Poor Outcome for Neural Surgery

(Epineurotomy or Neurolysis) for Carpal

Tunnel Syndrome Compared with Carpal
Tunnel Release Alone: A Meta-Analysis of
Global Outcomes
Richard Chapell, Ph.D., Vivian Coates, M.B.A., and Charles Turkelson, Ph.D.
Plymouth Meeting, Pa.

A meta-analysis was performed on the results of eight in their efforts to improve the quality of health
studies that compared the global outcomes of patients
who received carpal tunnel release with the global out-
care in the United States.
comes of patients who received carpal tunnel release and Carpal tunnel syndrome results from com-
neurolysis or epineurotomy. The meta-analysis suggests pression of the median nerve as it passes
that patients who received such neural surgery tended to through the carpal tunnel from the wrist to the
have poorer global outcomes than those who did not
(odds ratio, 0.54; 95 percent confidence interval, 0.32 to
hand, leading to progressive sensory and mo-
0.90). The data are homogenous, and linear-regression tor disturbances. The overall prevalence of car-
analysis indicates that patient attrition did not influence pal tunnel syndrome in the United States may
the outcome of the meta-analysis. The results of this meta- be as high as 1.9 million people, and each year
analysis indicate that neural surgery is potentially harmful
for most patients with carpal tunnel syndrome. The pos-
there are 300,000 to 500,000 operations for the
sibility remains that neural surgery may be helpful in condition, at a total cost of more than $2 bil-
special cases, such as in the presence of marked scarring lion.1 There are no widely accepted figures for
or neural adhesion, but no available evidence specifically what fraction of these cases requires surgical
documents the benefits and harms of surgery among such
patients. (Plast. Reconstr. Surg. 112: 983, 2003.)
intervention. Estimates range from nearly half
of all carpal tunnel syndrome patients with
occupational disease to a “small percentage” of
This article summarizes the findings of a all patients.1
meta-analysis conducted as part of an evidence The standard procedure for the treatment of
report and technology assessment entitled Di- carpal tunnel syndrome is the transection of
agnosis and Treatment of Worker-Related Musculo- the transverse carpal ligament, known as carpal
skeletal Disorders of the Upper Extremity written for tunnel release.2 This operation is typically per-
the Agency for Healthcare Research and Qual- formed in an ambulatory surgical center under
ity. The agency sponsors these reports through regional anesthesia. In addition to the stan-
its Evidence-Based Practice Centers to assist dard operation, adjunctive procedures may be
public-sector and private-sector organizations performed, including synovectomy of the

