ULTRASOUND IMAGING OF THE MEDIAN NERVE AS A PROGNOSTIC

FACTOR FOR CARPAL TUNNEL DECOMPRESSION

JEREMY D.P. BLAND, MB, ChB1 and STEPHAN M. RUDOLFER, PhD2
1
Department of Clinical Neurophysiology, Kent and Canterbury Hospital, Ethelbert Road, Canterbury, Kent CT1 3NG, UK
2
Centre for Biostatistics, Institute of Population Health, University of Manchester, Manchester, UK
Accepted 13 August 2013

ABSTRACT: Introduction: The diagnostic value of ultrasound
imaging in carpal tunnel syndrome is established, but reports
on its prognostic value have been contradictory. Methods: This
investigation was an observational study of subjective surgical
results, evaluated by symptom severity and functional status
scales, and an ordinal scale for overall outcome, for 145 carpal
tunnel decompressions in relation to preoperative measurement
of median nerve cross-sectional area. Results: The surgical
success rate was 86%. In univariate analyses no significant
correlation existed between outcome and preoperative cross-
sectional area, nor with preoperative nerve conduction studies
or patient variables, except for body mass index and gender. A
multivariate model including electrophysiological, imaging, and
patient variables was moderately predictive of success with an
area under the receiver operating characteristic curve of 0.82.
Conclusions: Cross-sectional area alone is unlikely to be a
sufficiently reliable predictor of outcome for use in counseling
individual patients, but imaging results may be useful in multi-
variate prognostic models.
Muscle Nerve 49: 741–744, 2014
Ultrasound imaging of the median nerve has been
shown to have utility in the diagnosis of carpal tun-
nel syndrome (CTS), with sensitivity and specificity
levels approaching those of electrodiagnostic stud-
ies.1 Concentration on the diagnostic sensitivity and
specificity of the 2 types of investigation has dis-
tracted attention from other reasons for their use.
The clinical diagnosis of CTS is usually easy, and the
major role of investigation is to aid management by
helping to elicit etiology, discover concurrent dis-
ease, and measure severity. The outcome of carpal
tunnel decompression is related to preoperative
severity of changes in nerve conduction studies
(NCS),2 although the relationship is complex. To
date, most studies of ultrasound imaging in relation
to surgery have addressed the issue of whether pre-
operative enlargement of the median nerve
decreases after surgery. One study, for example, sug-
gested that smaller nerves after surgery are corre-
lated with better outcomes.3 However, the
prognostic value of ultrasound has been assessed in
only 3 studies. One study of 112 wrists showed the
Abbreviations: BMI, body mass index; CSA, cross-sectional area; CTS,
carpal tunnel syndrome; FSS, Functional Status Scale; NCS, nerve con-
duction studies; ROC, receiver operating characteristic; SSS, Symptom
Severity Scale
Key words: carpal tunnel syndrome; multivariate modeling; prognosis;
surgery; ultrasound imaging
Correspondence to: J.D.P. Bland; e-mail: jeremy.bland@nhs.net
C 2013 Wiley Periodicals, Inc.
V
Published online 27 August 2013 in Wiley Online Library (wileyonlinelibrary.
com). DOI 10.1002/mus.24058
Ultrasound Prognosis in CTS
preoperative cross-sectional area (CSA) of the
median nerve at the carpal tunnel outlet to be a bet-
ter predictor of outcome than NCS, with larger
nerve CSA measurements being predictive of good
outcomes.4 The second study (88 wrists) showed no
correlation between the preoperative ultrasound
and surgical success.5 Finally, an Italian study of 67
patients showed that smaller pre-operative CSA at
the tunnel inlet was predictive of full patient satis-
faction with the outcome of surgery, normalization
of Boston Questionnaire scores, and normalization
of the postoperative CSA measurement, although
the lower limits of the 95% confidence intervals for
all of these variables were near 1.6
Many other prognostic factors have been identi-
fied for carpal tunnel surgery, including age,7
occupational status,8 and the presence of coinci-
dent disease.7 Evaluating the predictive value of
individual preoperative factors therefore requires
multivariate analyses with large numbers of
patients. Not all poor surgical outcomes are pre-
dictable; some will result from surgical errors or
the coincidental development of another disease
unrelated to the patient’s preoperative state. How-
ever, patients who contemplate surgery should be
given the most accurate prediction possible of
their chances of success.
