ARTICLES
The Carpal Tunnel Syndrome: Diagnostic Utility of the History and Physical Examination Findings Jeffrey N. Katz, M D ; Martin G. Larson, ScD; Amin Sabra, M D ; Christian Krarup, M D ; Craig R. Stirrat, M D ; Rajesh Sethi, M D ; Holley M. Eaton, RN; Anne H. Fossel; and Matthew H. Liang, M D , M P H
Study Objective: To assess the value of a history and physical examination findings in diagnosing the carpal tunnel syndrome, and to determine whether constellations of clinical findings identify patients at high or low risk for the carpal tunnel syndrome. Design: Comparison of diagnostic tests with neurophysiology testing. Setting: Patients with upper extremity complaints of diverse causes referred to a neurophysiology laboratory for diagnostic studies. Methods: Before nerve conduction testing, a history, demographic and physical examination data, and a hand pain diagram were obtained from each patient. Diagrams were categorized as indicating the classic carpal tunnel syndrome, or as probable, possible, or unlikely to indicate the carpal tunnel syndrome. Associations between clinical data and nerve conduction results were examined in univariate and multivariate analyses. Results: Of 110 patients in the study, 44 (40%) had the carpal tunnel syndrome. Individually, the best predictors were hand pain diagram rating (positive predictive value, 0.59; 95% CI, 0.48 to 0.68) and Tinel sign (positive predictive value, 0.55; CI, 0.45 to 0.65). The combination of a positive Tinel sign and a probable or classic diagram rating had a positive predictive value of 0.71; CI, 0.53 to 0.85. Other findings from physical examination and the history were less useful. Just 9% of patients under 40 years of age with possible or unlikely diagram ratings had the carpal tunnel syndrome. Conclusions: With the exceptions of age, Tinel sign, and hand pain diagram rating, findings from the physical examination and the history had limited diagnostic utility. Patients under 40 years of age with possible or unlikely diagram ratings were at low risk for the carpal tunnel syndrome. This finding, which should be confirmed in an independent population, suggests that subsets of patients may be managed without nerve conduction studies.
Annals of Internal Medicine. 1990;112:321-327. From Robert B. Brigham Multipurpose Arthritis Center; Brigham and Women's Hospital; Harvard Medical School; and Longwood Hand Associates, Boston, Massachusetts. For current author addresses, see end of text.
Ihe carpal tunnel syndrome affects 0.1 % of the general population and as many as 15% of workers in high-risk industries ( 1 , 2 ) . Early diagnosis can lead to effective conservative therapy, with surgical referral reserved for patients unresponsive to wrist splinting, anti-inflammatory drugs, and injection of corticosteroids ( 3 ) . Early recognition might also prompt identification of biomechanical factors in the workplace, such as repetitive microtrauma, which can exacerbate symptoms ( 4 ) . Many patients with the carpal tunnel syndrome present with nocturnal paresthesias in the digits innervated by the median nerve, but atypical presentations are common and present a diagnostic challenge ( 5 ) . The diagnostic gold standard, nerve conduction testing ( 6 ) , costs $150 to $500. The sensitivity of nerve conduction testing is approximately 9 0 % (7, 8); specificity has not been established. Computed tomography ( 9 ) , magnetic resonance imaging (10), vibrometry (11), and thermography (12) have also been used for diagnosing the carpal tunnel syndrome, but expense precludes their use in primary care settings. SemmesWeinstein monofilament testing for threshold pressure sensation is 9 1 % sensitive and 8 0 % specific for the carpal tunnel syndrome (13) but is not widely used because it requires considerable time and expertise. The clinical history and physical examination, including provocative maneuvers, are more appropriate screening tools for the ambulatory setting, but their diagnostic utility is not well established. Atrophy and weakness of the thenar muscles occur in less than 1 5 % of patients with the carpal tunnel syndrome (14). Two-point discrimination remains intact until late in the course of the disease (15) and has a sensitivity of approximately 3 3 % (13). Estimates of the sensitivity of Phalen and Tinel signs vary from 2 5 % to 7 5 % , and specificities are estimated at 7 0 % to 9 0 % (13, 16, 17). Most studies (13, 17) assessing Phalen and Tinel signs and other diagnostic tests have evaluated patients already presumed to have the carpal tunnel syndrome. The only controlled study (13) used asymptomatic subjects as a comparison group. The diagnostic utility of these tests has not been evaluated in patients who have upper extremity symptoms with unclear causes. We assess the value of the history and physical examination in diagnosing the carpal tunnel syndrome in patients with upper extremity complaints due to a wide range of disorders. We also identify constellations of non-invasive clinical findings that define subgroups of patients with particularly high or low probabilities for having the carpal tunnel syndrome.
© 1990 American College of Physicians
Downloaded From: http://annals.org/pdfaccess.ashx?url=/data/journals/aim/19699/ by a University of California San Diego User on 01/30/2017
321
Methods
Table 1. Clinical and Demographic lation (n = 110)
Features of Study
Popu-
Patients Feature Patients referred to the Brigham and Women's Hospital Neurophysiology Laboratory for electrophysiologic evaluation of upper extremity complaints were eligible for our study. Patients under 18 years of age were excluded. Eligible patients were identified by review of the laboratory schedule from January 1988 through June 1988. Examiners were available to evaluate patients two days per week, and patients scheduled on these days were contacted. All patients contacted agreed to participate and gave informed consent. Before neurophysiology testing, patients were evaluated by one of four examiners (a rheumatologist and three research assistants) who administered a questionnaire and did a standardized physical examination. The examiners were thoroughly trained and periodically retrained in clinical testing. Examiners were unaware of the results of nerve conduction tests. The protocol was approved by the Brigham and Women's Hospital Committee for the Protection of Human Subjects. Clinical D a t a The questionnaire included routine questions for demographic data and questions on known risk factors for the carpal tunnel syndrome ( 6 ) , including history of carpal tunnel syndrome, diabetes, thyroid disease, alcohol use, chronic renal disease, inflammatory arthritis of the wrist, and occupation. The duration and character of symptoms were also recorded. Patients indicated on a visual analog pain scale the maximum discomfort they had experienced in the previous week. Patients completed a self-administered hand pain diagram (18) that depicted both hands with dorsal and palmar views. Patients were asked to mark areas on the diagram corresponding to the location of their symptoms and to indicate the quality of their discomfort. Diagrams were later rated by a person who was blinded to the results of the clinical and nerve conduction tests. The diagrams were rated as indicating the classic carpal tunnel syndrome, or as probable, possible, or unlikely to indicate the carpal tunnel syndrome. Moving two-point discrimination (19) was done with electrocardiogram calipers with tips set 4 mm apart. The index and fifth fingertips were stroked five times with either one or two caliper tips. Failure to identify correctly the number of points on two or more strokes was considered abnormal. Thenar atrophy was defined as concavity of the thenar muscle group along the plane parallel to the palm and was scored as present or absent. Strength was assessed by resisted thumb abduction and resisted pinching of the thumb and fifth digit; these measurements were scored as normal or as slightly or markedly impaired. The Tinel sign was done by dropping the square end of a reflex hammer five times on the distal wrist crease from a height of 12 cm. A positive response required pain or paresthesias in at least one finger innervated by the median nerve (20). The Phalen sign was elicited by asking patients to flex both wrists 90 degrees with the dorsal aspects of their hands held in apposition for 60 seconds. A positive response required pain or paresthesias in at least one finger innervated by the median nerve ( 2 1 ) . Test-retest and interrater reliability were 9 1 % and 84% ( 1 8 ) , respectively, for hand diagram ratings and were not studied for physical examination maneuvers. Before nerve conduction testing, the neurologists did their normal evaluations that included an abbreviated history taking and physical examination. Using only this evaluation and blinded to the results of the non-invasive testing protocol and subsequent nerve conduction tests, the neurologists estimated the likelihood of having the carpal tunnel syndrome as probable, possible, or unlikely. N e r v e C o n d u c t i o n Tests The protocol included bilateral median and ulnar sensory and motor testing and electromyographic recording from the
322
Subjects n, %
Male White Duration of symptoms, mo* < 2 2 to 12 > 12 Comorbidity Previous carpal tunnel release Diabetes Thyroid disease Rheumatoid arthritis Work status Disabled or receiving compensation Retired or homemaker Working Electrophysiologic diagnosis The carpal tunnel syndrome Normal Cervical radiculopathy Ulnar neuropathy Other
37, 33.6 82, 78.8 21, 19.6 42, 39.3 44,41.1 15, 13.6 13, 11.8 8,7.2 4,3.6 16, 14.5 30, 27.2 64, 58.2 44, 29, 13, 11, 13,
40.0 26.4 11.8 10.0 11.8
* Data missing for three patients.
abductor pollicus brevis on the most symptomatic hand. (Some patients refused to have electromyography, and some refused nerve conduction testing of asymptomatic hands.) Testing was done with standard techniques (7) on a Disa 1500 (Copenhagen, Denmark) or Teca 42 (Pleasantville, New York) apparatus, with skin temperature maintained at 34° to 37° C. Standard criteria for cut-off values were used ( 7 ) . Diagnosis of the carpal tunnel syndrome was made if patients had median motor latency greater than 4.0 ms, sensory latency greater than 3.7 ms, or sensory velocity less than 50 m/s. Diagnoses of cervical radiculopathy were supported by electromyography. The neurologists were blind to the hand diagram ratings and to the results of physical examinations done by the study investigators.
