On the other hand, a variety of previous studies have indicated predictive values of NCS for functional outcome of CTS after different treatment methods [15-17]. By contrast, several studies indicated no prognostic value at of NCS for predicting functional outcome after decompression [18-21]. The discrepancy may result from a variety of factors, such as false-negative diagnosis rate at 10-20% [22], diverse practical and gradation standards [7, 8]. Since morphological information obtained by ultrasound can also diagnose and reflect severity of the disease [6, 10, 23-26], several multivariate regression studies regarding prognostic values of ultrasound were performed, which had discussed its predictability for diagnostic confirmation and clinical outcome after surgery [27, 28]. Nevertheless, to the best of our knowledge, there has not been studies performed to examine if functional outcome for CTS associated with axonal degeneration detected by ultrasound can be predicted ahead of treatment, yet with no evidences examined prognostic values of those newly identified ultrasound cut-off values. Therefore, this study was proposed, aiming at examining the prognostic value of ultrasound for functional recovery after treatment. It is of clinical significance to improve patients’ overall function and avoid unnecessary disease deterioration or relapse, provided ultrasound bears its nature of economic benefits, simplicity and non-invasiveness [9]. Our null hypothesis is that ultrasound parameters cannot predict functional recovery for CTS associated with axonal degeneration.
Materials and methods
This retrospective study was approved by the Institutional Review Board of the University of Hong Kong/Hospital Authorities in Hong Kong West (HKU/HA HKW IRB, Ref. Number: UW No.: 17-076) and registered in institutional clinical trial registry (Clinical registration number: HKUCTR-2201), Hong Kong SAR, China. Oral consent was obtained before conducting telephone interview with patients’ confidentiality and rights well protected.
Subjects who were admitted from January of 2012 to December of 2017 were screened from the Clinical Management System (CMS) and their records of NCS and ultrasound data before treatment, which were obtained following standard procedures that has been described in our previous studies [11], were fully retrieved. The inclusion criteria were set as follow: (1) Age < 70 years, (2) Primarily diagnosed as CTS; (3) Right handedness; (5) R-CSA (ratio of cross-sectional area) > 1.85, R-P (ratio of perimeter) > 1.68, DCSA (changes of CSA from wrist to distal one third forearm) > 6.98 mm2 and DP (changes of perimeter from wrist to distal one third forearm) > 5.77 mm[11] measured by ultrasound; while the exclusion criteria were: (1) Age 70 years, (2) Abnormal anatomical structure in median nerve such as bifid structure; (3) R-CSA 1.85, R-P 1.68, DCSA 6.98 mm2 and DP 5.77 mm [11] measured by ultrasound; (4) no complete records of NCS or ultrasound reports. Then, the structured telephone interview was performed by an experienced Occupational Therapist, asking subjects to grade their recovery on the affected hand, ranging from (1) fully-recovered, (2) improved, (3) residual syndrome, (4) similar and (5) deteriorated.
We summarized and categorized the following variables as predictors when establishing the multivariate ordinal logistic regression model as follow: (1) Age: unit in year; (2) Gender: male or female; (3) NCS severity grade: from mild to very severe grade based on the Bland’s classification [29]; (4) Treatment methods: a. surgical method: surgical decompression; b. conservative method: including resting splint, tendon/nerve gliding exercise, educational session at outpatient clinic and taking non-steroidal anti-inflammatory drugs, ; (5) Affected hand side: left or right side; (6) Comorbidities: a. CTS-relevant comorbidity: comorbidities that may cause CTS, including diabetes mellitus, thyroid dysfunction, fluid retention, menopause, high blood pressure, obesity, rheumatoid arthritis, hemodialysis, pseudo-gout, polymyalgia rheumatic, acromegaly, carpal stenosis, fractures or trauma to the wrist, patent median artery; b. CTS-irrelevant comorbidities: comorbidities which may not cause CTS, such as skin abscess, endometrial polyp, cellulitis, sepsis, cataract, constipation, gastritis, glaucoma and c. No comorbidities; (7) Surgical history: surgical record within recent 2 years were reviewed and categorized as follow: a. CTS-relevant surgical history: surgeries which were associated with CTS-relevant comorbidities mentioned above, including thyroidectomy, hysterectomy, oophorectomy, open reduction and internal fixation (ORIF) at wrist; b. CTS-irrelevant surgery: surgeries which were disassociated with CTS-relevant comorbidities, including hysteroscopy, skin lipoma excision, laminoplasty, prostatectomy, cholecystectomy, laparoscopy, sigmoidectomy, angioplasty and c. no surgical history within two years; (8) Length of follow-up: days from pre-treatment NCS/ultrasound to date of telephone interview; (9) ultrasound parameters: R-CSA, R-P and DP. The dependent variable is the grade of functional recovery reported by the subject.