From ECRI. Received for publication May 30, 2002; revised November 19, 2002.
This article is based on research conducted by the ECRI Evidence-Based Practice Center under contract to the Agency for Healthcare Research
and Quality (contract no. 290-97-0020). ECRI is a nonprofit health services research agency and a Collaborating Center for Healthcare Technology
Assessment of the World Health Organization.
The authors of this article are responsible for its contents, including any clinical or treatment recommendations. No statement in this article
should be construed as an official position of the Agency for Healthcare Research and Quality or the U.S. Department of Health and Human
Services.
DOI: 10.1097/01.PRS.0000076222.77125.1F
983
984 PLASTIC AND RECONSTRUCTIVE SURGERY, September 15, 2003
flexor tendons or a procedure on the median cluded all studies that compared external neu-
nerve (such as neurolysis or epineurotomy). rolysis, internal neurolysis, or epineurotomy to
The term “neurolysis” is used to encompass a control group receiving open carpal tunnel
several different procedures, including both release. This resulted in the inclusion of eight
internal and external neurolysis.3 External trials described in 10 reports: three trials of
neurolysis involves removal of adhesions from epineurotomy, one trial of external neurolysis
the connective tissue surrounding the nerve with flexor synovectomy, and four trials of in-
(the epineurium), thereby relieving pressure ternal neurolysis. The databases searched and
within the epineurium by means of a longitu- the key words used are listed in Table I.
dinal incision, or removal of a segment of the Meta-analyses of studies of neural surgery
epineurium. The terms “epineurotomy” and were conducted by using Hedges’ d as a mea-
“external neurolysis” are often used sure of each study’s effect size and then com-
interchangeably.4 puting the precision-weighted summary d from
The term “internal neurolysis” refers to pro- the combined results of all studies.5 Hedges’ d
cedures that attempt to free individual fascicles is the difference between the means of any
within the nerve from adhesions or compress- study’s two groups expressed in SD units. For
ing scar tissue. In interfascicular neurolysis, a computation of effect sizes derived from di-
common type of internal neurolysis, fascicles chotomous outcomes, we converted the odds
are carefully teased or cut apart and separated ratio to Hedges’ d as described by Hasselblad
from each other. Internal neurolysis is usually and Hedges.6 For computation of effect sizes
performed under an operating microscope. derived from rating scale data, we calculated a
Access to the fascicles may be gained by a total mean for each group as described by Torger-
epineurectomy or by an epineurotomy son.7 An advantage of this method is that it
incision.3 does not assume that all patients employ the
same boundaries for each category in a rating
MATERIALS AND METHODS scale. As an aid to interpretation, effect sizes
We searched 17 bibliographic databases for were also expressed as odds ratios calculated
studies of neural surgery for carpal tunnel syn- according to Hasselblad and Hedges.6 We em-
drome. Each database was searched from its ployed two tests for heterogeneity: the Q statis-
inception through October of 2001. We em- tic and each study’s standardized residual.8 We
ployed broad inclusion criteria. Thus, we in- regarded the data as heterogeneous if the re-

TABLE I
Search Strategy: Databases Searched

Databases searched
CINAHL (Cumulative Index to Nursing and Allied Health Literature; 1967 through October 18, 2001)
ClinicalTrials.gov (through October of 2001)
Cochrane Database of Systematic Reviews (through 2001, issue 4)
Cochrane Registry of Clinical Trials (through 2001, issue 4)
Cochrane Review Methodology Database (through 2001, issue 4)
CRISP (Computer Retrieval of Information on Scientific Projects; through October of 2001)
Database of Reviews of Effectiveness (Cochrane Library; through 2001, issue 4)
ECRI Health Devices Alerts (1977 through October of 2001)
ECRI Health Devices Sourcebase (through October of 2001)
ECRI Healthcare Standards (1975 through October of 2001)
ECRI International Health Technology Assessment (IHTA; through October of 2001)
ECRI Library Catalog (through October of 2001)
ECRI TARGET (Technology Assessment Resource Guide for Emerging Technologies; through October of 2001)
Embase (through October 18, 2001)
PubMed (includes MEDLINE and HealthSTAR; through October 18, 2001)
U.K. National Health Service (NHS) Centre for Reviews and Dissemination (through October of 2001)
U.S. National Guideline Clearinghouse (NGC; through October of 2001)
Terms searched
The search strategies employed a number of freetext keywords as well as controlled vocabulary terms including (but not limited to) the
following concepts:
Controlled trials: randomized controlled trials; controlled clinical trials (MeSH heading, publication type, and textword); meta-analysis;
random allocation; single-blind method; double-blind method, evidence-based medicine (includes randomized controlled trials, outcomes
research, and meta-analysis)
Disorder: carpal tunnel syndrome; CTS; carpal tunnel; median nerve; entrap
Vol. 112, No. 4 / NEURAL SURGERY FOR CARPAL TUNNEL SYNDROME 985
sults of either test were statistically significant. size of the effect was determined by examining
A statistically significant Q value indicates that the effect of attrition on between-group vari-
there are statistically significant differences ance using the QE statistic as described by
among study results. These differences typi- Hedges.14 A statistically significant QE statistic
cally arise when study results are influenced by indicates that the regression has failed to ex-
interstudy design differences or differences in plain a statistically significant amount of the
the types of patients enrolled in them. Regard- differences among study results.
less of its cause, a statistically significant Q The outcome we examined was global out-
value indicates that summarizing the data in come. This was the only patient-oriented
terms of a single meta-analytic summary statis- outcome reported by all eight studies. Global
tic is not appropriate. Similarly, a study with a outcomes attempt to capture the overall result
statistically significant standardized residual is of a treatment. The definitions of global out-
an outlier, which also suggests the presence of comes as assessed for each study are listed in
heterogeneity. Table II.
To further assist in interpreting the results of
our meta-analyses, we present the results of our RESULTS
fixed-effects model in terms of a forest plot and The eight studies in our meta-analysis, along
as a pair of normal curves. Each curve repre- with their study designs and the type of neural
sents the distribution of results in a study’s two surgical procedure they used, are shown in
groups. The difference between the means of Table III. They were described in a total of 10
these two normal curves represents d, the ef- publications. The quality of these trials was
fect size. We quantified the degree of the non- generally poor. One trial was not described as
overlap of these two curves by using the 艛 randomized, and one was retrospective. Two
statistics described by Cohen.9 The degree of others utilized a randomization method that
nonoverlap between groups provides a mea- may not have been truly stochastic,4,10 and only
sure of the proportion of patients who will one provided sufficient information to allow
experience a different outcome from that of readers to conclude that the randomization
the patients in the other group. For example, if method was stochastic. 12 Six trials were
the degree of nonoverlap is 10 percent, then blinded. Five studies experienced patient attri-
the highest-scoring 5 percent of patients will all tion, which ranged from 6 percent to 54 per-
be in one group, and the lowest-scoring 5 per- cent. None of these five studies reported re-
cent of patients will be in the other group. sults on an intent-to-treat basis. Only three of
There was patient attrition in five of the the eight studies in our analysis reported that
studies included in our analysis, ranging from patients had undergone a period of unsuccess-
6 percent to 54 percent.4,10 –13 Attrition can af- ful conservative treatment before surgery.10,11,15
fect a study’s results in unknown ways. To as- All but one study violated statistical assump-
certain whether attrition influenced the results tions of independence by including patients
of our analysis, we conducted a linear-regres- with bilateral carpal tunnel syndrome.18 The
sion analysis. The influence of attrition on the impact of this violation in terms of the number
TABLE II
Global Outcomes