Analysis The primary outcome variable was the result of nerve conduction studies, expressed as either positive or negative for the carpal tunnel syndrome. Patients receiving the neurophysiology diagnosis of the carpal tunnel syndrome in one or both extremities were classified as positive and others, as negative. Provocative tests and physical findings were considered positive if a positive finding was noted in one or both hands. Thus, the patient, and not the individual hand, served as the unit of analysis in our study. We chose this approach because data from both hands of a person are probably not independent. Additional analyses done with the individual hand rather than the patient as the unit of analysis yielded similar results and are not reported. The sensitivity (proportion of patients with the carpal tunnel syndrome who had positive clinical findings), specificity (proportion of those without the carpal tunnel syndrome who had negative findings), positive predictive value (proportion of those with positive findings who had the carpal tunnel syndrome), and negative predictive value (proportion of those with negative findings who did not have the carpal tunnel syndrome) were calculated for each clinical finding (22). Confidence intervals for the predictive values were calculated with an iterative proportional fitting algorithm assuming a hypergeometric distribution. Predictive values in a hypothetical population with a 15% prevalence of the carpal tunnel syndrome were calculated using Bayes' theorem (22). The odds ratio was chosen as the measure of
1 March 1990 • Annals of Internal Medicine • Volume 112 • Number 5
Downloaded From: http://annals.org/pdfaccess.ashx?url=/data/journals/aim/19699/ by a University of California San Diego User on 01/30/2017
association, because it depends only on the sensitivity and specificity of diagnostic tests and not on the prevalence of disease (23), p. 426). Because there are no published data on the prevalence of the carpal tunnel syndrome among unselected patients with upper extremity complaints, the odds ratio makes our findings more generalizable. The BreslowDay test for the homogeneity of the odds ratio (24), p. 173) was used to evaluate variability among examiners in the associations between clinical tests and nerve conduction results. Statistical analyses were done with the SAS statistical package (25) on an IBM PS 2 Model 60 computer. Tests of association among categorical variables were done with a chi-squared statistic. A significance level of 0.05 was used for each test; no adjustments were made for multiple testing. A logistic regression model (24), p. 192) was developed with the CATMOD program (25) in SAS. Independent variables included age category, unilateral or bilateral symptoms, nocturnal symptoms, Phalen and Tinel signs, and hand diagram rating. Variables not associated with the carpal tunnel syndrome at a significance level of 0.05 were dropped from the model.
Results Baseline Data One hundred and ten patients entered our study. The mean age was 45.6 years (SD, 14.4). Other baseline demographic and clinical features of the patients in our study are summarized in Table 1. All patients with diabetes, thyroid disease, and rheumatoid arthritis were receiving appropriate treatment for these disorders. Of the 110 patients, 5 0 % had discomfort in both upper extremities; 2 5 % , in only the right; and 2 5 % , in only the left. Numbness, pain, and tingling were reported by 7 9 % , 7 8 % , and 8 4 % of all patients, respectively. Seventy-two percent indicated that their symptoms were worse at night or awakened them from sleep. The specific diagnoses for each patient are displayed in Table 1. Forty-four patients ( 4 0 % ) were diagnosed as having the carpal tunnel syndrome in at least one upper extremity.
Variability among Examiners The associations between clinical test results and nerve conduction outcomes were examined separately for each of the four examiners. Using the Breslow-Day test for homogeneity of the odds ratio (24), p. 173), we found no significant differences among examiners in the associations between the carpal tunnel syndrome and the Phalen sign (P = 0.19), Tinel sign (P = 0.31), or two-point discrimination (P = 0.09).
Bivariate Analyses Age was partitioned into three categories: under 40, 40 to 55, and over 55 years of age. The prevalence of the carpal tunnel syndrome in these categories was 2 4 % , 4 4 % , and 5 4 % , respectively (chi-squared = 6.31, 2 df, P = 0.043). Patients presenting with symptoms in both hands (Table 2) had a higher prevalence of the carpal tunnel syndrome than patients with unilateral complaints (chi-squared = 3.79, 1 df, P = 0.05). Gender ( P = 0.46), the duration of symptoms (P = 0.71), and the presence of nocturnal symptoms (P = 0.58) were not significantly associated with diagnosis. Exposure to either vibrating tools, frequent wrist flexion, or repetitive pinching was not significantly associated with the carpal tunnel syndrome (P = 0.93). The prevalence of the carpal tunnel syndrome was 3 8 % in workers, 3 1 % in subjects who were disabled or receiving Workmen's Compensation, and 6 0 % in retirees or homemakers. Slight or marked weakness in thenar muscle strength was detected in 3 6 % of patients, but these findings correlated poorly with diagnosis of the carpal tunnel syndrome (P = 0.62). Thenar atrophy was detected in 10% of patients; only half of these patients had the carpal tunnel syndrome (P = 0.22). Hand diagram ratings were grouped into those indicating the classic carpal tunnel syndrome or probably indicating the carpal tunnel syndrome and those possi-
Table 2. Bivariate Analyses of Clinical Findings in the Carpal Tunnel Syndrome Finding
Tinel sign Phalen sign Sensory loss Hand diagram ratingt Neurologist's assessment:): Age > 40 years Nocturnal symptoms Bilateral symptoms
Sensitivity Specificity
Odds Ratio (95% CI)*
In our!Sample with 40% :Prevalence Negative Positive Predictive Value Predictive Value (95% CI) (95% CI)
In Hypothetical Sample \vith 15% Prevj ilence Positive Negative Predictive Predictive Value Value
0.60 0.75 0.32
0.67 0.47 0.81
3.13 (1.40 to 7.00) 2.65(1.10 to 6.20) 1.98(0.81 to 4.55)
0.55(0.45 to 0.65) 0.48(0.40 to 0.54) 0.54(0.37 to 0.70)
0.72(0.63 to 0.79) 0.74(0.62 to 0.84) 0.63(0.58 to 0.68)
0.25 0.20 0.23
0.91 0.91 0.87
0.61
0.71
3.92(1.75 to 8.77)
0.59(0.48 to 0.68)
0.73(0.66 to 0.80)
0.27
0.91
0.84 0.80
0.72 0.42
13.16(4.95 to 34.48) 2.87(1.19 to 6.92)
0.67(0.58 to 0.73) 0.48(0.41 to 0.53)
0.87(0.78 to 0.93) 0.76(0.63 to 0.86)
0.34 0.20
0.96 0.92
0.77
0.28
1.29(0.33 to 4.01)
0.42(0.36 to 0.47)
0.64(0.48 to 0.78)
0.16
0.87
0.61
0.58
2.16(0.98 to 4.72)
0.49(0.40 to 0.58)
0.69(0.60 to 0.78)
0.20
0.89
* Relative odds for the carpal tunnel syndrome with 9 5 % CI. t Probable or classic compared with possible or unlikely hand diagram rating. t Probable or possible compared with unlikely neurologist's assessment.
1 March 1990 • Annals ofInternal Medicine • Volume 112 • Number 5
Downloaded From: http://annals.org/pdfaccess.ashx?url=/data/journals/aim/19699/ by a University of California San Diego User on 01/30/2017
323
Table 3. Bivariate Analyses Finding
Phalen or Tinel signs Phalen and Tinel signs Tinel sign or hand diagram ratingt Tinel sign and hand diagram rating! Phalen sign or hand diagram ratingf Phalen sign and hand diagram ratingf
of Combinations
Sensitivity Specificity
of Clinical Findings in the Carpal Tunnel
Syndrome
Odds Ratio (95% CI)*
In our Sample with 4 0 % Prevalence Positive Negative Predictive Value Predictive Value (95% CI) (95% CI)
In Hypothetical Sample with 1 5 % Prevalence Positive Negative p r edictive Predictive Value Value
0.88
0.41
5.21 (1.80 to 14.92)
0.50 (0.44 to 0.54) 0.84 (0.70 to 0.93)
0.21
0.95
0.46
0.73
2.34 (1.02 to 5.35)
0.53 (0.40 to 0.65) 0.68 (0.61 to 0.74)
0.23
0.89
0.82
0.50
4.50 (1.82 to 1.11)
0.52 (0.45 to 0.57) 0.81 (0.69 to 0.89)
0.22
0.94
0.39
0.89
5.31 (1.97 to 14.30)
0.71 (0.53 to 0.85) 0.69 (0.64 to 0.72)
0.39
0.89
0.82
0.38
2.74 (1.10 to 6.85)
0.47 (0.41 to 0.5) 0.76 (0.62 to 0.87)
0.19
0.92
0.49
0.83
4.77 (1.97 to 11.52)
0.66 (0.51 to 0.78) 0.71 (0.65 to 0.77)
0.34
0.90
* Relative odds for the carpal tunnel syndrome with 9 5 % CI. t Probable or classic compared with possible or unlikely hand diagram rating.