Data analysis
Enrolled subjects were divided in to surgical (Group One) and conservative group (Group Two). IBM SPSS Statistics 24 (SPSS, Chicago, IL, USA) was used for analyzing the data. Demographics were generated descriptively. To compare demographics, NCS & US performance between groups, Chi-square test was used to examine categorical variables, including gender, affected hand side, comorbidities and surgical history; Mann-Whitney U test was adopted to examine NCS severity grade and functional recovery while independent t-test was applied to examine age, length of follow-up, NCS and US parameters.
Multicollinearity diagnostic test was performed via linear regression analysis to examine the variance inflation factors (VIF) among predictors. As moderate to strong correlation exists among ultrasound parameters (r=0.324-0.979, p<0.0001), any combinations of ultrasound parameters with VIF value > 3.0 were omitted. Finally, it was found that two pairs of ultrasound parameters were appropriate for model establishment, both with VIF < 3.0 (R-CSA & R-P, VIF: 1.142; R-CSA & DP, VIF: 1.12).
Then, by taking different combinations of ultrasound parameters into account, two ordinal regression models were established. The ordinal and categorical variables, which included NCS severity grade, treatment methods, gender, surgical history, comorbidities and affected hand side, were put into factors; whereas the continuous variable were considered as covariates, including length of follow-up and ultrasound parameters. A series of tests, including model fitting test, goodness-of
-fit test, Pseudo
R-square test and test of parallel line were conducted to examine the model performance, with estimates and 95% confidence interval(CI) obtained. Finally, bootstrapping method was applied to improve the model robustness.
Result
Demographic, neurophysiological and ultrasonographic characteristics
98 subjects with 196 hands were enrolled, with 75 hands treated by surgical decompression (Group One) while 121 hands treated by conservative method (Group Two). There were no significant differences in age (t=-0.727, p=0.468), affected hand side (Pearson Chi-square=1.058, p=0.304) and surgical history (Pearson Chi-square=5.5, p=0.064) while significant difference in gender (Chi-square=3.9, p=0.048), diagnostic severity (Z=-2.257, p=0.024), functional recovery (Z=-5.785, p<0.0001), length of follow-up (t=2.761, p=.007) and comorbidities (Chi-square=6.241, p=0.044) between groups.
Regarding performance of NCS, there were no significant differences in palmar DML (distal motor latency, t=1.372, p=0.173), wrist ML (motor latency, t=1.909, p=0.06), wrist CMAP (t=-1.96, p=0.054), wrist MCV (Motor conduction velocities, t=-1.466, p=-1.47), DSL (Distal sensory latency, t=1.197, p=0.27), SNAP (sensory nerve action potential, t=-0.861, p=0.506) and SCV (sensory conduction velocity, t=-0.626,p=0.557) only except palmar CMAP (compound motor action potential, t=-2.91, p=0.004) between groups. As for ultrasound performance, there were no significant differences in R-CSA (t=-0.507, p=0.613), R-P (t=-1.26, p=0.209), CSA (t=1.583, p=0.116) and P (t=-0.78, p=0.436). Overall, it appeared Group One was slightly severe than Group Two but the functional recovery was better than Group Two.