Study Definition
15
Leinberry et al., 1997 Number of patients free of symptoms
Blair et al., 19964 1. Number of patients happy or satisfied with their treatment*
2. Number of patients stating that they would have the surgery again
3. Number of patients reporting permanent total relief, permanent partial relief,
temporary total relief, temporary partial relief, no change, or worse symptoms after
surgery
Foulkes et al., 199410 1. Number of patients cured or improved*
2. Numerical symptom severity rating
Mackinnon et al., 199111 Number of patients reporting relief of all or the majority of their symptoms
Lowry and Follender, 198812 Categorical rating: numbers of patients with excellent, good, fair, or poor results
Gelberman et al., 198716 Number of patients with complete resolution of symptoms and objective signs
Holmgren and Rabow, 198713 Number of patients with complete resolution of symptoms
Freshwater and Arons, 197817 Number of patients with complete resolution of symptoms
* Two studies reported more than one measure of global outcome. In both cases, the outcome used to calculate the effect size is the first one listed.
986 PLASTIC AND RECONSTRUCTIVE SURGERY, September 15, 2003
TABLE III
Type of Evidence

Study Study Design Type of Surgery

Leinberry et al., 199715 Prospective randomized controlled trial Epineurotomy

Blair et al., 19964 Prospective randomized controlled trial Epineurotomy
Mackinnon et al., 199111 Prospective randomized controlled trial Internal neurolysis
Foulkes et al., 199410 Prospective randomized controlled trial Epineurotomy
Lowry and Follender, 198812 Prospective randomized controlled trial Internal neurolysis
Holmgren and Rabow, 198713; Prospective randomized controlled trial Internal neurolysis
Holmgren-Larsson et al., 198518
Freshwater and Arons, 197817 Prospective controlled trial External neurolysis and flexor synovectomy
Gelberman et al., 198716; Rhoades Retrospective controlled trial Internal neurolysis
et al., 198519