METHODS
In this observational study we assessed anony-
mized routine clinical records. Institutional review
board approval was not required. Patients who
attended a dedicated CTS clinic between 2007 and
2011 and had ultrasound imaging were followed
up, as were all patients who attend this clinic, to
assess the outcome of treatment. We present data
on those treated surgically and in whom the out-
come of surgery was assessed, as recorded in the
computerized clinic records. It is routine practice
in this clinic to obtain patient-reported outcomes
of treatment 3–6 months after surgery, so that the
data for this study can be available without further
inconvenience to the patients. Ultrasound scan-
ning has been performed on some patients who
attended the clinic since 2007, initially a random
selection of patients while we were becoming famil-
iar with the technique. More recently, although we
have continued to scan patients opportunistically
whenever there is time to add ultrasound
MUSCLE & NERVE May 2014 741
scanning, we have also made an effort to image
patients with atypical clinical features, such as dis-
proportionately severe pain with mild NCS abnor-
malities, and those being referred for surgery.
Comparisons of the ultrasound measurements
made in the early part of the study with those
made in the last year suggest that there has been
no significant change in expertise or technique of
the ultrasonographer during this period.
The computerized clinic records include data
on prognostic factors such as age, gender, body
mass index (BMI), occupation, smoking, length of
history, and presence of diabetes. Patients com-
pleted the Boston CTS-specific Symptom Severity
Scale (SSS) and Functional Status Scale (FSS),9
both at the time of neurophysiological assessment
and at follow-up after surgery, and we also ask
patients to rate the overall outcome of surgery on
a simple 5-point ordinal scale from “worse” to
“completely cured.”
NCS in the CTS care pathway were carried out
to AANEM standards, and the neurophysiological
severity of CTS detected was recorded using the
Canterbury Severity Scale.10 Ultrasound scanning
was performed using a 7–14-MHZ small-footprint
linear-array transducer (Sonosite, Inc., Bothell,
Washington). One unblinded sonographer (J.B.)
performed all imaging. The median nerve was
imaged longitudinally and in cross-section from
palm to forearm and, for the purposes of this
study, the largest nerve cross-sectional area (CSA)
immediately proximal to the carpal tunnel was
measured by the method of continuous tracing
just inside the hyperechoic epineurium. Where the
nerve was already dividing into 2 branches proxi-
mal to the wrist, both branches were measured sep-
arately and the areas summed. In all patients in
this series the site of greatest nerve enlargement
was proximal to the carpal tunnel.
RESULTS
Since April 2007, 12,857 sets of nerve conduc-
tion studies were performed for CTS, and 1223 of
the studies were accompanied by ultrasound
images. There were 1146 patients who had surgery
performed in the local care pathway for CTS. Two
hundred thirty hands with ultrasound imaging had
subsequently been decompressed surgically, and
postoperative patient-reported outcomes were avail-
able for 150 hands at the time of data extraction.
Two operations proved to be reexplorations after
previous failed surgery, and in 3 patients there was
good evidence that the reason for surgical failure
was incomplete section of the transverse carpal lig-
ament. These 5 patients were excluded from fur-
ther analysis, leaving 145 operations in 124
patients for analysis (21 bilateral operations).
742 Ultrasound Prognosis in CTS
Forty-two operations were performed on men and
103 on women (mean age 61 years, range 31–95
years). The majority of hands had moderate to
severe CTS (Canterbury Scale score: grade 0: 4
hands; grade 1: 9 hands; grade 2: 17 hands; grade
3: 29 hands; grade 4: 24 hands; grade 5: 60 hands;
grade 6: 2 hands). The mean median nerve CSA of
the operated hands before surgery was 12.97 mm2
(standard deviation 4.08 mm2, range 5–25 mm2).
Patients were asked to rate the overall outcome of
surgery for each hand operated separately and, on
these overall ratings, 55 operations (38%) were
reported as a complete cure, 70 much improved
(48%), 12 slightly improved (8%), and 7 worse
after surgery (5%). One patient did not provide an
overall rating but did complete the SSS and FSS.
No patients considered themselves unchanged. We
consider the desired outcome of carpal tunnel
decompression to be a patient overall rating of
“cured” or “much improved” and, on this basis,
125 (86%) of the operations could be considered
successful.
Women showed greater improvement than
men. The mean change in SSS was 21.77 for
women and 21.42 for men (P 5 0.05), and for
FSS the changes were: women 5 21.10 and men
5 20.66 (P 5 0.01). The overall success rate was
88% in women and 78% in men.
In this relatively small data set we were unable
to show significant relationships in univariate anal-
yses between outcomes and the preoperative neu-
rophysiological grade, operating surgeon, age,
previous response to steroid injection, occupation
(coded as manual, non-manual, or retired), dura-
tion of history, smoking status, or diabetes (9
patients). There was a weak but significant correla-
tion between improvement in SSS and BMI, with
heavier patients showing greater improvement (P
< 0.05, r 5 20.16)
The mean preoperative CSA for successful
operations was 13.1 mm2, and for unsuccessful pro-
cedures it was 11.9 mm2 (P 5 0.2). There was no
clear relationship between CSA and the change in
SSS or FSS with surgery (Figs. 1 and 2).