Model performance
The model fitting information revealed all of the regression coefficients in Model 1 and Model 2 were not equal to zero, indicating significant effect of the predictor variables (Model 1: Chi-square=399.56, p<0.0001; Model 2: Chi-square=395.387, p<0.0001). Goodness-of-fit test revealed the predicted probabilities were not deviated from the observed probabilities (Model 1: Chi-square=758.678, p=0.216, Model 2: Chi-square=594.208, P=1.0). The overall performance of Psuedo R-square tests revealed both models may explain 49-52% of the total variance in Model 1 while 50-53% of that in Model 2 as were shown in Cox & Snell as well as Nagelkerke tests. Meanwhile, the result of McFadden test revealed both models were far better than the intercept models. Test of parallel lines revealed the ordered logit coefficients were equal across the levels of the outcome in both models (Model 1: Chi-square=53.24, p=0.19; Model 2: Chi-square=42.52, p=0.58). Therefore, the overall result indicated both models were appropriate to predict functional outcome.
Performance of the ordinal logistic regression model and bootstrapping method
In the Model 1, it revealed R-CSA (OR= 4.4(1.15), 95% CI=2.14-6.66, p<0.001), R-P (OR=11.56(1.73), 95% CI=8.16-14.95, p<0.001), age (OR= -0.07(0.02), 95% CI=-0.109- -0.024, p<0.05), treatment method (OR=-1.72(0.35), 95%CI=-2.4- -1.04, p<0.001) and NCS severity grade (mild) (OR=3.56(1.26), 95% CI=1.09-6.03, p<0.05) as significant predictors associated with dependent variable. For one unit increase of R-CSA, the probability of functional recovery decreased 4.4 times (The functional recovery gets worse as the score is rated higher); for one unit increase of R-P, the probability of functional recovery decreased 11.56 times. It may also get 3.56 times worse should there be one unit increase of NCS severity grade. As for treatment method, it may result in 1.72 times worse of functional recovery when it switched from surgical to conservative treatment.
On the other hand, in the Model 2, it revealed R-CSA (OR=4.93(1.16), 95%CI= 2.66 – 7.19, p<0.0001), P (OR=1.12(0.17), 95%CI= 0.79 — 1.44, p<0.0001), age (OR=0.07(0.02), 95%CI=-0.11 — -0.025, p<0.05), treatment methods (OR= -1.79(0.35), 95% CI= -2.48 — -1.11, p<0) and NCS severity grade (mild) (OR=4.13(1.28), 95%CI=1.62-6.65, p<0.05) as significant predictors associated with functional recovery. For one unit increase of R-CSA, the probability of functional recovery decreased 3.4.93 times; whereas for one unit increase of P, the probability of functional recovery decreased 1.12 times. It may also witness 4.13 times worse should the NCS severity grade increase one-unit. There was 1.79 times of probability of increasing functional recovery when the conservative treatment was converted into surgical decompression. The performance of bootstrapping method also indicated consistent findings in both models. Conclusively, both models demonstrated consistent performance based on ultrasonographic results.
Discussion
This is the first study to examine predictability of ultrasound for functional recovery after different treatment methods for CTS accompanied by axonal degeneration. By demonstrating overall satisfactory prognostic value of correspondent ultrasound parameters at pre-treatment stage, our findings imply that detection of axonal degeneration by ultrasound is therapeutically instructive for appropriate treatment prescription. It is suggested that surgical treatment should be prioritized when axonal degeneration associated in CTS was detected by ultrasound. Numerous robust evidences have also demonstrated better treatment effectiveness using surgical treatment compared to non-surgical methods [30-34]. In addition, our findings also contributed to clarify the ambiguity of appropriate treatment prescription particularly for those diagnosed below the severe grade whereas potential axonal degeneration may co-exist. Among the enrolled hands in our current study, 82/196 (48.1%) were in fact diagnosed below the severe grade based on Bland’s severity classification system [29]. But all of them exhibited axonal degeneration reflected by ultrasound. This finding was similar with a previous study which found over half of the enrolled subjects within moderate grade were suffering from axonal degeneration [14]. All these findings further reinforced our point of view that detection of axonal degeneration is more clinically informative regardless of severity gradation. Therefore, we propose that early detection of axonal degeneration in CTS by ultrasound can be therapeutically instructive for physicians.