of times an erroneous conclusion of statistical ure 1. The overall effect size is significantly
significance was drawn cannot be determined different from zero (d ⫽ ⫺0.34; 95 percent
from the published data, but it does affect the confidence interval, ⫺0.62 to ⫺0.06; p ⫽
confidence one can have in the results of any 0.018). The negative effect size indicates that
statistical analyses. The results of our meta- patients receiving neural surgery generally had
analysis are shown in Table IV. The calculated less favorable global outcomes than did pa-
effect sizes are not heterogeneous (Q ⫽ 2.35; tients not receiving neural surgery. This effect
p ⫽ 0.94; no statistically significant standard- is equivalent to an odds ratio of 0.54 (95 per-
ized residuals). The lack of heterogeneity sug- cent confidence interval, 0.32 to 0.90)
gests that although different researchers per- Although the difference between groups is
formed different types of neural surgery, these statistically significant, there is still consider-
differences did not lead to statistically signifi- able (76.4 percent) overlap between the global
cant differences in global outcomes. Similarly, outcome scores of the two groups (Fig. 2).
the different definitions of global outcome, Another way to express this overlap is to note
different amounts of patient attrition, and dif- that 11.8 percent of patients receiving neural
ferences in follow-up time among studies did surgery will have a less favorable outcome than
not result in statistically significant heterogene- all patients who do not receive this surgery.
ity. It was therefore statistically valid to com- Several points about our results bear men-
bine these studies for meta-analysis. The sizes tion. First, it was only possible to calculate an
of effects in the individual studies as well as the effect size from one (number of patients cured
overall effect size, and their 95 percent confi- or improved) of the two measures of global
dence intervals, are depicted graphically in Fig- outcome reported by Foulkes et al.10 Also, we
TABLE IV
Results of Conservative Meta-Analysis of Global Outcome among Patients Treated with Neural Surgery for Carpal Tunnel
Syndrome

No. of Standardized
Study Patients Effect Size* 95% CI p Odds Ratio Residual†

Leinberry et al., 199715 50 ⫺0.09 ⫺0.69 to 0.52 0.78 0.85 0.92

Blair et al., 19964 75 ⫺0.94 ⫺2.57 to 0.70 0.26 0.18 ⫺0.73
Foulkes et al., 199410 23 ⫺0.30 ⫺2.13 to 1.53 0.75 0.58 0.04
Mackinnon et al., 199111 63 ⫺0.26 ⫺0.99 to 0.46 0.48 0.62 0.22
Lowry and Follender, 198812 47 ⫺0.28 ⫺0.85 to 0.30 0.34 0.60 0.23
Gelberman et al., 198716 69 ⫺0.61 ⫺1.14 to ⫺0.07 0.03 0.33 ⫺1.15
Holmgren and Rabow, 198713 37 0.00 ⫺2.19 to 2.19 1.00 1.00 0.30
Freshwater and Arons, 197817 26 ⫺0.23 ⫺1.43 to 0.97 0.71 0.66 0.18
p Value Overall p Value of
Q‡ of Q Effect Size 95% CI Effect Size Odds Ratio

2.35 0.94 ⫺0.34 ⫺0.62 to ⫺0.06 0.018 0.54

* Effect size is the difference between mean group outcomes measured in SD units.
† Standardized residual is the difference between an individual effect size and the weighted mean of the other effect sizes in the meta-analysis. A standardized
residual greater than 1.96 indicates an outlier.
‡ Q is an omnibus measure of between-study variance. A statistically significant Q indicates that studies are heterogeneous and should not be meta-analytically
combined using a fixed effects model.
Vol. 112, No. 4 / NEURAL SURGERY FOR CARPAL TUNNEL SYNDROME 987

FIG. 1. Results of meta-analysis of the effect of neurolysis on global outcome.