bly indicating or unlikely to indicate the carpal tunnel syndrome. Similarly, the neurologists' assessments were categorized as those probably or possibly indicating the carpal tunnel syndrome and those unlikely to indicate the carpal tunnel syndrome. As shown in Table 2, the hand diagram ratings (positive predictive value, 0.59; 9 5 % CI, 0.43 to 0.68), and neurologists' assessments (positive predictive value, 0.67; CI, 0.58 to 0.73) were better predictors of the carpal tunnel syndrome than Phalen or Tinel signs. The prevalence or pretest probability of the carpal tunnel syndrome in our sample was 0.40. If the positive predictive value of a test is not substantially greater than the sample prevalence, the test is of little value in distinguishing outcomes among subjects with a positive test. Thus, the positive predictive values associated with Tinel and Phalen signs and two-point discrimination (Table 2) are quite modest. Similarly, if the negative predictive value of a test is less than the prevalence of subjects without the carpal tunnel syndrome (0.60), the test is of little value in distinguishing outcomes among subjects with negative test results. The neurologists' assessments had the highest negative predictive value (0.87; CI, 0.78 to 0.93), followed by Tinel sign, Phalen sign, and the hand diagram ratings. In a hypothetical population whose prevalence of the carpal tunnel syndrome was 1 5 % , the positive predictive values would be much lower and the negative predictive values would be higher. For example, the positive predictive value of Tinel sign would be 0.25 and the negative predictive value would be 0.91. Predictive values of other findings in a population with a prevalence of 1 5 % are shown in Table 2. We assessed the diagnostic utility of tests used in various combinations; the results are shown in Table 3. Requiring that just one of a pair of tests be positive raises sensitivity but sacrifices specificity; similarly, requiring that both tests be positive raises specificity but
324
lowers sensitivity. The combination of a positive Tinel and a probable or classic diagram rating was very specific (0.89) and had a higher positive predictive value (0.71; CI, 0.53 to 0.85) than either test used singly. The same was true of a positive Phalen and a probable or classic diagram rating. The combination of a positive Phalen or Tinel was very sensitive (0.88) and had a high negative predictive value (0.84; CI, 0.70 to 0.93). Table 3 shows that the positive predictive values of these combinations would be lower and the negative predictive values higher in a population with a lower prevalence of the carpal tunnel syndrome. Multivariate Analysis In the logistic regression analysis, nocturnal symptoms, unilateral (as compared with bilateral) complaints, and Phalen sign were not associated with the carpal tunnel syndrome at a significance level of 0.05 and were dropped from the model. A term was added to account for a suspected interaction between hand diagram rating and age. The final model included hand diagram rating (P = 0.001), Tinel sign (P = 0.01), and the interaction between age category and hand diagram rating (P = 0.06). The adjusted odds ratios from the logistic regression are presented in Table 4. The hand pain diagram rating was more predictive of the carpal tunnel syndrome in the youngest age group (odds ratio, 12.5; CI, 1.95 to 80.58) than in the middle (odds ratio, 1.19; CI, 0.35 to 4.11) and oldest age groups (odds ratio, 8.10; CI, 1.23 to 53.28). We evaluated the independent contributions of the neurologists' assessments and the clinical findings (age, hand diagram rating, and Tinel sign) by comparing multivariate models containing clinical findings, the neurologists' assessments, or both. The log likelihood statistic was used to test for significance
1 March 1990 • Annals ofInternal Medicine • Volume 112 • Number 5
Downloaded From: http://annals.org/pdfaccess.ashx?url=/data/journals/aim/19699/ by a University of California San Diego User on 01/30/2017
(24, p. 205). The neurologists' assessments were added to the model comprised of clinical findings with significant improvement in fit {P < 0.001). Age category and Tinel sign no longer contributed significantly to the model when the assessment was added, but the hand diagram rating remained a significant predictor of the outcome ( P = 0.02). Addition of the clinical findings to a model containing the neurologists' assessments also significantly improved the fit (P < 0.05). Table 5 shows the risk for the carpal tunnel syndrome in patients stratified by age category, hand diagram rating, and Tinel sign. In general, the risk for the carpal tunnel syndrome increases with the presence of additional positive findings. The table identifies strata associated with particularly low and high risks for the carpal tunnel syndrome. Patients under 40 years of age with possible or unlikely hand diagram ratings (first two strata) had an 8.7% ( 2 / 2 3 ) prevalence of the carpal tunnel syndrome, whereas patients over 55 years of age with positive Tinel signs and probable or classic hand diagram ratings had an 88.9% ( 8 / 9 ) prevalence of the carpal tunnel syndrome.
Table 5. Prevalence of the Carpal Tunnel Syndrome in Subgroups of Subjects Defined by Particular Clinical Features
Discussion
that included pregnant patients with hand symptoms. The most informative data from the history appeared to be the anatomic localization of symptoms. A self-administered hand pain diagram, on which patients indicated the location and quality of their complaints, had a higher predictive value (0.59; CI, 0.48 to 0.68) than any provocative test or physical examination finding. The diagram was more predictive of the carpal tunnel syndrome in patients under 40 years of age (odds ratio, 12.5) than in the middle and oldest age groups (odds ratios, 1.2 and 8.1, respectively). This finding was not explained by variation in the distribution of diagnoses across age categories. Older patients may have had more difficulty completing the diagram accurately. Neurologists with experience in evaluating upper extremity disorders identified patients at low, moderate, and high risk for the carpal tunnel syndrome with impressive accuracy. In logistic regression analyses, addition of the neurologists' assessments to a model containing age, Tinel sign, the hand diagram rating, and the age-diagram interaction significantly improved the fit, indicating that the neurologists' assessments incorporated diagnostically valuable information that was not captured by the clinical tests. Similarly, addition of the clinical tests to a model containing the neurologists' assessments significantly improved the fit, indicating that the clinical tests added important information to the neurologists' assessments. The hand diagram rating was the only clinical test that contributed significantly to the model that included clinical tests and the neurologists' assessments. These data suggest that although experienced clinicians use information not captured by our study variables, experienced clinicians' assessments might be improved by using the results of clinical tests, especially hand diagram ratings. Using combinations of tests may increase the diagnostic value of the tests (27, p. 41). The presence of
Previous reviews (6, 21) of the carpal tunnel syndrome have emphasized the diagnostic utility of provocative tests, bilaterality of symptoms, and nocturnal paresthesias, but our data suggest that these findings are of limited value. Specifically, nocturnal symptoms were not significantly associated with the carpal tunnel syndrome, bilateral involvement had little independent predictive value in the multivariate analysis, and physical findings traditionally associated with the carpal tunnel syndrome, including Phalen and Tinel signs, two-point discrimination, and thenar muscle atrophy and weakness, were of limited utility. Of these tests, Tinel sign had the highest positive predictive value (0.55; CI, 0.45 to 0.65) and Phalen sign had the highest negative predictive value (0.74; CI, 0.62 to 0.84). We found that age is also a risk factor for the carpal tunnel syndrome, in agreement with other studies (26). This finding would be less striking in a sample Table 4. Adjusted Odds Ratios for the Carpal Tunnel drome Associated with Specific Clinical Features
Syn-
Clinical Feature
CI)*
Age Categoryt, y Under 40 40 to 55 Over 55 Positive Tinel signj Probable or classic hand diagram rating§ Entire sample Patients under 40 Patients 40 to 55 Patients over 55
Odds Ratio (95% 1.00 2.82 3.74 3.11 4.95 12.55 1.19 8.10
(not applicable) (0.91 to 8.69) (0.99 to 14.13) (1.26 to 7.70) (1.87 (1.95 (0.35 (1.23
to to to to
13.08) 80.58) 4.11) 53.28)
* 9 5 % CI = 9 5 % confidence interval for the relative odds for the carpal tunnel syndrome. t Reference age category is under 40 years of age. % Compared with negative Tinel sign. § Compared with possible or unlikely hand diagram rating.
Tinel Sign*
_ + — — — + — + — + + +
Total
Patients in Subgroup Number Percentage with Hand Diagram Age Ratingf Category} the Carpal Tunnel Syndrome
_ — — — + — + + + — + +
1 1 3 2 1 3 2 2 3 2 1 3
13 10 9 16 6 3 11 8 5 10 7 9 107§
8 10 22 25 33 33 45 50 60 70 71 89
* — = negative, + = positive Tinel sign. t — = possible or unlikely, + = probable or classic hand diagram rating. %\ = under 40, 2 = 40 to 55, 3 = over 55. § Data missing for three patients.
