Magnetic resonance imaging findings in respect to carpal tunnel syndrome

Magnetic resonance imaging findings in respect to carpal tunnel syndrome. Scand J Work Environ Health 2003;29(3):189–196. The objective of this systematic review was to examine the technical and diagnostic capability of magnetic resonance imaging (MRI) to identify carpal tunnel syndrome (CTS). Two independent authors retrieved and screened the existing data and extracted the clinical and validity data adhering to predefined inclusion criteria. The MRI methods and findings were analyzed by two experienced radiologists. Disagreements were solved in consensus. The MRI findings of 373 affected wrists in 13 studies were compared with asymptomatic referents or a series of patients with non-CTS wrist pain. Increased T2-signal, cross-sectional area, and flattening of the median nerve inside the carpal tunnel, as well as bowing of the flexor retinaculum, were the most frequently occurring signs in CTS. Reliable assessment of the sensitivity and specificity of certain MRI signs in respect to CTS remained difficult due to study heterogeneity. There is an obvious need for imaging studies in which validated diagnostic criteria are used.

Carpal tunnel syndrome (CTS) is a disabling condition characterized by pain and paresthesias of the hand and forearm, caused by median nerve neuropathy in the wrist. In 1997 the prevalence rate for clinically certain CTS in the Swedish general population was 2.8% for men and 4.6% for women (1). The dominant hand is affected more frequently, and reports of bilateral involvement range from 7% to 80% (2)(3)(4).
Several inflammatory (traumatic synovitis, rheumatoid arthritis), metabolic (diabetes, amyloidosis, hypothyreoidism), and physiological processes (pregnancy, menopause, work-related hypertrophy of muscles and tendons), as well as space-occupying lesions and congenital abnormalities that either narrow the carpal tunnel or increase the pressure within it, can result in entrapment of the median nerve (5). Another possible mechanism is circulatory disturbance in the nerve, whether due to ischemia or venous congestion (6).
The diagnosis of CTS is often based on typical symptoms alone. Tinel's sign and Phalen's test are prone to false-positive and false-negative results (5). From 53% to 98% of patients with clinically diagnosed CTS have abnormalities in nerve conduction tests (7). Imaging techniques had minor importance in the assessment of CTS before the introduction of magnetic resonance imaging (MRI) (8).
Although MRI is not used in the routine diagnosis of CTS, there are some important justifications for an imaging technique able to reveal macroscopic CTS-specific neural and other soft tissue changes and bony details. Ideally, MRI would distinguish cases refractory to conservative treatment at an early stage, enabling patients' more accurate allocation to surgery. In clinical research and compensation decisions, documented images allow for more objective follow-up. Detailed and repeated anatomic imaging may also clarify the pathogenesis of CTS.
Our objective was to review the medical literature systematically to determine what is known about the technical and diagnostic performance of MRI in relation to CTS.

Subjects and methods
All studies in which patients with CTS were imaged with magnetic resonance were included in the review. Excluded were reviews, studies with small samples of less than 10 patients, and studies on CTS due to trauma, rheumatoid arthritis, gout, or tumors only. Where the same or partly the same results were presented in two or more publications, the most recent publication was chosen for the review.
We prepared structured forms for clinical and imaging data extraction, clinical relevance, and internal validity by adapting the recommendations of the Cochrane Methods Working Group on Systematic Review of Screening and Diagnostic Tests (9).
Two independent reviewers (IP, HF) selected the studies first by title, then by abstract, and finally by the whole text. The reference lists of the pertinent articles were screened. The reviewers then extracted the clinical data and assessed the validity and clinical relevance of the studies included. Two radiologists (PT, TV), both with several years of experience with MRI, analyzed the imaging methods and findings independently. The interrater agreement on the assessments of external and internal validity was not measured. Disagreements were solved in consensus.