FIG. 2. Overlap between effects of neural surgery and control.

computed an effect size from only one of the come than the number of patients stating they
three measures of global outcome reported by would have surgery again. The latter could be
Blair et al.4 We did not include their data on more of a measure of whether the patient felt
patient perceptions about symptom relief be- that it was worthwhile to attempt any treatment
cause these results were presented in a manner rather than whether they were content with the
that is difficult to quantify. For example, they success of this particular one.
reported that some patients experienced per- We further examined the credibility of our
manent partial relief, while others experienced results by performing a sensitivity analysis. We
temporary total relief. It is difficult to deter- ran the meta-analysis again, omitting the re-
mine which of these outcomes the patients sults of the study by Blair et al., which had
considered superior. Of the remaining two out- yielded the largest (i.e., farthest from zero)
comes in the report by Blair et al., the number effect size. With these results omitted, the re-
of patients who reported that they were happy sults of the remaining studies were homoge-
or satisfied with their treatment was considered nous (Q ⫽ 1.82, p ⫽ 0.94) and the summary
to be a more accurate measure of global out- effect size was statistically significantly different
988 PLASTIC AND RECONSTRUCTIVE SURGERY, September 15, 2003
from zero (d ⫽ ⫺0.32; 95 percent confidence tions because they are satisfied with their con-
interval, ⫺0.60 to ⫺0.03, p ⫽ 0.028). ditions, this may make the treatment appear
The study by Gelberman et al.16 was the only less effective. When analyzing study results, the
retrospective study in the meta-analysis and was analyst may attempt to compensate for patient
the only individual study to report a statistically attrition by assuming that treatment had failed
significant effect. When the meta-analysis was (or succeeded) for all missing patients.
rerun with the results of Gelberman omitted, In this meta-analysis, no attempt was made to
the remaining results were homogenous (Q ⫽ compensate for missing patients for two rea-
1.02, p ⫽ 0.98). However, the overall effect size sons. First, many of the published reports did
was not statistically significant (d ⫽ ⫺0.24, 95 not identify the group assignment of the miss-
percent confidence interval, ⫺0.57 to 0.09, p ⫽ ing patients. Second, some studies had dramat-
0.16). Thus, the inclusion of this lower-quality ically more attrition from one group than the
study in the meta-analysis is necessary to other. Adding a greater number of treatment
achieve statistical significance. However, this failures or successes to one group than to the
may be due to the increased statistical power of other can cause dramatic differences in the
the meta-analysis when this study is included calculated effect size.
and not due to biased results in this study. That
Rather than attempt to compensate mathe-
this study did not have biased results is, in turn,
matically for missing patients, we used linear
suggested by the fact that its results were not
regression to examine the influence of patient
different from those of the other studies in the
meta-analysis. This latter point is supported by attrition on the calculated effect size from each
the results of our homogeneity testing and by study. Figure 3 depicts a plot of effect size
examination of Figure 1. versus percent attrition for the eight studies.
Also important is that a result that is statisti- No clear relationship can be discerned. The
cally not significant indicates that there is no slope of the regression line (␤1 ⫽ ⫺0.06) is not
detectable difference between groups. While significantly different from zero (p ⫽ 0.96). In
such a result does not support the conclusion addition, the QE statistic is also not significant
that neural surgery is harmful, it also does not (QE ⫽ 2.35, p ⫽ 0.89), indicating that there is
lead to the conclusion that neural surgery is no statistically significant variation among the
beneficial. results of these studies that is not accounted for
Another factor that may influence the out- by this regression line (given that the Q statistic
come of a study is patient attrition. If patients in this meta-analysis was not significant, this
are so dissatisfied with their treatment that they finding is not surprising). Similar regression
refuse to continue to participate in a study, analyses (not shown) found no effect of fol-
their omission may make the treatment appear low-up time (QE ⫽ 2.13, p ⫽ 0.91) or percent-
to be more effective than it is. Conversely, if age of patients with bilateral carpal tunnel syn-
patients fail to return for follow-up examina- drome (QE ⫽ 2.19, p ⫽ 0.90).

FIG. 3. Effect of patient attrition on effect size.