1 March 1990 • Annals of Internal Medicine • Volume 112 • Number 5
Downloaded From: http://annals.org/pdfaccess.ashx?url=/data/journals/aim/19699/ by a University of California San Diego User on 01/30/2017
325
both a positive Tinel sign and a probable or classic hand diagram rating was associated with impressive specificity (0.89) and high positive predictive value (0.71; CI, 0.53 to 0.85). The presence of a positive Phalen sign and a probable or classic hand diagram rating was also valuable (Table 3 ) . The combinations of Phalen or Tinel signs and a Tinel sign or a positive hand diagram rating were nonspecific but had high sensitivity and negative predictive value. These data illustrate that the optimal choice of tests or combination of tests depends on whether the goal is to establish diagnosis of the carpal tunnel syndrome or to exclude it. Predictive values vary with the prevalence of disease. Table 2 indicates that in a sample with a 1 5 % prevalence of the carpal tunnel syndrome, as might be encountered in a high-risk industry ( 2 ) , the positive predictive values of non-invasive tests would be just 0.20 to 0.30, and the negative predictive values would be approximately 0.90. This finding has important implications for occupational screening. Most asymptomatic workers screened with these non-invasive tests will not have the carpal tunnel syndrome regardless of test results. The tests are valuable in diagnosing the syndrome only if the previous probability of the carpal tunnel syndrome is substantially higher than the baseline prevalence, as would be expected in symptomatic workers. An important goal of our study was to determine whether specific constellations of clinical signs and symptoms could identify patients with particularly high or low probabilities for having the carpal tunnel syndrome. In particular, identifying patients who could be managed without nerve conduction tests would be valuable. As shown in Table 5, risk categories can be defined by age, Tinel sign, and hand diagram rating. Patients under 40 years of age with possible or unlikely hand diagram ratings had a low (8.7) likelihood of having the carpal tunnel syndrome. Although the low-risk group included only 23 patients, this stratum would likely account for a larger proportion in a less select population. This clinically defined approach to identifying low-risk patients should be validated in an independent sample. If confirmed, these findings would support the practice of excluding the carpal tunnel syndrome without doing nerve conduction tests in selected patients. Since the carpal tunnel syndrome is neither fatal nor rapidly progressive, clinicians would probably accept an 8% to 9 % false negative rate for this potential prediction rule. Our study has two potential limitations: First, patients referred to a hospital-based laboratory may have atypical findings or multiple causes for their symptoms, lowering the specificity of non-invasive tests. Referred patients might also have the carpal tunnel syndrome particularly severely, raising sensitivity (28). Therefore, our results must be generalized to the primary care setting cautiously. Second, using nerve conduction tests to define the carpal tunnel syndrome might be questioned, because their sensitivity is approximately 9 0 % ( 8 ) ; moreover, their accuracy may vary among laboratories and neurophysiologists.
326
However, no alternative tests have been shown to be comparable or superior. Further research might be done in the occupational setting or with outpatients presenting de novo to primary providers to determine whether the performance characteristics of these tests change significantly in different populations. It would be particularly valuable to confirm that non-invasive clinical findings can identify a subgroup of patients at particularly low risk for the carpal tunnel syndrome who can be managed effectively without nerve conduction studies. Acknowledgments: The authors thank E. Francis Cook, SD, Lee Goldman, M D , and David Bates, M D , for their contributions; Barbara Shulman, RPT, Susan Steindorf, RPT, and Lisa Bail for doing clinical evaluations; Patricia Butler and Margaret Mills for scheduling patients; and Mari Garcia and Jacqueline Mazzie for manuscript preparation. Grant Support: By N I H Grants AR36308 and AR07530 and the Kellogg Program for Training in Research in Clinical Effectiveness. Requests for Reprints: Jeffrey N. Katz, MD, Arthritis Center, Brigham and Women's Hospital, 75 Francis Street, Boston, MA 02115. Current Author Addresses: Dr. Katz: Arthritis Center, Brigham and Women's Hospital, 75 Francis Street, Boston, MA 02115. Dr. Liang: Department of Rheumatology-Immunology, Brigham and Women's Hospital, 75 Francis Street, Boston, MA 02115. Dr. Larson, Ms. Eaton, and Ms. Fossel: Robert B. Brigham Multipurpose Arthritis Center, 221 Longwood Avenue, Boston, MA 02115. Dr. Stirrat: Longwood Hand Center, 830 Boylston Street, Chestnut Hill, MA 02167. Dr. Sethi: 317 Patrick Street Southwest, Vienna, VA 22180. Dr. Sabra: Department of Medicine, New England Baptist Hospital, 125 Parker Hill Avenue, Boston, MA 02120. Dr. Krarup: Department of Neurophysiology, Rigshospitalet, Blegdamsvej 9, 2100 Copenhagen, Denmark.
References 1. The National Ambulatory Medical Care Survey. Summary, United States. Rockville, Maryland: U.S. Department of Health, Education, and Welfare, Public Health Service, Health Resources Administration, National Center for Health Statistics; 1975: D H E W No. ( H R A ) 76-1772. 2. Masear VR, Hayes JM, Hyde AG. An industrial cause of carpal tunnel syndrome. J Hand Surg [Am]. 1986;11:222-7. 3. Gelberman RH, Aronson D, Weisman MH. Carpal tunnel syndrome. Results of a prospective trial of steroid injection and splinting. / Bone Joint Surg [Am]. 1980;62:1181-4. 4. Armstrong TJ. An Ergonomics Guide to Carpal Tunnel Syndrome. Akron, Ohio: American Industrial Hygiene Association; 1983. 5. Hoffman DE. Carpal tunnel syndrome. Importance of sensory nerve conduction studies in diagnosis. JAMA. 1975;233:983-4. 6. Dorwart BB. Carpal tunnel syndrome: a review. Semin Arthritis Rheum. 1984;14:134-40. 7. Stevens J C . A A E E minimonograph # 2 6 : the electrodiagnosis of carpal tunnel syndrome. Muscle Nerve. 1987;10:99-113. 8. Grundberg AB. Carpal tunnel decompression in spite of normal electromyography. J Hand Surg [Am]. 1983;8:348-9. 9. Bleeker ML. Medical surveillance for carpel tunnel syndrome in workers. / Hand Surg. 1987;12:845-8. 10. Weiss KL, Beltran J, Lubbers LM. High-fold MR surface-coil imaging of the hand and wrist. Part II. Pathologic correlations and clinical relevance. Radiology. 1986;160:147-52. 11. Lundberg G, Lie-Stenstrom AK, Sollerman C, Strombers T, Pykko I. Digital vibrogram: a new diagnostic tool for sensory testing in compression neuropathy. J Hand Surg [Am]. 1986;11:693-9. 12. Herrick RT, Herrick SK. Thermography in the detection of carpal tunnel syndrome and other compressive neuropathies. / Hand Surg [Am]. 1987;12:943-9. 13. Gellman H, Gelberman RH, Tan AM, Botte MJ. Carpal tunnel syndrome. An evaluation of the provocative diagnostic tests. / Bone Joint Surg [Am]. 1986;68:735-7. 14. Posch JL, Marcotte DR. Carpal tunnel syndrome: an analysis of 1201 cases. Orthop Rev. 1976;5:25-34. 15. Gelberman RH, Szabo RM, Williamson RV, Dimick MP. Sensibility testing in peripheral-nerve compression syndromes. An experimental study in humans. J Bone Joint Surg [Am]. 1983;65:632-8. 16. Golding DN, Rose DM, Selvarajah K. Clinical tests for carpal tunnel syndrome: an evaluation. Br J Rheumatol. 1986;25:388-90. 17. Seror P . Electroclinical correlations in the carpal tunnel syndrome.
1 March 1990 • Annals ofInternal Medicine • Volume 112 • Number 5
Downloaded From: http://annals.org/pdfaccess.ashx?url=/data/journals/aim/19699/ by a University of California San Diego User on 01/30/2017
Apropos of 100 cases. Rev Rhum Mai Osteoartic. 1987;54:643-8. 18. Katz JN, Stirrat CS. A self-administered hand diagram for the diagnosis of carpal tunnel syndrome. / Hand Surg [Am]. 1989; 14[In press]. 19. Dellon AL. The moving two-point discrimination test: clinical evaluation of the quickly adapting fiber/receptor system. / Hand Surg [Am]. 1978;3:474-81. 20. Mossman SS, Blau JN. TinePs sign and the carpal tunnel syndrome. Br Med J [Clin Res]. 1987;294:680. 21. Phalen GS. The carpal tunnel syndrome. Seventeen years' experience in diagnosis and treatment of six hundred fifty-four hands. / Bone Joint Surg [Am]. 1966;48:211-28. 22. McNeil BJ, Keeler E, Adelstein SJ. Primer on certain elements of medical decision making. N Engl J Med. 1975;293:211-5. 23. Feinstein AR. Clinical Epidemiology: The Architecture of Clinical
Research. Philadelphia: W.B. Saunders Company; 1985. 24. Davis W, Breslow NE, Day NE. Statistical Methods in Cancer Research. 1. Lyon: International Agency for Research on Cancer; 1980. 25. SAS Institute, Inc. SAS/STAT User's Guide, release 6.03 ed. Cary, North Carolina: SAS Institute, Inc.; 1988:189-282. 26. Stevens JC, Sun S, Beard CM, O'Fallon WM, Kurland LT. Carpal tunnel syndrome in Rochester, Minnesota, 1961 to 1980. Neurology. 1988;38:134-8. 27. Galen RS, Gambino SR. Beyond Normality: The Predictive Value and Efficiency of Medical Diagnoses. New York: John Wiley and Sons; 1975. 28. Ransohoff DF, Feinstein AR. Problems of spectrum and bias in evaluating the efficacy of diagnostic tests. AT Engl J Med. 1978;229:926-30.
