). No consistent correlation between BMI and cross-sectional area of the median nerve was found, and further studies are needed to establish this relationship.

Table 3. Summary of discriminatory and nondiscriminatory criteria found for sonography of the carpal tunnel in carpal tunnel syndrome*

* Techniques used were 1) patient position: wrist flat with fingers semiextended and resting on a hard surface; 2) measurement technique: outline the median nerve inside the echogenic rim (or nerve sheath) and directly trace nerve margin, rather than calculate using the ellipsoid formula.

1. Useful discriminatory criteria

a. Cross-sectional area of median nerve immediately proximal to carpal tunnel

b. Cross-sectional area of median nerve immediately at carpal tunnel inlet

c. Cross-sectional area of median nerve immediately at carpal tunnel outlet

2. Possibly useful discriminatory criteria

a. Cross-sectional area of median nerve branches immediately distal to carpal tunnel

b. Bowing of the carpal tunnel at the tunnel outlet

c. Maximum depth of carpal tunnel

3. Not useful discriminatory criteria

a. Flattening ratio of median nerve at tunnel inlet

b. Flattening ratio of median nerve at tunnel outlet

c. Cross-sectional area of median nerve in distal forearm

d. Thickness of flexor retinaculum

In analyzing the studies to date (Figures 2A–C), there seems to be good consistency between studies with regard to the range and threshold values. Nevertheless, one must realize that median nerve swelling is a continuum and as a result, threshold values are arbitrary. Several studies, including the present study, have shown that there is considerable overlap in the caliber of the median nerve between CTS patients and normal subjects. Since the objective of this study was to define the optimal sonographic diagnostic criteria in CTS patients, only patients with symptoms and confirmatory electrophysiologic changes were included to allow us to maximize the likelihood that our CTS patients actually had CTS. Likewise, asymptomatic controls were purposely chosen to minimize the impact of overlap between the 2 groups and thereby allow a clear delineation of the sonographic diagnostic criteria. The invariable tradeoff of this inclusion criterion is that the results might not be applicable to “ milder” cases of CTS, i.e., those with CTS symptoms but normal electrophysiologic findings. Further studies will be needed to confirm the usefulness of sonography in this group of patients.

The ability of ultrasound to differentiate CTS from non-CTS patients with coexisting hand conditions remains unknown, since such a group of controls is lacking in this study. It would be interesting to investigate the sonographic appearances of the median nerve in CTS patients as compared with non-CTS patients with hand conditions such as trigger finger, given the awareness that a portion of these patients will demonstrate electrophysiologic evidence of CTS (14). An increase in false-positive results will be inevitable against this group of patients.

On this note, one must exercise caution on the use of sonography of the median nerve as a stand-alone diagnostic criterion for CTS, because it might result in a loss of specificity, as previously happened with electrodiagnostic tests for CTS (15). The implication is that sonography of the median nerve is unlikely to solve all of the diagnostic issues in CTS, but its role will need to be further evaluated against not just electrophysiologic studies, but also magnetic resonance imaging (MRI), the other applicable imaging method. Currently, the role of imaging in CTS is not yet defined; the majority of CTS patients do not undergo imaging studies. MRI can better demonstrate the carpal tunnel margins, the retinacular attachments, and nerve edema (16). Nevertheless, the main benefits of ultrasound over MRI in the diagnosis of CTS are the ease of availability, shorter examination time, and reduced cost.