<scp>AAOS</scp> 2024 <scp>CTS</scp> Guidelines—Comments by Members of <scp>AANEM CTS</scp> Severity Scale Workgroup

The recently promulgated guideline on the diagnosis of carpal tunnel syndrome (CTS) from the American Academy of Orthopedic Surgeons (AAOS) recommends that the CTS-6 clinical assessment is sufficient for the diagnosis of CTS, and that the routine use of high-resolution ultrasonography (HRUS) or nerve conduction studies/electromyography (NCS/EMG) is not required. This recommendation, which can be found here aaos.org/cts2cpg, is based on the AAOS's review of the literature and consensus development process, which compared the diagnostic sensitivity and specificity of these three methods. Much is made in the final recommendations of the CTS-6 [1]. This is presented as a validated diagnostic tool with high sensitivity and specificity, considered by the AAOS to be as reliable as NCS or HRUS in diagnosing CTS. Validation, in this context, implies that the diagnosis of CTS made by the CTS-6 has been compared against the diagnosis made by other methods and has demonstrated substantial agreement. This assertion is supported mainly by studies from Fowler et al. [2-7]. It is worth examining the development and validation of this tool in detail. The CTS-6, developed by Graham, is an evaluation tool that assesses the likelihood of CTS based on a combination of medical history and physical examination findings. It evaluates the presence of six findings: numbness in the median nerve distribution, nocturnal symptoms, thenar atrophy or weakness, a positive Phalen test, loss of 2-point discrimination, and a positive Tinel sign. Each of these parameters is scored as being present (1) or absent (0), and each parameter is weighted based on the relative contribution to the diagnosis, and the responses are summed. The sum scores are related to the probability of CTS, and in most studies, a score of > 12 is taken as indicating the presence of CTS, though other thresholds are possible. Confusingly, there is a different CTS-6 [8], a reduced version of the Boston Carpal Tunnel Questionnaire (BCTQ) [9], devised by Isam Atroshi, used as a patient-reported outcome measure to assess treatment outcomes for CTS. Confusion between these tools has occurred in published studies, in which research on the Atroshi survey has been misinterpreted as pertaining to the Graham CTS-6 [10]. The Graham CTS-6 was originally constructed through a Delphi consensus process among “CTS-experts,” predominantly hand surgeons. This process relied on expert opinion rather than real-world patient data and reflects the perspectives of a highly specialized group who primarily see more severe CTS cases. Most CTS patients, however, never consult a surgeon [11]. Validation of the CTS-6 involved comparison with the opinions of a similar group of experts, resulting in agreement that demonstrates face validity but lacks convergent or external validity. The recommendation that the CTS-6 instrument is sufficiently accurate for routine diagnosis is based upon 15 studies. According to the AAOS clinical practice guideline, 10 were rated as high quality and 5 as moderate; however, the criteria for selecting and rating these studies remain unclear. A review suggests the studies are generally small (40–265 patients) and highly heterogeneous. Some focused on specific patient groups, such as individuals with diabetes [12] or hypothyroidism [13], and most were based on convenience samples from hand surgery clinics, unrepresentative of the broader CTS population. Methodological concerns abound as well, including treating the right and left hands of one patient as independent observations [2, 14] (which they are not), or retrospectively selecting optimal cutoff values from receiver operating characteristic (ROC) curves without validation in prospective datasets. One study relied on sensitivity and specificity figures from the literature to demonstrate, via Bayes' theorem, that additional tests minimally change the likelihood of diagnosis when the prior probability of CTS is very high [15], offering no direct insight into the sensitivity or specificity of the tests themselves. Of the 15 studies, only one assesses the diagnostic properties of the CTS-6 [2]. This used the statistical technique of latent class analysis in a sample of 187 hands and estimated that the sensitivity and specificity of the three tests were as follows: NCS: 97%/40% Ultrasound: 89%/72% CTS-6: 75%/61% An earlier paper by the same group [5] (not included in the 15 AAOS papers) used the same method in a sample of 85 patients and reported the following estimates for sensitivity and specificity: NCS: 91%/83% Ultrasound: 91%/94% CTS-6: 95%/91% The rather striking difference in specificity figures for all three tests between these two papers probably calls for an expert statistical review of the methods. Apart from these two studies using latent class analysis, only two other conventional validation attempts for the Graham CTS-6 exist, neither of which is included in the AAOS 15. One study reported the CTS-6 achieving 31%–75% sensitivity and 59%–99% specificity against a clinical reference diagnosis, depending on the cutoff score used to define a diagnosis of CTS [6]. The cutoff with the best performance, as judged by the area under the ROC curve, achieved 75% sensitivity and 59% specificity. Another study, using NCS confirmation in 98 patients [16], found: 50% probability of CTS on CTS-6: Sensitivity 87%, Specificity 60%. 80% probability of CTS on CTS-6: Sensitivity 60%, Specificity 90%. However, the latter study used a dichotomous group of subjects selected by hand surgeons, artificially simplifying the task of distinguishing clear CTS cases from non-CTS cases, thereby inflating test performance through selection bias. Pooling these four studies suggests that the CTS-6 achieves about 82% sensitivity and 68% specificity. Notably, three studies were conducted by the same authors [2, 5, 6], and the fourth has significant methodological flaws. Thus, claims of the CTS-6's validation and accuracy are weakly supported, and its performance appears only moderately effective in surgical hand clinic populations. A well-designed study in a more representative CTS population is urgently needed to establish its true diagnostic accuracy. A striking feature of the studies by Fowler et al. is the consistently low specificity of NCS for CTS, ranging from 27% to 83% (mean 67%), implying high false-positive rates. This contrasts with a larger AANEM review, which