1 March 1990 • Annals of Internal Medicine • Volume 112 • Number 5
Downloaded From: http://annals.org/pdfaccess.ashx?url=/data/journals/aim/19699/ by a University of California San Diego User on 01/30/2017
327
The Carpal Tunnel Syndrome: Diagnostic Utility of the History and Physical Examination Findings Jeffrey N. Katz, M D ; Martin G. Larson, ScD; Amin Sabra, M D ; Christian Krarup, M D ; Craig R. Stirrat, M D ; Rajesh Sethi, M D ; Holley M. Eaton, RN; Anne H. Fossel; and Matthew H. Liang, M D , M P H
Study Objective: To assess the value of a history and physical examination findings in diagnosing the carpal tunnel syndrome, and to determine whether constellations of clinical findings identify patients at high or low risk for the carpal tunnel syndrome. Design: Comparison of diagnostic tests with neurophysiology testing. Setting: Patients with upper extremity complaints of diverse causes referred to a neurophysiology laboratory for diagnostic studies. Methods: Before nerve conduction testing, a history, demographic and physical examination data, and a hand pain diagram were obtained from each patient. Diagrams were categorized as indicating the classic carpal tunnel syndrome, or as probable, possible, or unlikely to indicate the carpal tunnel syndrome. Associations between clinical data and nerve conduction results were examined in univariate and multivariate analyses. Results: Of 110 patients in the study, 44 (40%) had the carpal tunnel syndrome. Individually, the best predictors were hand pain diagram rating (positive predictive value, 0.59; 95% CI, 0.48 to 0.68) and Tinel sign (positive predictive value, 0.55; CI, 0.45 to 0.65). The combination of a positive Tinel sign and a probable or classic diagram rating had a positive predictive value of 0.71; CI, 0.53 to 0.85. Other findings from physical examination and the history were less useful. Just 9% of patients under 40 years of age with possible or unlikely diagram ratings had the carpal tunnel syndrome. Conclusions: With the exceptions of age, Tinel sign, and hand pain diagram rating, findings from the physical examination and the history had limited diagnostic utility. Patients under 40 years of age with possible or unlikely diagram ratings were at low risk for the carpal tunnel syndrome. This finding, which should be confirmed in an independent population, suggests that subsets of patients may be managed without nerve conduction studies.
Annals of Internal Medicine. 1990;112:321-327. From Robert B. Brigham Multipurpose Arthritis Center; Brigham and Women's Hospital; Harvard Medical School; and Longwood Hand Associates, Boston, Massachusetts. For current author addresses, see end of text.
Ihe carpal tunnel syndrome affects 0.1 % of the general population and as many as 15% of workers in high-risk industries ( 1 , 2 ) . Early diagnosis can lead to effective conservative therapy, with surgical referral reserved for patients unresponsive to wrist splinting, anti-inflammatory drugs, and injection of corticosteroids ( 3 ) . Early recognition might also prompt identification of biomechanical factors in the workplace, such as repetitive microtrauma, which can exacerbate symptoms ( 4 ) . Many patients with the carpal tunnel syndrome present with nocturnal paresthesias in the digits innervated by the median nerve, but atypical presentations are common and present a diagnostic challenge ( 5 ) . The diagnostic gold standard, nerve conduction testing ( 6 ) , costs $150 to $500. The sensitivity of nerve conduction testing is approximately 9 0 % (7, 8); specificity has not been established. Computed tomography ( 9 ) , magnetic resonance imaging (10), vibrometry (11), and thermography (12) have also been used for diagnosing the carpal tunnel syndrome, but expense precludes their use in primary care settings. SemmesWeinstein monofilament testing for threshold pressure sensation is 9 1 % sensitive and 8 0 % specific for the carpal tunnel syndrome (13) but is not widely used because it requires considerable time and expertise. The clinical history and physical examination, including provocative maneuvers, are more appropriate screening tools for the ambulatory setting, but their diagnostic utility is not well established. Atrophy and weakness of the thenar muscles occur in less than 1 5 % of patients with the carpal tunnel syndrome (14). Two-point discrimination remains intact until late in the course of the disease (15) and has a sensitivity of approximately 3 3 % (13). Estimates of the sensitivity of Phalen and Tinel signs vary from 2 5 % to 7 5 % , and specificities are estimated at 7 0 % to 9 0 % (13, 16, 17). Most studies (13, 17) assessing Phalen and Tinel signs and other diagnostic tests have evaluated patients already presumed to have the carpal tunnel syndrome. The only controlled study (13) used asymptomatic subjects as a comparison group. The diagnostic utility of these tests has not been evaluated in patients who have upper extremity symptoms with unclear causes. We assess the value of the history and physical examination in diagnosing the carpal tunnel syndrome in patients with upper extremity complaints due to a wide range of disorders. We also identify constellations of non-invasive clinical findings that define subgroups of patients with particularly high or low probabilities for having the carpal tunnel syndrome.
© 1990 American College of Physicians
Downloaded From: http://annals.org/pdfaccess.ashx?url=/data/journals/aim/19699/ by a University of California San Diego User on 01/30/2017
321
Methods
Table 1. Clinical and Demographic lation (n = 110)
Features of Study
Popu-
Patients Feature Patients referred to the Brigham and Women's Hospital Neurophysiology Laboratory for electrophysiologic evaluation of upper extremity complaints were eligible for our study. Patients under 18 years of age were excluded. Eligible patients were identified by review of the laboratory schedule from January 1988 through June 1988. Examiners were available to evaluate patients two days per week, and patients scheduled on these days were contacted. All patients contacted agreed to participate and gave informed consent. Before neurophysiology testing, patients were evaluated by one of four examiners (a rheumatologist and three research assistants) who administered a questionnaire and did a standardized physical examination. The examiners were thoroughly trained and periodically retrained in clinical testing. Examiners were unaware of the results of nerve conduction tests. The protocol was approved by the Brigham and Women's Hospital Committee for the Protection of Human Subjects. Clinical D a t a The questionnaire included routine questions for demographic data and questions on known risk factors for the carpal tunnel syndrome ( 6 ) , including history of carpal tunnel syndrome, diabetes, thyroid disease, alcohol use, chronic renal disease, inflammatory arthritis of the wrist, and occupation. The duration and character of symptoms were also recorded. Patients indicated on a visual analog pain scale the maximum discomfort they had experienced in the previous week. Patients completed a self-administered hand pain diagram (18) that depicted both hands with dorsal and palmar views. Patients were asked to mark areas on the diagram corresponding to the location of their symptoms and to indicate the quality of their discomfort. Diagrams were later rated by a person who was blinded to the results of the clinical and nerve conduction tests. The diagrams were rated as indicating the classic carpal tunnel syndrome, or as probable, possible, or unlikely to indicate the carpal tunnel syndrome. Moving two-point discrimination (19) was done with electrocardiogram calipers with tips set 4 mm apart. The index and fifth fingertips were stroked five times with either one or two caliper tips. Failure to identify correctly the number of points on two or more strokes was considered abnormal. Thenar atrophy was defined as concavity of the thenar muscle group along the plane parallel to the palm and was scored as present or absent. Strength was assessed by resisted thumb abduction and resisted pinching of the thumb and fifth digit; these measurements were scored as normal or as slightly or markedly impaired. The Tinel sign was done by dropping the square end of a reflex hammer five times on the distal wrist crease from a height of 12 cm. A positive response required pain or paresthesias in at least one finger innervated by the median nerve (20). The Phalen sign was elicited by asking patients to flex both wrists 90 degrees with the dorsal aspects of their hands held in apposition for 60 seconds. A positive response required pain or paresthesias in at least one finger innervated by the median nerve ( 2 1 ) . Test-retest and interrater reliability were 9 1 % and 84% ( 1 8 ) , respectively, for hand diagram ratings and were not studied for physical examination maneuvers. Before nerve conduction testing, the neurologists did their normal evaluations that included an abbreviated history taking and physical examination. Using only this evaluation and blinded to the results of the non-invasive testing protocol and subsequent nerve conduction tests, the neurologists estimated the likelihood of having the carpal tunnel syndrome as probable, possible, or unlikely. N e r v e C o n d u c t i o n Tests The protocol included bilateral median and ulnar sensory and motor testing and electromyographic recording from the
322
Subjects n, %
Male White Duration of symptoms, mo* < 2 2 to 12 > 12 Comorbidity Previous carpal tunnel release Diabetes Thyroid disease Rheumatoid arthritis Work status Disabled or receiving compensation Retired or homemaker Working Electrophysiologic diagnosis The carpal tunnel syndrome Normal Cervical radiculopathy Ulnar neuropathy Other
37, 33.6 82, 78.8 21, 19.6 42, 39.3 44,41.1 15, 13.6 13, 11.8 8,7.2 4,3.6 16, 14.5 30, 27.2 64, 58.2 44, 29, 13, 11, 13,
40.0 26.4 11.8 10.0 11.8
* Data missing for three patients.
abductor pollicus brevis on the most symptomatic hand. (Some patients refused to have electromyography, and some refused nerve conduction testing of asymptomatic hands.) Testing was done with standard techniques (7) on a Disa 1500 (Copenhagen, Denmark) or Teca 42 (Pleasantville, New York) apparatus, with skin temperature maintained at 34° to 37° C. Standard criteria for cut-off values were used ( 7 ) . Diagnosis of the carpal tunnel syndrome was made if patients had median motor latency greater than 4.0 ms, sensory latency greater than 3.7 ms, or sensory velocity less than 50 m/s. Diagnoses of cervical radiculopathy were supported by electromyography. The neurologists were blind to the hand diagram ratings and to the results of physical examinations done by the study investigators.