Multivariate analyses were performed using
logistic regression models as predictors of the
binary outcome of success (cured or much
improved) or failure. To address the difficulty of
including 2 hands from some patients in fitting
the logistic regression models, we treated patient
ID as a cluster variable, causing the model to treat
the 2 hands as being correlated when they were 2
sides of the same patient. The result of such mod-
eling is to reduce the standard errors of the varia-
bles, thus making the model more conservative
(needing larger values to achieve significance).
When only 1 hand was involved, the clustering
MUSCLE & NERVE May 2014
FIGURE 1. Relationship of preoperative ultrasound imaging to
change in symptom severity score after surgery. Negative val-
ues for the SSS indicate improved symptoms. Pearson regres-
sion with 95% confidence limits shown (r 5 0.038, P 5 0.64).
made no difference. The performance of each
model was assessed by constructing receiver operat-
ing characteristic (ROC) curves. The best resulting
model (that with the greatest area under the ROC
curve) included age, BMI, CSA, SSS, FSS, symptom
duration, neurophysiological grade, occupation,
gender, side, and smoking status. The ROC curve
for this model is shown in Figure 3.
Two variables (diabetes and surgeon) were
dropped from the model. There were 9 diabetics
among the patients studied, all with successful out-
comes. This deterministic aspect of diabetes in pre-
dicting successful outcome means that, for
technical reasons, any statistical model using it as a
predictor must exclude those patients with diabe-
tes. This would have reduced the number of hands
used in the model fitting to 136, an unsatisfactory
situation. There were 3 surgeons in this study. One
only operated on 2 carpal tunnels, both of which
FIGURE 2. Relationship of preoperative ultrasound imaging to
change in FSS score after surgery. Negative for the FSS indi-
cate improved symptoms. Pearson regression with 95% confi-
dence intervals shown (r 5 0.124, P 5 0.14).
Ultrasound Prognosis in CTS
FIGURE 3. ROC curve for an 11-variable regression model
attempting to predict success (“complete cure” or “much
improved” on subjective report) of carpal tunnel decompression.
Area under curve 5 0.82.
resulted in unsuccessful outcomes. As with diabe-
tes, this would have necessitated dropping 2 hands
from the analysis. Of the remaining surgeons, 1
did twice as many wrists as the other (103 vs. 40),
yet both had broadly similar success rates (89.3%
vs. 82.5%, respectively). For these reasons, the sur-
geon variable was dropped from the model. The
11 variables left in the model fell into 2 groups,
those increasing the probability of operational suc-
cess and those decreasing this probability. Varia-
bles that increased the probability of success were
higher BMI, non-manual occupation (or retired),
right side surgery as opposed to left side, higher
neurophysiological grade, and larger CSA. Those
that decreased the probability of success were
higher FSS, older age, male gender, smoking,
higher SSS, and longer symptom duration.
Another aspect of model fitting is the signifi-
cance of each covariate within the overall model.
This is traditionally measured by the P-value—a
small P-value indicates a significant contribution to
the model and a large P-value (near 1) indicates
the opposite. The formal significance level for P-
values is usually 5%; in this study, none achieved
this level. The lowest P-value was 9.9%, and the
highest was 94.5%. The formal lack of significance
can be attributed to the relatively small sample size
for 11 covariates, yet they may be ordered in terms
of P-value to give an indication of their relative
importance in the model. In increasing order of
magnitude of P-value, the variables were found to
be FSS (9.9%), age (12.1%), BMI (13.2%), male
(29.7%), retired (36.5%), non-manual occupation
(37.9%), right hand (38.1%), present smoker
(48.5%), neurophysiological grade (50.8%), CSA
(73.8%), SSS (78.5%), previous smoker (90.5%),
and symptom duration (94.5%).
MUSCLE & NERVE May 2014 743
DISCUSSION
Predictions of the subjective outcome of sur-
gery will never be perfect, because random events
subsequent to the time at which one is trying to
make the prediction can affect the patient’s
impressions of the result. It is clear, however, that
there are preoperative factors associated with
poorer outcomes, and patients can be warned
when their surgical prospects are less than optimal.
The factors showing the best predictive value so far
appear to be measures of functional impairment,
the NCS and the preoperative FSS score, and this
is consistent with the advice usually given to
patients before carpal tunnel decompression.
Patients are told that they can expect positive
symptoms of pain and tingling to resolve, but
established loss of function, numbness, and weak-
ness of the thenar muscles may persist.