We scored the external validity criteria of the included studies (documentation of the reference diagnosis and the clinical characteristics of the patients and referents) on a scale from 0 to 9 according to the fulfillment of the following questions: (i) were any of the existing consensus criteria for CTS applied (4, 10-13), (ii) were the symptoms described, (iii) were the clinical findings described, (iv) was a nerve conduction study (NCS) included, (v) was the severity of the disease graded, (vi) were the patient exclusion criteria stated, (vii) were the comorbid conditions stated, (viii) were the hand-loading activities described, and (ix) were any possible invasive treatments of the forearm described? The score for the internal validity of the study method (scale 0-3) was based on the following aspects: (i) blinding of the magnetic resonance image interpretation, (ii) interob-server or intraobserver agreement concerning the image assessment, and (iii) comparability between the cases and referents. Cases and referents were considered comparable in their age and gender distribution if the mean ages or gender ratios of the referents were within a 10% range of the corresponding values of the cases. The two studies (14,15) in which the referents represented a relevant consecutive series of patients with similar symptoms of CTS and the three studies (16)(17)(18) that used the contralateral symptom-free wrist of the patient as reference were scored for comparability.
The MRI method score (0-10) included the description of the following factors: (i) field strength, (ii) coil type, (iii) arm-wrist position, (iv) imaging direction, (v) pulse sequences, (vi) times to repetition, times to echo and flip angle, (vii) number of acquisitions, (viii) slice thickness, (ix) gap between slices, and (x) pixel size or field of view/matrix size. The image quality of the figures was rated as 0=poor, 1=fair, 2=good, or 3=excellent. Agreement concerning the radiologists' ratings of image quality was computed with the intraclass correlation (weighted kappa). Spearman's correlation (rs) was used to correlate the year of publication, magnetic field strength, method scores, number of sequences, and mean image quality with each other. SPSS 10.0 software (SPSS Inc, Chicago, Il, USA) was used.
The reviewers evaluated the following MRI findings: median nerve dimensions, median nerve flattening, median nerve signal intensity, contrast enhancement of the median nerve, bowing of the flexor retinaculum, dimensions of the carpal tunnel, peritendon or synovial pathology, possible anatomic variations, and other pathology. The existence of these signs was tabulated, and the statistically significant difference between the study group and the reference group was noted. If the statistically significant difference was not given in the original article, we used Fisher's exact test for retrospective testing whenever possible. We paid attention to whether quantitative measurements or merely subjective estimations were presented.
The gender distribution and the mean ages, as well as the scores for internal validity, are presented in table 1. The internal validity score correlated positively with the year of publication (rs=0.57, P=0.041). The clinical features of the patients and the means of diagnosing their CTS were described incompletely in all of the original studies (tables 2 and 3). In most studies, the diagnosis of CTS was verified by means of pathological nerve conduction studies. The study of Brahme et al (8), in contrast, included only patients with normal nerve conduction studies. In three studies, the findings of the pa-tients' nerve conduction studies were either pathological or normal.
The MRI findings are summarized in table 4. Five studies applied only qualitative estimations, whereas most studies (8 of 13) used quantitative measurements at least for some of the magnetic resonance parameters (dimensions or signal intensities). Statistically significant differences between the CTS patients and the referents were found for the following median nerve abnormalities: enlargement of the cross-sectional area (4 studies), flattening (3 studies), a combination of enlargement and flattening (1 study), and an increased T2 signal (7 studies). Other statistically significant pathological   findings among the CTS patients included flexor retinaculum bowing (6 studies), synovial swelling (1 study), and the presence of anatomic variants-an absent hamulus, median nerve interposition, and double branching of the median nerve (1 study).
Nine studies gave dichotomous qualitative estimations of the occurrence of the previously mentioned magnetic resonance findings among the patients and referents. We used these data to calculate the sensitivities and specificities. For enlargement of the crosssectional area of the median nerve, the sensitivity was 35% and the specificity was 84%. The corresponding values were 54% and 95% for median nerve flattening, 70% and 93% for flexor retinaculum bowing, and 75% and 66% for an increased T2-signal intensity of the median nerve.
The studies of Bonel et al (14) and Radack et al (15), which had the highest validity scores, used a relevant clinical patient population with various wrist pathologies as their referents. The results of these studies are very similar to the overall results. Similarly, the results of a larger subgroup of studies with internal validity scores of 2 or 3 (5 studies) did not deviate from the average. The three studies with apparently different results (16,17,21) did not represent any special quality category. Their internal validity scores were 1, 2, and 1, with external validity scores of 5, 2 and 3 and MRI method scores of 9, 9 and 7.