Analysis The primary outcome variable was the result of nerve conduction studies, expressed as either positive or negative for the carpal tunnel syndrome. Patients receiving the neurophysiology diagnosis of the carpal tunnel syndrome in one or both extremities were classified as positive and others, as negative. Provocative tests and physical findings were considered positive if a positive finding was noted in one or both hands. Thus, the patient, and not the individual hand, served as the unit of analysis in our study. We chose this approach because data from both hands of a person are probably not independent. Additional analyses done with the individual hand rather than the patient as the unit of analysis yielded similar results and are not reported. The sensitivity (proportion of patients with the carpal tunnel syndrome who had positive clinical findings), specificity (proportion of those without the carpal tunnel syndrome who had negative findings), positive predictive value (proportion of those with positive findings who had the carpal tunnel syndrome), and negative predictive value (proportion of those with negative findings who did not have the carpal tunnel syndrome) were calculated for each clinical finding (22). Confidence intervals for the predictive values were calculated with an iterative proportional fitting algorithm assuming a hypergeometric distribution. Predictive values in a hypothetical population with a 15% prevalence of the carpal tunnel syndrome were calculated using Bayes' theorem (22). The odds ratio was chosen as the measure of
1 March 1990 • Annals of Internal Medicine • Volume 112 • Number 5
Downloaded From: http://annals.org/pdfaccess.ashx?url=/data/journals/aim/19699/ by a University of California San Diego User on 01/30/2017
association, because it depends only on the sensitivity and specificity of diagnostic tests and not on the prevalence of disease (23), p. 426). Because there are no published data on the prevalence of the carpal tunnel syndrome among unselected patients with upper extremity complaints, the odds ratio makes our findings more generalizable. The BreslowDay test for the homogeneity of the odds ratio (24), p. 173) was used to evaluate variability among examiners in the associations between clinical tests and nerve conduction results. Statistical analyses were done with the SAS statistical package (25) on an IBM PS 2 Model 60 computer. Tests of association among categorical variables were done with a chi-squared statistic. A significance level of 0.05 was used for each test; no adjustments were made for multiple testing. A logistic regression model (24), p. 192) was developed with the CATMOD program (25) in SAS. Independent variables included age category, unilateral or bilateral symptoms, nocturnal symptoms, Phalen and Tinel signs, and hand diagram rating. Variables not associated with the carpal tunnel syndrome at a significance level of 0.05 were dropped from the model.
Results Baseline Data One hundred and ten patients entered our study. The mean age was 45.6 years (SD, 14.4). Other baseline demographic and clinical features of the patients in our study are summarized in Table 1. All patients with diabetes, thyroid disease, and rheumatoid arthritis were receiving appropriate treatment for these disorders. Of the 110 patients, 5 0 % had discomfort in both upper extremities; 2 5 % , in only the right; and 2 5 % , in only the left. Numbness, pain, and tingling were reported by 7 9 % , 7 8 % , and 8 4 % of all patients, respectively. Seventy-two percent indicated that their symptoms were worse at night or awakened them from sleep. The specific diagnoses for each patient are displayed in Table 1. Forty-four patients ( 4 0 % ) were diagnosed as having the carpal tunnel syndrome in at least one upper extremity.
Variability among Examiners The associations between clinical test results and nerve conduction outcomes were examined separately for each of the four examiners. Using the Breslow-Day test for homogeneity of the odds ratio (24), p. 173), we found no significant differences among examiners in the associations between the carpal tunnel syndrome and the Phalen sign (P = 0.19), Tinel sign (P = 0.31), or two-point discrimination (P = 0.09).
Bivariate Analyses Age was partitioned into three categories: under 40, 40 to 55, and over 55 years of age. The prevalence of the carpal tunnel syndrome in these categories was 2 4 % , 4 4 % , and 5 4 % , respectively (chi-squared = 6.31, 2 df, P = 0.043). Patients presenting with symptoms in both hands (Table 2) had a higher prevalence of the carpal tunnel syndrome than patients with unilateral complaints (chi-squared = 3.79, 1 df, P = 0.05). Gender ( P = 0.46), the duration of symptoms (P = 0.71), and the presence of nocturnal symptoms (P = 0.58) were not significantly associated with diagnosis. Exposure to either vibrating tools, frequent wrist flexion, or repetitive pinching was not significantly associated with the carpal tunnel syndrome (P = 0.93). The prevalence of the carpal tunnel syndrome was 3 8 % in workers, 3 1 % in subjects who were disabled or receiving Workmen's Compensation, and 6 0 % in retirees or homemakers. Slight or marked weakness in thenar muscle strength was detected in 3 6 % of patients, but these findings correlated poorly with diagnosis of the carpal tunnel syndrome (P = 0.62). Thenar atrophy was detected in 10% of patients; only half of these patients had the carpal tunnel syndrome (P = 0.22). Hand diagram ratings were grouped into those indicating the classic carpal tunnel syndrome or probably indicating the carpal tunnel syndrome and those possi-
Table 2. Bivariate Analyses of Clinical Findings in the Carpal Tunnel Syndrome Finding
Tinel sign Phalen sign Sensory loss Hand diagram ratingt Neurologist's assessment:): Age > 40 years Nocturnal symptoms Bilateral symptoms
Sensitivity Specificity
Odds Ratio (95% CI)*
In our!Sample with 40% :Prevalence Negative Positive Predictive Value Predictive Value (95% CI) (95% CI)
In Hypothetical Sample \vith 15% Prevj ilence Positive Negative Predictive Predictive Value Value
0.60 0.75 0.32
0.67 0.47 0.81
3.13 (1.40 to 7.00) 2.65(1.10 to 6.20) 1.98(0.81 to 4.55)
0.55(0.45 to 0.65) 0.48(0.40 to 0.54) 0.54(0.37 to 0.70)
0.72(0.63 to 0.79) 0.74(0.62 to 0.84) 0.63(0.58 to 0.68)
0.25 0.20 0.23
0.91 0.91 0.87
0.61
0.71
3.92(1.75 to 8.77)
0.59(0.48 to 0.68)
0.73(0.66 to 0.80)
0.27
0.91
0.84 0.80
0.72 0.42
13.16(4.95 to 34.48) 2.87(1.19 to 6.92)
0.67(0.58 to 0.73) 0.48(0.41 to 0.53)
0.87(0.78 to 0.93) 0.76(0.63 to 0.86)
0.34 0.20
0.96 0.92
0.77
0.28
1.29(0.33 to 4.01)
0.42(0.36 to 0.47)
0.64(0.48 to 0.78)
0.16
0.87
0.61
0.58
2.16(0.98 to 4.72)
0.49(0.40 to 0.58)
0.69(0.60 to 0.78)
0.20
0.89
* Relative odds for the carpal tunnel syndrome with 9 5 % CI. t Probable or classic compared with possible or unlikely hand diagram rating. t Probable or possible compared with unlikely neurologist's assessment.
1 March 1990 • Annals ofInternal Medicine • Volume 112 • Number 5
Downloaded From: http://annals.org/pdfaccess.ashx?url=/data/journals/aim/19699/ by a University of California San Diego User on 01/30/2017
323
Table 3. Bivariate Analyses Finding
Phalen or Tinel signs Phalen and Tinel signs Tinel sign or hand diagram ratingt Tinel sign and hand diagram rating! Phalen sign or hand diagram ratingf Phalen sign and hand diagram ratingf
of Combinations
Sensitivity Specificity
of Clinical Findings in the Carpal Tunnel
Syndrome
Odds Ratio (95% CI)*
In our Sample with 4 0 % Prevalence Positive Negative Predictive Value Predictive Value (95% CI) (95% CI)
In Hypothetical Sample with 1 5 % Prevalence Positive Negative p r edictive Predictive Value Value
0.88
0.41
5.21 (1.80 to 14.92)
0.50 (0.44 to 0.54) 0.84 (0.70 to 0.93)
0.21
0.95
0.46
0.73
2.34 (1.02 to 5.35)
0.53 (0.40 to 0.65) 0.68 (0.61 to 0.74)
0.23
0.89
0.82
0.50
4.50 (1.82 to 1.11)
0.52 (0.45 to 0.57) 0.81 (0.69 to 0.89)
0.22
0.94
0.39
0.89
5.31 (1.97 to 14.30)
0.71 (0.53 to 0.85) 0.69 (0.64 to 0.72)
0.39
0.89
0.82
0.38
2.74 (1.10 to 6.85)
0.47 (0.41 to 0.5) 0.76 (0.62 to 0.87)
0.19
0.92
0.49
0.83
4.77 (1.97 to 11.52)
0.66 (0.51 to 0.78) 0.71 (0.65 to 0.77)
0.34
0.90
* Relative odds for the carpal tunnel syndrome with 9 5 % CI. t Probable or classic compared with possible or unlikely hand diagram rating.
