Background

Diagnosis of carpal tunnel syndrome (CTS) is usually based on a combination of symptoms, signs and findings from neurophysiological investigations. Demonstration of impaired distal sensory nerve conduction (SNC) in the median as compared with the ulnar or radial nerve is generally regarded as the most sensitive neurophysiological marker for the disorder[1]. However, definition of when median SNC is abnormal has been somewhat arbitrary.

We recently examined the relationship of median SNC to symptoms and signs in a consecutive series of patients, who were being investigated for suspected CTS at a general hospital[2]. Even for the combination of symptoms and signs which was associated with the greatest mean reduction in median SNC, there was overlap of the distribution of SNC velocities with that in “normal” hands which had no symptoms or signs. The clearest discrimination from normal hands was achieved with a cut-point of 8 m/s for the difference between little and index finger SNC velocities, and on this basis, we proposed that with the method of testing employed, a value of > 8 m/s was a reasonable definition for an abnormal difference between SNC velocities in the little and index fingers (this would include hands in which no signal could be detected when the index finger was tested, indicating extreme impairment of conduction).

In the absence of a satisfactory diagnostic gold standard, the validity of a case definition is best judged by its practical utility in discriminating categories of illness which might benefit from different preventive strategies (i.e. with different risk factors), or different clinical management (i.e. with different responses to treatment or prognosis)[3]. Using the same case series, we have shown that our definition of abnormal SNC distinguished patient groups which differed importantly in their exposures to risk factors[4]. For example, those with abnormal SNC had markedly higher body mass index (BMI) than controls, whereas no association with BMI was apparent for the group with normal SNC. We here present findings from a follow-up of the earlier study, in which we examined whether our definition of abnormal SNC distinguished patients with different response to surgical treatment.

Methods

The selection and recruitment of patients has been described previously[2]. They comprised a consecutive series of men and women aged 20–64 years, who attended the neurophysiology department at Southampton General Hospital during January 2007 to September 2009 for investigation of suspected CTS.

At baseline and before the results of nerve conduction studies were known, those who agreed to take part in the study were asked to complete a self-administered questionnaire. Among other things, this covered sex, age, height, weight, smoking habits, somatising tendency, mental health, diabetes, and the occurrence of symptoms during the past month in each hand (numbness, tingling and pain) and in the neck (pain). Information on height and weight was used to derive body mass index (BMI) in kg/m2. Questions about somatising tendency were taken from the Brief Symptom Inventory[5], and patients were classified according to the number of common physical symptoms from a total of five (faintness or dizziness, pains in the heart or chest, nausea or upset stomach, trouble getting breath, and hot or cold spells) that had been at least moderately distressing in the past week. Mental health was assessed using questions from the relevant domain of the Short Form-36 (SF-36) Questionnaire[6], and scores were grouped to approximate thirds of the distribution in the study sample (good, intermediate and poor). Where patients reported numbness or tingling in a hand, they were asked to mark the anatomical distribution on a diagram. From this, the distribution was classified according to whether it included most of the area lying within the sensory distribution of the median nerve (“extensive median”), a smaller part of this area (“limited median”), or only other parts of the hand (“other”)[2].

Also at baseline, and again before the results of neurophysiological tests were known, a research nurse (CL) carried out a standardised physical examination of the hands, which included Tinel’s and Phalen’s tests. Details of the methods used are reported elsewhere[2].

Nerve conduction studies were carried out by a physician or clinical physiologist with a Nicolet machine, according to the normal practice of the department. Among other things, measurements were made of orthodromic SNC from the index and little fingers to the wrist, with surface recordings over the median or ulnar nerves proximal to the distal wrist crease. Median SNC was deemed to be abnormal if there was no detectable signal when the index finger was tested, or if the difference in SNC velocity between the little and the index finger was > 8 m/s.

At the time of completing the baseline questionnaire, participants were asked whether they would be willing to receive a follow-up questionnaire. Those who agreed were sent a further, shorter questionnaire by post after an interval of approximately 18 months (followed by a reminder to those who did not respond). This asked about treatment for hand and arm symptoms since baseline (surgery on the wrist or hand; injections; and physical therapy), and about numbness, tingling and pain in the hands during the past four weeks.

Statistical analysis was carried out with Stata version 11.1 software. As in our earlier report[2], analysis was based on hands, and was restricted to those which had not been treated surgically for CTS before baseline. We also excluded hands in which there had been no numbness, tingling or pain in the four weeks before the baseline questionnaire was answered.