High-field MRI equipment (≥1.0 T) was used in 10 studies, and to some extent in one of the remaining three studies. The MRI method was described well, the scores ranging from 3 to 10 and the median being 9 (table 5). The readers agreed on 11 of 13 image quality ratings and disagreed on two; no disagreement was greater than one score (intraclass correlation 0.80). No significant correlations were found between the methodology variables. Table 3. Description of the comorbid conditions, hand loading activities, and invasive treatment prior to the magnetic resonance imaging (MRI) of the patients in the included studies and the total external valitity score of the studies. (NM = not mentioned, NCS = nerve conduction study, CTS = carpal tunnel syndrome) An increase in median nerve cross-section of up 1989 (22,23) arthritis that may have caus-to two-to threefold as it entered carpal tunnel (bilateral) ed CTS (clerk, typist, in the pisiform level indicated CTS the best barber), 3 had unusual upper-limb activity (painting, hammering, racket sports)

Discussion
An increased T2-signal intensity in the median nerve and bowing of the flexor retinaculum seem to be the most sensitive magnetic resonance signs in CTS. Other potential signs are an increased cross-sectional area, flattening of the median nerve, and peritendon pathology.
Previous nonsystematic reviews on MRI in respect to CTS (5,26) found problems similar to ours as regards the design of the original studies: poor description of the reference diagnosis and recruitment of asymptomatic referents only.
There are several classifications or consensus criteria for CTS, including symptoms, clinical findings, and electrophysiological criteria (4,(10)(11)(12)(13). Different diagnostic criteria lead to differences in patient spectra, and such differences certainly have some effect on estimates of test accuracy (27). Therefore we would have expected moreadequate descriptions of the symptoms and signs of the Table 4. Summary of the magnetic resonance imaging (MRI) findings in the included studies. a (+ / + = increased value of the parameter or presence of the sign , -/-= decreased value of the parameter or absence of the sign, + -/ + -= indifferent, * = statistically significant difference when compared with referents as reported in the article or calculated afterwards, (*) = some of the recorded values significant, FR = flexor retinaculum, CT = carpal tunnel) spectrum bias, which tends to increase both the sensitivity and specificity of the test. In this review, the two studies that used contralateral symptom-free hands as reference (16,17) showed no sign of enlargement of the cross-sectional area of the median nerve, a sign that occurs frequently in studies in which healthy volunteers serve as referents. Such findings may be related to constitutional factors or aging rather than to CTS. Another setting, represented by two studies in this review (14,15), is a prospective consecutive series of a relevant clinical population with suspected disease. All of the patients undergo an imaging test and their diagnosis is assessed according to proper consensus criteria during the follow-up. Age, gender, occupation, and other activities alongside the comorbid conditions are all relevant sources of bias when wrists affected with CTS and symptom-free wrists are compared. As long as there is no perfect gold standard for the diagnosis of CTS, any other disease that causes symptoms similar to those of CTS in the forearm can be considered a confounding factor. This kind of study, in which referents represent a clinical population with suspected disease, overcomes these difficulties and can be seen as more relevant to clinical practice.
In presenting the imaging results (table 4), a distinction was drawn between MRI assessments based on quantitative measurements and those based on subjective impressions. This important issue of study validity was not employed in the validity scorings, and therefore may be a weakness of this review. Due to the heterogeneity of the studies in terms of reference diagnosis and type of control group, as well as the small sample size, the values for sensitivity and specificity calculated in this review should be considered with caution. Because of this apparent heterogeneity, we decided not to dichotomize all the data and pool it to obtain a summary receiver operating characteristic curve (28). There are also limitations with the use quality scores. Scoring implies some loss of original information and increases observer bias (29). Another possible shortcoming of this review is the omission of the interrater agreement assessment, which may have given a useful numerical estimation of the subjectivity of the assessment of external and internal validity in the original studies.
The most promising magnetic resonance signs in CTS were an increased T2-signal, an increased crosssectional area, and flattening of the median nerve inside the carpal tunnel, as well as bowing of the flexor retinaculum. The heterogeneity and suboptimal quality of the original studies still make reliable assessment of the sensitivity and specificity of these signs impossible. There is an obvious need for MRI studies that use validated diagnostic criteria for CTS and that describe the hand-loading activities during work or leisure time.