bly indicating or unlikely to indicate the carpal tunnel syndrome. Similarly, the neurologists' assessments were categorized as those probably or possibly indicating the carpal tunnel syndrome and those unlikely to indicate the carpal tunnel syndrome. As shown in Table 2, the hand diagram ratings (positive predictive value, 0.59; 9 5 % CI, 0.43 to 0.68), and neurologists' assessments (positive predictive value, 0.67; CI, 0.58 to 0.73) were better predictors of the carpal tunnel syndrome than Phalen or Tinel signs. The prevalence or pretest probability of the carpal tunnel syndrome in our sample was 0.40. If the positive predictive value of a test is not substantially greater than the sample prevalence, the test is of little value in distinguishing outcomes among subjects with a positive test. Thus, the positive predictive values associated with Tinel and Phalen signs and two-point discrimination (Table 2) are quite modest. Similarly, if the negative predictive value of a test is less than the prevalence of subjects without the carpal tunnel syndrome (0.60), the test is of little value in distinguishing outcomes among subjects with negative test results. The neurologists' assessments had the highest negative predictive value (0.87; CI, 0.78 to 0.93), followed by Tinel sign, Phalen sign, and the hand diagram ratings. In a hypothetical population whose prevalence of the carpal tunnel syndrome was 1 5 % , the positive predictive values would be much lower and the negative predictive values would be higher. For example, the positive predictive value of Tinel sign would be 0.25 and the negative predictive value would be 0.91. Predictive values of other findings in a population with a prevalence of 1 5 % are shown in Table 2. We assessed the diagnostic utility of tests used in various combinations; the results are shown in Table 3. Requiring that just one of a pair of tests be positive raises sensitivity but sacrifices specificity; similarly, requiring that both tests be positive raises specificity but
324
lowers sensitivity. The combination of a positive Tinel and a probable or classic diagram rating was very specific (0.89) and had a higher positive predictive value (0.71; CI, 0.53 to 0.85) than either test used singly. The same was true of a positive Phalen and a probable or classic diagram rating. The combination of a positive Phalen or Tinel was very sensitive (0.88) and had a high negative predictive value (0.84; CI, 0.70 to 0.93). Table 3 shows that the positive predictive values of these combinations would be lower and the negative predictive values higher in a population with a lower prevalence of the carpal tunnel syndrome. Multivariate Analysis In the logistic regression analysis, nocturnal symptoms, unilateral (as compared with bilateral) complaints, and Phalen sign were not associated with the carpal tunnel syndrome at a significance level of 0.05 and were dropped from the model. A term was added to account for a suspected interaction between hand diagram rating and age. The final model included hand diagram rating (P = 0.001), Tinel sign (P = 0.01), and the interaction between age category and hand diagram rating (P = 0.06). The adjusted odds ratios from the logistic regression are presented in Table 4. The hand pain diagram rating was more predictive of the carpal tunnel syndrome in the youngest age group (odds ratio, 12.5; CI, 1.95 to 80.58) than in the middle (odds ratio, 1.19; CI, 0.35 to 4.11) and oldest age groups (odds ratio, 8.10; CI, 1.23 to 53.28). We evaluated the independent contributions of the neurologists' assessments and the clinical findings (age, hand diagram rating, and Tinel sign) by comparing multivariate models containing clinical findings, the neurologists' assessments, or both. The log likelihood statistic was used to test for significance
1 March 1990 • Annals ofInternal Medicine • Volume 112 • Number 5
Downloaded From: http://annals.org/pdfaccess.ashx?url=/data/journals/aim/19699/ by a University of California San Diego User on 01/30/2017
(24, p. 205). The neurologists' assessments were added to the model comprised of clinical findings with significant improvement in fit {P < 0.001). Age category and Tinel sign no longer contributed significantly to the model when the assessment was added, but the hand diagram rating remained a significant predictor of the outcome ( P = 0.02). Addition of the clinical findings to a model containing the neurologists' assessments also significantly improved the fit (P < 0.05). Table 5 shows the risk for the carpal tunnel syndrome in patients stratified by age category, hand diagram rating, and Tinel sign. In general, the risk for the carpal tunnel syndrome increases with the presence of additional positive findings. The table identifies strata associated with particularly low and high risks for the carpal tunnel syndrome. Patients under 40 years of age with possible or unlikely hand diagram ratings (first two strata) had an 8.7% ( 2 / 2 3 ) prevalence of the carpal tunnel syndrome, whereas patients over 55 years of age with positive Tinel signs and probable or classic hand diagram ratings had an 88.9% ( 8 / 9 ) prevalence of the carpal tunnel syndrome.
Table 5. Prevalence of the Carpal Tunnel Syndrome in Subgroups of Subjects Defined by Particular Clinical Features
Discussion
that included pregnant patients with hand symptoms. The most informative data from the history appeared to be the anatomic localization of symptoms. A self-administered hand pain diagram, on which patients indicated the location and quality of their complaints, had a higher predictive value (0.59; CI, 0.48 to 0.68) than any provocative test or physical examination finding. The diagram was more predictive of the carpal tunnel syndrome in patients under 40 years of age (odds ratio, 12.5) than in the middle and oldest age groups (odds ratios, 1.2 and 8.1, respectively). This finding was not explained by variation in the distribution of diagnoses across age categories. Older patients may have had more difficulty completing the diagram accurately. Neurologists with experience in evaluating upper extremity disorders identified patients at low, moderate, and high risk for the carpal tunnel syndrome with impressive accuracy. In logistic regression analyses, addition of the neurologists' assessments to a model containing age, Tinel sign, the hand diagram rating, and the age-diagram interaction significantly improved the fit, indicating that the neurologists' assessments incorporated diagnostically valuable information that was not captured by the clinical tests. Similarly, addition of the clinical tests to a model containing the neurologists' assessments significantly improved the fit, indicating that the clinical tests added important information to the neurologists' assessments. The hand diagram rating was the only clinical test that contributed significantly to the model that included clinical tests and the neurologists' assessments. These data suggest that although experienced clinicians use information not captured by our study variables, experienced clinicians' assessments might be improved by using the results of clinical tests, especially hand diagram ratings. Using combinations of tests may increase the diagnostic value of the tests (27, p. 41). The presence of
Previous reviews (6, 21) of the carpal tunnel syndrome have emphasized the diagnostic utility of provocative tests, bilaterality of symptoms, and nocturnal paresthesias, but our data suggest that these findings are of limited value. Specifically, nocturnal symptoms were not significantly associated with the carpal tunnel syndrome, bilateral involvement had little independent predictive value in the multivariate analysis, and physical findings traditionally associated with the carpal tunnel syndrome, including Phalen and Tinel signs, two-point discrimination, and thenar muscle atrophy and weakness, were of limited utility. Of these tests, Tinel sign had the highest positive predictive value (0.55; CI, 0.45 to 0.65) and Phalen sign had the highest negative predictive value (0.74; CI, 0.62 to 0.84). We found that age is also a risk factor for the carpal tunnel syndrome, in agreement with other studies (26). This finding would be less striking in a sample Table 4. Adjusted Odds Ratios for the Carpal Tunnel drome Associated with Specific Clinical Features
Syn-
Clinical Feature
CI)*
Age Categoryt, y Under 40 40 to 55 Over 55 Positive Tinel signj Probable or classic hand diagram rating§ Entire sample Patients under 40 Patients 40 to 55 Patients over 55
Odds Ratio (95% 1.00 2.82 3.74 3.11 4.95 12.55 1.19 8.10
(not applicable) (0.91 to 8.69) (0.99 to 14.13) (1.26 to 7.70) (1.87 (1.95 (0.35 (1.23
to to to to
13.08) 80.58) 4.11) 53.28)
* 9 5 % CI = 9 5 % confidence interval for the relative odds for the carpal tunnel syndrome. t Reference age category is under 40 years of age. % Compared with negative Tinel sign. § Compared with possible or unlikely hand diagram rating.
Tinel Sign*
_ + — — — + — + — + + +
Total
Patients in Subgroup Number Percentage with Hand Diagram Age Ratingf Category} the Carpal Tunnel Syndrome
_ — — — + — + + + — + +
1 1 3 2 1 3 2 2 3 2 1 3
13 10 9 16 6 3 11 8 5 10 7 9 107§
8 10 22 25 33 33 45 50 60 70 71 89
* — = negative, + = positive Tinel sign. t — = possible or unlikely, + = probable or classic hand diagram rating. %\ = under 40, 2 = 40 to 55, 3 = over 55. § Data missing for three patients.