For the hands on which follow-up information was obtained, we first used Poisson regression to explore baseline predictors of subsequent surgery. To account for the hierarchical structure of the data, we employed multi-level modelling, clustering by patient. Associations were summarised by prevalence rate ratios (PRRs) with associated 95% confidence intervals (CIs).

Next, we examined the frequency with which symptoms had resolved by follow-up (i.e. had been absent throughout the four weeks before answering the follow-up questionnaire) according to baseline characteristics and treatments received during the follow-up period. Two parallel sets of multilevel Poisson regression analyses were carried out, the first with resolution of all sensory symptoms (numbness, tingling and pain) as the outcome, and the second with resolution of numbness and tingling as the outcome. The second set of analyses was restricted to hands in which numbness or tingling had been reported at baseline.

Finally, we compared the association of surgical treatment with symptom resolution according to the presence or absence of abnormal median SNC and other clinical features at baseline. The statistical significance of differences in association was quantified by examining the interaction between the baseline feature and surgical treatment in a regression analysis which took resolution of symptoms as the dependent variable.

Ethical approval for the study was provided by the Southampton and South West Hampshire NHS Research Ethics Committee.

Results

After exclusion of patients who declined to take part in the study (27%) and 10 hands that had already been treated surgically for CTS, our previous analysis was based on 1806 hands in 905 patients[2]. They included 1506 hands with numbness, tingling or pain in 901 patients, 704 of whom agreed to receive a follow-up questionnaire (Figure 1). Among these 704 patients, 394 (56%) completed the follow-up questionnaire, giving a total of 656 originally symptomatic hands that could be included in the current analysis (both hands from 262 patients and one hand from 132 patients). Numbness or tingling had been reported in 633 of the hands at baseline, while the other 23 hands had been painful but without numbness or tingling.

Figure 1
figure 1

Hands and patients analysed and reasons for exclusions.

Full size image

Table 1 shows the numbers of hands analysed according to characteristics assessed at baseline. Response rates at follow-up tended to be higher in women than in men, and at older ages. However, there were no important differences in the completeness of follow-up data according to nerve conduction or other clinical findings at baseline. Among the 394 patients who provided follow-up data, the mean interval from baseline to follow-up was 19.2 months (median 18.9, range 15.0 to 31.0, all but four patients between 17 and 24 months).

Table 1 Number of hands analysed according to characteristics assessed at baseline
Full size table

In total, 154 hands (23%) had been treated surgically during the follow-up period, 47 (7%) by injection (of which 18 were also treated surgically), and 88 (13%) by physical therapies (of which 31 were also treated surgically). The strongest baseline predictor of subsequent surgery was abnormal median SNC (PRR after adjustment for other risk factors 3.6, 95% CI 2.3-5.5) (Table 2). After allowance for neurophysiological abnormality, surgery was also more common when Phalen’s test was positive (PRR 2.4, 95% CI 1.4-4.0). However, associations with numbness/tingling in an extensive median distribution, and with positive Tinel’s test, both disappeared after adjustment for other predictors.

Table 2 Baseline predictors of carpal tunnel surgery
Full size table

Overall, 241 hands (37%) were free from pain and sensory disturbance at the time of follow-up, while the occurrence of numbness or tingling had resolved in 312 (49%) of the 633 hands in which these symptoms had been reported at baseline. Table 3 shows the frequency of symptom resolution in relation to baseline characteristics and treatment received during follow-up. Resolution of symptoms was unrelated to sex or age, but was more frequent in patients treated by surgery (PRR 1.4 for both outcomes). After allowance for these variables, it was less likely in current smokers (PRR 0.7 for both outcomes) and in patients with a strong tendency to somatise (PRR 0.7 for both outcomes). In addition, complete resolution of sensory symptoms, including pain as well as numbness and tingling, was less common when neck pain was reported at baseline (PRR 0.7, 95% CI 0.6-0.9) and more common in patients with diabetes (PRR1.6, 95% CI 1.0-2.5), while numbness and tingling were less likely to resolve when BMI was ≥30 kg/m2 (PRR 0.7, 95% CI 0.5-1.0).

Table 3 Frequency of symptom resolution in hands according to baseline characteristics and treatment during follow-up
Full size table

Table 4 compares associations between symptom resolution and surgical treatment according to clinical features assessed at baseline. Risk estimates are presented with and without adjustment for the five risk factors in Table 3 which showed significant (p < 0.05) associations with resolution of symptoms (smoking habits, somatising tendency, neck pain, diabetes and BMI). The adjustment had only a small impact. Among patients with abnormal median SNC, surgery was associated with resolution both of numbness, tingling and pain (adjusted PRR 1.5, 95% CI 1.0-2.2) and of numbness and tingling specifically (adjusted PRR 1.8, 95% CI 1.3-2.6). In contrast, no association with surgery was apparent for either outcome where median SNC was classed as normal. Similarly, there was evidence of a differential outcome after surgery where numbness and tingling were originally present in regions of the hand falling exclusively within the sensory distribution of the median nerve. However, negative findings from both Tinel’s and Phalen’s tests did not preclude a better outcome in patients treated surgically. The difference in associations between surgery and resolution of numbness and tingling according to median SNC abnormality was of borderline statistical significance (p = 0.05).