Vol. 112, No. 4 / NEURAL SURGERY FOR CARPAL TUNNEL SYNDROME 989
DISCUSSION hands could be in this group. We chose to
Our results suggest that patients who receive assume that there were 14 hands in this group,
neurolysis report worse global outcomes than thereby biasing the outcome in favor of neural
those who do not receive it. Nevertheless, attri- surgery. The fact that we obtained a statistically
tion was present in most of the trials we exam- significant negative effect size despite this bias
ined. Thus, results were not reported for 20 tends to strengthen our conclusion that neural
percent of the 494 patients who were enrolled surgery is harmful.
in these studies. Our regression analysis sug- It is possible that our results do not extend to
gests that the results of our meta-analyses are all patients with carpal tunnel syndrome. The
complication/failure rate for carpal tunnel re-
not likely to be overturned by the results of
lease surgery is approximately 3 percent to 19
“missing” data.
percent. Many of these cases require re-
In the study by Freshwater and Arons,17 pa-
exploration. For these patients, external neu-
tients in the experimental group received
rolysis (nerve mobilization) has been advocat-
flexor synovectomy in addition to external neu-
ed.21 In some cases, the surgical incision may
rolysis. Because patients in the control group
reveal extensive scarring, adhesions, and con-
did not receive flexor synovectomy, there is no striction of the median nerve by soft tissues in
way to determine whether differences between the area of the carpal tunnel. In such cases, it
groups are the result of the neural or the ten- may be advisable to perform some form of
don surgery. It is therefore possible that the neural surgery to address these complications.
poorer global outcomes among patients However, there are no clinical trials that have
treated with neural surgery in this group may addressed whether there are any special cir-
be the result of the flexor tenosynovectomy cumstances under which neural surgery should
rather than the neural surgery. However, the be performed.
effect size calculated from the results of this
study led to a small standardized residual CONCLUSIONS
(0.18), indicating that this study did not exert
a large influence on the overall effect size. Patients receiving neural surgery tend to re-
Also problematic is that many studies vio- port significantly worse global outcomes than
lated the statistical assumption of indepen- patients who do not undergo such surgeries at
dence. This occurred because more than one the time of carpal tunnel release. The possibil-
ity remains that neural surgery may be helpful
procedure was performed on a single patient.
in special cases, such as in the presence of
All of the studies assessed here included pa-
marked scarring or neural adhesion, but no
tients who received bilateral procedures, with
available published evidence specifically docu-
the exception of the study by Holmgren et
ments the benefits and harms of neurolysis
al.13,18 Thus, in all but one study, assumptions
among such patients.
of independence between and within groups
Richard Chapell, Ph.D.
were violated. This can lead to underestima- ECRI
tion of standard errors and spurious statisti- 5200 Butler Pike
cally significant results (type I errors).20 The Plymouth Meeting, Pa. 19462
extent of the violation depends on the percent- [email protected]
age of patients with bilateral procedures. The
more bilateral patients included, the more se- REFERENCES
vere the violation. However, linear-regression 1. Palmer, D. H., and Hanrahan, L. P. Social and eco-
analysis found that differences between studies nomic costs of carpal tunnel surgery. Instr. Course Lect.
were not influenced by the percentage of bilat- 44: 167, 1995.