reported a sensitivity of 93% and specificity of 95% for the combined sensory index, and an overall sensitivity of 85%–90% and specificity of 82%–85% for NCS in general [17]. The neurophysiological methods and criteria used by Fowler et al. are often vaguely described, but in one study, their stringent criteria for normal NCS included median nerve motor latency < 4.0 ms, compound muscle action potential (CMAP) amplitude in abductor pollicis brevis (APB) 10 mV, median nerve sensory latency < 3.5 ms, median sensory nerve action potential (SNAP) amplitude > 6 μV, and relative median-ulnar nerve sensory latency < 0.5 ms, resulting in a specificity of 27% [3]. These cutoffs are more strict than AANEM standards, which, for example, consider an APB CMAP amplitude > 4.2 mV normal for typical CTS patients [18]. Such criteria also suffer from the issue identified by Redmond and Rivner, where making multiple measurements and considering the overall test positive if any single value is outside the reference range greatly increases the false positive rate [19]. Of the 15 studies referenced in the AAOS guidelines, only 4 use neurophysiological methods that clearly conform to the AANEM guidelines on how testing should be performed for CTS [14, 20-22]. The methods employed in the remaining 11 are either out of date or too poorly described to be assessed. The AAOS guidelines appear to have only relied on the 15 referenced studies to assess the diagnostic properties of ultrasound imaging in CTS. This is odd given the extensive literature on this topic with several well-performed meta-analyses [23-28], and even an “umbrella” meta-analysis of the meta-analyses [29]. These publications overall suggest that ultrasound imaging is more accurate diagnostically than the 82% sensitivity and 68% specificity achieved by the CTS-6, generally in the region of 80%–90% sensitivity and specificity. In clinical practice, diagnostic tests are often unnecessary for patients with possible CTS. For roughly one-third of patients, the diagnosis is clinically obvious, and no test (CTS-6, HRUS, or electrodiagnostic studies will alter this) [15]. Similarly, in another third, it is clear the diagnosis is not CTS, thus rendering such tests redundant. In these circumstances, they can at most only show the presence of asymptomatic median neuropathy (EDX) or incidental anatomical anomalies (HRUS). Tests are only needed for diagnosis in the remaining third of patients in whom the diagnosis is uncertain, and here sensitivity, specificity, and predictive values are crucial. In cases with a firm CTS diagnosis, tests are needed to objectively assess severity, guide treatment, and monitor progression. EDX provides the most validated measure of physiological impairment in CTS, with evidence supporting its prognostic value for surgery and its responsiveness to change [30]. Several studies have attempted to determine whether either the ultrasound findings or the CTS-6 constitutes a valid measure of CTS severity. Ideally, this would be evaluated in the same fashion as EDX, through studies of prognostic value and responsiveness to change. However, such studies are lacking, and the existing literature only examines correlations of CTS-6 scores or ultrasound CSA measurements with NCS-assessed severity. The results are not encouraging: Martikkala et al. [31] and Moran et al. [14] concluded that ultrasound CSA could not be used to predict neurophysiological severity, though Moran et al. found slight differences in CSA between mild and very severe CTS. Kanagasabai [32] demonstrated a relationship between CSA measurements and neurophysiological severity categorized using Stevens' criteria [33], but the presentation of the results is such that it is impossible to extract any estimate of the strength of correlation. However, the means and standard deviations for CSA measurements in the different severity groups suggest that there is significant overlap between groups, for example, between “mild” disease (mean CSA +2 standard deviations, for “CSA 2” = 14.6) and “moderate” disease (mean CSA −2 standard deviations for “CSA 2” = 11.9), suggesting that the performance of ultrasound in assigning neurophysiological severity classifications would be poor. Nkrumah et al. [34], another study from Fowler et al. of 274 hands, attempted to determine whether either CTS-6 or ultrasound CSA could identify severe CTS (absent sensory/motor responses). In this task, the CSA measurement achieved 37% sensitivity and 82% specificity, with an area under the ROC curve of 0.79, while the area under the ROC curve was 0.62 for the CTS-6 only. Yang et al. [10] showed a significant correlation between CTS-6 and NCS severity in 105 patients, but an r value of only 0.39 indicates that it would not be possible to substitute one measurement for the other in individual patients. There is also a study from Dr. Fowler's group correlating the CTS-6 score with the ultrasound CSA that proves to be significant, but with a rather poor r value of 0.53 [35]. In a much larger series of 1560 right and 1748 left hands [30], the r values for the correlations between CSA and NCS grade were 0.43 and 0.45, respectively—significant, but not good enough to be useful in individual patients. We think that the Graham-CTS-6 has not yet satisfactorily demonstrated its performance as a diagnostic tool in patients who do not already have a firm clinical diagnosis of CTS, is unsuited for the assessment of disease severity in deciding treatment, and does not provide the structural insights given by ultrasound imaging. It, therefore, cannot be recommended as the sole assessment tool when evaluating a patient with possible CTS. Jeremy Bland: conceptualization, writing – original draft, writing – review and editing. Devon Rubin: writing – review and editing. Lawrence Robinson: writing – review and editing. Michael Cartwright: writing – review and editing. Michael Andary: writing – review and editing. John Kincaid: writing – review and editing. Zainab Al-lawati: writing – review and editing. Peter Grant: writing – review and editing. We confirm that we have read the Journal's position on issues involved in ethical publication and affirm that this report is consistent with those guidelines. Dr John Kincaid reports receiving royalties for authorship in the Rehabilitation Medicine Quick Reference series by Demos Medical Publishers. Carpal tunnel diagnosis, including EMG, is covered in the book. The other authors declare no conflicts of interest. The authors have nothing to report.