1 March 1990 • Annals of Internal Medicine • Volume 112 • Number 5
Downloaded From: http://annals.org/pdfaccess.ashx?url=/data/journals/aim/19699/ by a University of California San Diego User on 01/30/2017
325
both a positive Tinel sign and a probable or classic hand diagram rating was associated with impressive specificity (0.89) and high positive predictive value (0.71; CI, 0.53 to 0.85). The presence of a positive Phalen sign and a probable or classic hand diagram rating was also valuable (Table 3 ) . The combinations of Phalen or Tinel signs and a Tinel sign or a positive hand diagram rating were nonspecific but had high sensitivity and negative predictive value. These data illustrate that the optimal choice of tests or combination of tests depends on whether the goal is to establish diagnosis of the carpal tunnel syndrome or to exclude it. Predictive values vary with the prevalence of disease. Table 2 indicates that in a sample with a 1 5 % prevalence of the carpal tunnel syndrome, as might be encountered in a high-risk industry ( 2 ) , the positive predictive values of non-invasive tests would be just 0.20 to 0.30, and the negative predictive values would be approximately 0.90. This finding has important implications for occupational screening. Most asymptomatic workers screened with these non-invasive tests will not have the carpal tunnel syndrome regardless of test results. The tests are valuable in diagnosing the syndrome only if the previous probability of the carpal tunnel syndrome is substantially higher than the baseline prevalence, as would be expected in symptomatic workers. An important goal of our study was to determine whether specific constellations of clinical signs and symptoms could identify patients with particularly high or low probabilities for having the carpal tunnel syndrome. In particular, identifying patients who could be managed without nerve conduction tests would be valuable. As shown in Table 5, risk categories can be defined by age, Tinel sign, and hand diagram rating. Patients under 40 years of age with possible or unlikely hand diagram ratings had a low (8.7) likelihood of having the carpal tunnel syndrome. Although the low-risk group included only 23 patients, this stratum would likely account for a larger proportion in a less select population. This clinically defined approach to identifying low-risk patients should be validated in an independent sample. If confirmed, these findings would support the practice of excluding the carpal tunnel syndrome without doing nerve conduction tests in selected patients. Since the carpal tunnel syndrome is neither fatal nor rapidly progressive, clinicians would probably accept an 8% to 9 % false negative rate for this potential prediction rule. Our study has two potential limitations: First, patients referred to a hospital-based laboratory may have atypical findings or multiple causes for their symptoms, lowering the specificity of non-invasive tests. Referred patients might also have the carpal tunnel syndrome particularly severely, raising sensitivity (28). Therefore, our results must be generalized to the primary care setting cautiously. Second, using nerve conduction tests to define the carpal tunnel syndrome might be questioned, because their sensitivity is approximately 9 0 % ( 8 ) ; moreover, their accuracy may vary among laboratories and neurophysiologists.
326
However, no alternative tests have been shown to be comparable or superior. Further research might be done in the occupational setting or with outpatients presenting de novo to primary providers to determine whether the performance characteristics of these tests change significantly in different populations. It would be particularly valuable to confirm that non-invasive clinical findings can identify a subgroup of patients at particularly low risk for the carpal tunnel syndrome who can be managed effectively without nerve conduction studies. Acknowledgments: The authors thank E. Francis Cook, SD, Lee Goldman, M D , and David Bates, M D , for their contributions; Barbara Shulman, RPT, Susan Steindorf, RPT, and Lisa Bail for doing clinical evaluations; Patricia Butler and Margaret Mills for scheduling patients; and Mari Garcia and Jacqueline Mazzie for manuscript preparation. Grant Support: By N I H Grants AR36308 and AR07530 and the Kellogg Program for Training in Research in Clinical Effectiveness. Requests for Reprints: Jeffrey N. Katz, MD, Arthritis Center, Brigham and Women's Hospital, 75 Francis Street, Boston, MA 02115. Current Author Addresses: Dr. Katz: Arthritis Center, Brigham and Women's Hospital, 75 Francis Street, Boston, MA 02115. Dr. Liang: Department of Rheumatology-Immunology, Brigham and Women's Hospital, 75 Francis Street, Boston, MA 02115. Dr. Larson, Ms. Eaton, and Ms. Fossel: Robert B. Brigham Multipurpose Arthritis Center, 221 Longwood Avenue, Boston, MA 02115. Dr. Stirrat: Longwood Hand Center, 830 Boylston Street, Chestnut Hill, MA 02167. Dr. Sethi: 317 Patrick Street Southwest, Vienna, VA 22180. Dr. Sabra: Department of Medicine, New England Baptist Hospital, 125 Parker Hill Avenue, Boston, MA 02120. Dr. Krarup: Department of Neurophysiology, Rigshospitalet, Blegdamsvej 9, 2100 Copenhagen, Denmark.
References 1. The National Ambulatory Medical Care Survey. Summary, United States. Rockville, Maryland: U.S. Department of Health, Education, and Welfare, Public Health Service, Health Resources Administration, National Center for Health Statistics; 1975: D H E W No. ( H R A ) 76-1772. 2. Masear VR, Hayes JM, Hyde AG. An industrial cause of carpal tunnel syndrome. J Hand Surg [Am]. 1986;11:222-7. 3. Gelberman RH, Aronson D, Weisman MH. Carpal tunnel syndrome. Results of a prospective trial of steroid injection and splinting. / Bone Joint Surg [Am]. 1980;62:1181-4. 4. Armstrong TJ. An Ergonomics Guide to Carpal Tunnel Syndrome. Akron, Ohio: American Industrial Hygiene Association; 1983. 5. Hoffman DE. Carpal tunnel syndrome. Importance of sensory nerve conduction studies in diagnosis. JAMA. 1975;233:983-4. 6. Dorwart BB. Carpal tunnel syndrome: a review. Semin Arthritis Rheum. 1984;14:134-40. 7. Stevens J C . A A E E minimonograph # 2 6 : the electrodiagnosis of carpal tunnel syndrome. Muscle Nerve. 1987;10:99-113. 8. Grundberg AB. Carpal tunnel decompression in spite of normal electromyography. J Hand Surg [Am]. 1983;8:348-9. 9. Bleeker ML. Medical surveillance for carpel tunnel syndrome in workers. / Hand Surg. 1987;12:845-8. 10. Weiss KL, Beltran J, Lubbers LM. High-fold MR surface-coil imaging of the hand and wrist. Part II. Pathologic correlations and clinical relevance. Radiology. 1986;160:147-52. 11. Lundberg G, Lie-Stenstrom AK, Sollerman C, Strombers T, Pykko I. Digital vibrogram: a new diagnostic tool for sensory testing in compression neuropathy. J Hand Surg [Am]. 1986;11:693-9. 12. Herrick RT, Herrick SK. Thermography in the detection of carpal tunnel syndrome and other compressive neuropathies. / Hand Surg [Am]. 1987;12:943-9. 13. Gellman H, Gelberman RH, Tan AM, Botte MJ. Carpal tunnel syndrome. An evaluation of the provocative diagnostic tests. / Bone Joint Surg [Am]. 1986;68:735-7. 14. Posch JL, Marcotte DR. Carpal tunnel syndrome: an analysis of 1201 cases. Orthop Rev. 1976;5:25-34. 15. Gelberman RH, Szabo RM, Williamson RV, Dimick MP. Sensibility testing in peripheral-nerve compression syndromes. An experimental study in humans. J Bone Joint Surg [Am]. 1983;65:632-8. 16. Golding DN, Rose DM, Selvarajah K. Clinical tests for carpal tunnel syndrome: an evaluation. Br J Rheumatol. 1986;25:388-90. 17. Seror P . Electroclinical correlations in the carpal tunnel syndrome.
1 March 1990 • Annals ofInternal Medicine • Volume 112 • Number 5
Downloaded From: http://annals.org/pdfaccess.ashx?url=/data/journals/aim/19699/ by a University of California San Diego User on 01/30/2017
Apropos of 100 cases. Rev Rhum Mai Osteoartic. 1987;54:643-8. 18. Katz JN, Stirrat CS. A self-administered hand diagram for the diagnosis of carpal tunnel syndrome. / Hand Surg [Am]. 1989; 14[In press]. 19. Dellon AL. The moving two-point discrimination test: clinical evaluation of the quickly adapting fiber/receptor system. / Hand Surg [Am]. 1978;3:474-81. 20. Mossman SS, Blau JN. TinePs sign and the carpal tunnel syndrome. Br Med J [Clin Res]. 1987;294:680. 21. Phalen GS. The carpal tunnel syndrome. Seventeen years' experience in diagnosis and treatment of six hundred fifty-four hands. / Bone Joint Surg [Am]. 1966;48:211-28. 22. McNeil BJ, Keeler E, Adelstein SJ. Primer on certain elements of medical decision making. N Engl J Med. 1975;293:211-5. 23. Feinstein AR. Clinical Epidemiology: The Architecture of Clinical
Research. Philadelphia: W.B. Saunders Company; 1985. 24. Davis W, Breslow NE, Day NE. Statistical Methods in Cancer Research. 1. Lyon: International Agency for Research on Cancer; 1980. 25. SAS Institute, Inc. SAS/STAT User's Guide, release 6.03 ed. Cary, North Carolina: SAS Institute, Inc.; 1988:189-282. 26. Stevens JC, Sun S, Beard CM, O'Fallon WM, Kurland LT. Carpal tunnel syndrome in Rochester, Minnesota, 1961 to 1980. Neurology. 1988;38:134-8. 27. Galen RS, Gambino SR. Beyond Normality: The Predictive Value and Efficiency of Medical Diagnoses. New York: John Wiley and Sons; 1975. 28. Ransohoff DF, Feinstein AR. Problems of spectrum and bias in evaluating the efficacy of diagnostic tests. AT Engl J Med. 1978;229:926-30.
1 March 1990 • Annals of Internal Medicine • Volume 112 • Number 5
Downloaded From: http://annals.org/pdfaccess.ashx?url=/data/journals/aim/19699/ by a University of California San Diego User on 01/30/2017
327