In this small series we were unable to confirm
the findings of previous studies reporting either
small6 or large4 CSA measurements to be positive
predictive factors for surgical outcome. The various
findings in the 4 studies suggest that the CSA mea-
surement used in isolation is unlikely to be a suffi-
ciently strong predictor of outcome to be of utility
in counseling the individual patient. Similarly,
none of the other preoperative variables shown to
be correlated with surgical outcome in large stud-
ies have sufficient predictive value when used in
isolation to be useful to a patient considering sur-
gery. The multivariate analysis, however, suggests
that there may be considerable predictability to
surgical outcomes when multiple preoperative fac-
tors are taken into account, although this will
require further prospective study. It remains possi-
ble that the imaging findings will play a part in the
multivariate predictive model, and it is also possi-
ble that other imaging parameters, such as Dopp-
ler assessment of nerve vascularity or a measure of
nerve mobility, could contribute to such a model.
This study has some limitations. The data are
observational from a clinic in which patients with a
good prognosis are selected for surgery and thus
have a generally high rate of satisfactory outcomes
(e.g., 86% success compared with 63% in the study
by Naranjo et al.4). The very fact that all patients
had been referred to a specialist CTS clinic will
introduce an initial selection bias, and they may
not be representative of the entire population of
patients with CTS. Although the ultrasound find-
ings themselves were not used consciously in the
decision on whether or not to operate, there was
744 Ultrasound Prognosis in CTS
no attempt to make this decision blind to the
imaging results, and the selection of which cases to
collect ultrasound images from was not random.
Indeed, there is a deliberate selection bias in choos-
ing patients with atypical clinical features and, in
the more recent cases, those opting for immediate
surgical treatment for scanning. The relative rarity
of poor outcomes in this population makes it more
difficult statistically to predict them. The analysis
was carried out by operation rather than by patient,
with 21 patients included twice, once for each
hand. This increased the likelihood of type 1 statis-
tical errors, but likely did not obscure the lack of a
strong relationship between imaging results and sur-
gical outcome. The multivariate regression model
was generated from the same data on which we are
reporting its predictive performance and may be
overfitted to this particular data set. It may not per-
form as well with new data.
Despite these limitations this is the largest study
of this topic to date and has the advantage of
describing a normal clinic population without arti-
ficial exclusions so that the conclusions may be
generalized to routine practice. Further work is
required to develop useful prognostic models for
carpal tunnel surgery.
The data in this study were presented in part at the 59th annual
meeting of the American Association of Neuromuscular & Electro-
diagnostic Medicine, October 2012, Orlando, Florida.
REFERENCES
1. Fowler JR, Gaughan JP, Ilyas AM. The sensitivity and specificity of
ultrasound for the diagnosis of carpal tunnel syndrome: a meta-analy-
sis. Clin Orthop Relat Res 2011;469:1089–1094.
2. Bland JDP. Do nerve conduction studies predict the outcome of car-
pal tunnel decompression? Muscle Nerve 2001;24:935–940.
3. Vogelin E, Nuesch E, Juni P, Reichenbach S, Eser P, Ziswiler H-R.
Sonographic follow-up of patients with carpal tunnel syndrome
undergoing surgical or nonsurgical treatment: prospective cohort
study. J Hand Surg 2010;35A:1401–1409.
4. Naranjo A, Ojeda S, Arana V, Baeta P, Fernandez-Palacios J, Garcia-
Duque O, et al. Usefulness of clinical findings, nerve conduction
studies and ultrasonography to predict response to surgical release in
idiopathic carpal tunnel syndrome. Clin Exp Rheumatol 2009;27:
786–793.
5. Smidt MH, Visser LH. Carpal tunnel syndrome: clinical and sono-
graphic follow-up after surgery. Muscle Nerve 2008;38:987–991.
6. Mondelli M, Filippou G, Aretini A, Frediani B, Reale F. Ultrasonogra-
phy before and after surgery in carpal tunnel syndrome and relation-
ship with clinical and electrophysiological findings. A new outcome
predictor? Scand J Rheumatol 2008;37:219–224.
7. Atroshi I, Johnsson R, Ornstein E. Patient satisfaction and return to
work after endoscopic carpal tunnel surgery. J Hand Surg 1998;23A:
58–65.
8. Katz JN, Keller RB, Fossel AH, Punnett L, Bessette L, Simmons BP,
et al. Predictors of return to work following carpal tunnel release.
Am J Ind Med 1997;31:85–91.
9. Levine DW, Simmons BP, Koris MJ, Daltroy LH, Hohl GG, Fossel
AH, et al. A self-administered questionnaire for the assessment of
severity of symptoms and functional status in carpal tunnel syn-
drome. J Bone Joint Surg 1993;75A:1585–1592.
10. Bland JDP. A neurophysiological grading scale for carpal tunnel syn-
drome. Muscle Nerve 2000;23:1280–1283.
MUSCLE & NERVE May 2014