In terms of ultrasound predictors, previous studies suggested CSA as the most sensitive ultrasound parameter to predict postsurgical outcomes [27]. Nevertheless, since male subjects were enrolled in our study, we did not involve measurements at wrist (CSA and P) as predictor in the model due to concern of gender influence on nerve size. Since previous studies revealed better discriminating performance in R-CSA and DCSA [10], therefore, ratio and changes of ultrasound parameters were taken into account for model establishment.
As for other predictors in our models, they overall exhibited consistent performance compared with previous findings. To design the scale of functional recovery, we took Bland [15]’s study for reference. Although we did not find significant relationship between functional recovery and NCS severity grade in our study, correlation between functional recovery and R-CSA (r=0.319, p<0.0001), DML (r=-0.234, p=0.007) and MCV (r=0.233, p=0.011) were still identified, which can also affirm the consistency. Regarding the NCS severity grade, only mild grade was found as significant predictor for functional recovery. Bland [15] illustrated that patients with less severe neurophysiological abnormalities could have better outcome after surgical decompression. Fowler, Munsch [17] also shared similar opinion that patients with mild or moderate CTS based on pre-operative NCS experienced a faster recovery after surgical decompression. Congruent finding was found in our study. As NCS severity grade increases, it is less likely to witness its significant prognostic values.
All these further highligh
ted the importance of detection of axonal degeneration at early stage may help to avoid unnecessary delay of recovery or deterioration by opting to appropriate intervention. Regarding length of follow-up, we also found no correlation between NCS severity grade, which was consistent with the finding in Bland [15]’s study. Therefore, it is unlikely that a systematic bias co-exists in the observed relationship between NCS severity grade and functional recovery. Like many studies, which indicated age as a poor predictor [35, 36] and even with no correlation with functional performance/satisfaction after surgical decompression [35, 37], we also found age as a significant but poor predictor in both models. As for gender, our studies indicated gender as a nonsignificant prognostic factor in the models. Identically, it coincides with several logistic studies which indicated gender has no prognostic value towards functional outcome after treatment [38, 39]. As for comorbidities and surgical history, we identified and categorized the disorder factor associated with CTS into category of CTS-relevant comorbidities based on previous large epidemic studies [40-49]. Meanwhile we clustered types of surgery which are associated/disassociated with correspondent risk factors into CTS-relevant/irrelevant categories. Nevertheless, we didn’t find significant prognostic values of comorbidities and surgical history for functional recovery. We infer it may be due to small sample size, lack of a control population, which is also identical with points of view from previous epidemic studies [50].
Nevertheless, the primary goal of this study is not to compare treatment effectiveness based on detection of axonal degeneration versus current severity classification system. The diagnostic utility of ultrasound should not be overstated as it is fundamentally a morphological assessment tool, which cannot take place of other more accurate assessment methods for confirming axonal degeneration such as biopsy regardless of its invasiveness [51]. In general, we recommend ultrasound as an efficient, convenient and complimentary assessment tool for prescreening CTS with potential axonal degeneration to assist physician’s clinical judgement.
Limitation
There are several limitations in our study. First, we didn’t involve a validated clinical outcome assessment tool. Nevertheless, previous literature demonstrated simply binary responses to measure satisfaction and resolution of symptoms [52]. Statistical significant changes in clinical measurements is not necessarily equivalent to clinical importance [53]. A large minimal clinically important difference was also found in the Levine Carpal Tunnel Questionnaire, which may constrain its use in detecting clinical significant outcome [54]. On the other hand, small sample size, non-randomized study design and lack of a control group may also result in biased result. A further randomized clinical trial with a larger sample size is required to examine patients’ overall function and quality of life can be truly obtained by integrating use of ultrasound in early detection of axonal degeneration.
Conclusion
Ultrasound can be used to predict functional outcome for CTS associated with axonal degeneration regarding different treatment prescription. Detection of axonal degeneration at early stage of CTS can be promising in preventing disease deterioration with better functional outcome after more accurate treatment prescription.