eral patients included in each study. 2. Boggins-Magill, M. K. Carpal tunnel release: Scoping
out the carpal tunnel. Today’s OR Nurse 16: 27, 1994.
The study by Gelberman et al. was among 3. Frykman, G. K., Adams, J., and Bowen, W. W. Neurol-
those that included patients with bilateral car- ysis. Orthop. Clin. North Am. 12: 325, 1981.
pal tunnel syndrome. In the control group, 18 4. Blair, W. F., Goetz, D. D., Ross, M. A., Steyers, C. M., and
patients with 22 affected hands were cured. In Chang, P. Carpal tunnel release with and without
the neural surgery group, 10 patients were epineurotomy: A comparative prospective trial.
J. Hand Surg. (Am.) 21: 655, 1996.
cured, but the number of hands cured was not 5. Hedges, L. V., and Olkin, I. Statistical Methods for Meta-
reported. From information given in the text Analysis. Boston: Academic Press, 1985.
of the report,19 as few as 11 or as many as 14 6. Hasselblad, V., and Hedges, L. V. Meta-analysis of
990 PLASTIC AND RECONSTRUCTIVE SURGERY, September 15, 2003
screening and diagnostic tests. Psychol. Bull. 117: 167, L. V. Hedges (Eds.), The Handbook of Research Synthesis.
1995. New York: Russell Sage Foundation, 1994.
7. Torgerson, W. S. The law of categorical judgment. In 15. Leinberry, C. F., Hammond, N. L., III, and Siegfried, J. W.
Theory and Methods of Scaling, 6th Ed. New York: John The role of epineurotomy in the operative treatment
Wiley & Sons, 1958. of carpal tunnel syndrome. J. Bone Joint Surg. (Am.) 79:
8. Shadish, W. R., and Haddock, C. K. Combining esti- 555, 1997.
mates of effect size. In H. Cooper and L. V. Hedges 16. Gelberman, R. H., Pfeffer, G. B., Galbraith, R. T., Szabo,
(Eds.), The Handbook of Research Synthesis. New York: R. M., Rydevik, B., and Dimick, M. Results of treat-
Russell Sage Foundation, 1994. ment of severe carpal-tunnel syndrome without inter-
9. Cohen, J. Statistical Power Analysis for the Behavioral Sci- nal neurolysis of the median nerve. J. Bone Joint Surg.
ences, 2nd Ed. Hillsdale, N.J.: Lawrence Erlbaum As- (Am.) 69: 896, 1987.
sociates, 1988. 17. Freshwater, M. F., and Arons, M. S. The effect of various
10. Foulkes, G. D., Atkinson, R. E., Beuchel, C., Doyle, J. R., adjuncts on the surgical treatment of carpal tunnel
and Singer, D. I. Outcome following epineurotomy syndrome secondary to chronic tenosynovitis. Plast.
in carpal tunnel syndrome: A prospective, random- Reconstr. Surg. 61: 93, 1978.
ized clinical trial. J. Hand Surg. (Am.) 19: 539, 1994. 18. Holmgren-Larsson, H., Leszniewski, W., Linden, U.,
11. Mackinnon, S. E., McCabe, S., Murray, J. F., et al. In- Rabow, L., and Thorling, J. Internal neurolysis or
ternal neurolysis fails to improve the results of primary ligament division only in carpal tunnel syndrome: Re-
carpal tunnel decompression. J. Hand Surg. (Am.) 16: sults of a randomized study. Acta Neurochir. (Wien) 74:
211, 1991. 118, 1985.
12. Lowry, W. E., Jr., and Follender, A. B. Interfascicular 19. Rhoades, C. E., Mowery, C. A., and Gelberman, R. H.
neurolysis in the severe carpal tunnel syndrome: A Results of internal neurolysis of the median nerve for
prospective, randomized, double-blind, controlled severe carpal-tunnel syndrome. J. Bone Joint Surg. (Am.)
study. Clin. Orthop. 227: 251, 1988. 67: 253, 1985.
13. Holmgren, H., and Rabow, L. Internal neurolysis or 20. Newcombe, R. G., and Duff, G. R. Eyes or patients?
ligament division only in carpal tunnel syndrome: II. Traps for the unwary in the statistical analysis of oph-
A 3-year follow-up with an evaluation of various neu- thalmological studies. Br. J. Ophthalmol. 71: 645, 1987.
rophysiological parameters for diagnosis. Acta Neuro- 21. Botte, M. J., von Schroeder, H. P., Abrams, R. A., and
chir. (Wien) 87: 44, 1987. Gellman, H. Recurrent carpal tunnel syndrome.
14. Hedges, L. V. Fixed effects models. In H. Cooper and Hand Clin. 12: 731, 1996.