Table 4 Frequency of symptom resolution in hands treated or not treated by surgery according to clinical features assessed at baseline
Full size table

Discussion

In this follow-up investigation, the definition of abnormal median SNC that we had previously proposed distinguished a group of patients which appeared to benefit from surgery from another which did not. This further supports the utility of the definition as a diagnostic criterion in epidemiological research.

In longitudinal studies, the response rate most relevant to internal validity is that among subjects who enter follow-up – in our study, patients who consented at baseline to receive a second questionnaire. The response rate that we were able to achieve (56% overall) was less than ideal, and as in many studies, was lower in men and at younger ages. Nevertheless, important bias would have resulted only if responders were atypical as regards the relation of baseline characteristics to outcomes of interest – principally the resolution of numbness, tingling and pain, and its association with surgery. That response rates varied little according to clinical findings at baseline (Table 1) makes it unlikely that associations between symptom resolution and surgery would be biased differentially in relation to these clinical features.

A greater challenge to interpretation is the possibility of confounding by factors influencing selection for surgery. Although randomised controlled trials attest to the benefit of surgical decompression of the carpal tunnel in CTS[7], the associations which we observed between surgical treatment and symptom resolution may not fully reflect the impact of surgery. It could be, for example, that a decision was taken not to operate on some patients because their symptoms were already improving spontaneously, in which case the benefits of surgery would tend to be underestimated in our analysis. To minimise the potential for such confounding, we adjusted for risk factors at baseline which were identified as carrying an adverse prognosis independent of surgical treatment, and reassuringly, this adjustment had only minor impact on the risk estimates of main interest. We cannot rule out a residual effect of other, unrecognised confounders, but to produce important bias in relation to our study question, such confounding would have to be differential according to baseline nerve conduction results, which seems unlikely. In particular, there is no reason to expect that surgeons would be more likely to operate on patients with a poorer chance of responding to surgery if they had normal as compared with abnormal median nerve conduction.

Another limitation was the modest statistical power to demonstrate differences in associations of symptom resolution with surgery according to baseline characteristics. Thus although surgery was clearly associated with the disappearance of symptoms in hands with abnormal median SNC, and there was no indication of any benefit from surgery in those with normal median SNC, this divergence was only of borderline statistical significance. That said, our investigation was larger than most previously reported studies of outcome following surgery for CTS[8], and had the advantage of referent data from patients who were not treated surgically.

As would be expected, abnormality of median SNC was a strong predictor of surgical treatment (PRR 3.6, 95% CI 2.3-5.5). However, surgery was by no means limited to hands with abnormal median SNC according to our definition. This is not surprising since the definition, which was proposed for use in epidemiological studies and not in clinical practice, was formulated only after the collection of all baseline data for the study was complete. Moreover, it did not take account of other neurophysiological findings – for example, on motor nerve conduction. The importance of other clinical features is evidenced in the higher rate of surgery when Phalen’s test was positive, even after account was taken of whether median SNC was abnormal (Table 2).

After allowance for any surgical treatment, symptom resolution was less frequent in current smokers and in patients who tended to somatise. An adverse prognostic influence of smoking has not been reported previously, but somatising tendency has been found to predict persistence of musculoskeletal pain in other circumstances[9, 10], presumably because it increases patients’ awareness of, and causes them to dwell on, symptoms which others would dismiss. Complete resolution of sensory symptoms was also less likely in patients who reported neck pain at baseline, perhaps because in some of these patients, hand symptoms were attributable to pathology in the neck that was less likely to improve over time. In contrast, diabetes was associated with higher resolution of symptoms. In addition, numbness and tingling resolved less frequently in patients with high BMI. Obesity is a known risk factor for CTS[11–16], but has not been clearly associated with worse outcome following surgery[17]. The comparisons that we made of associations between surgery and symptom resolution were all carried out with and without adjustment for these prognostic variables.

Although many studies have examined outcomes after surgery for CTS according to pre-operative neurophysiological findings[18–35], these have not entailed comparisons with patients who had similar neurophysiological results but were not treated surgically. Thus, they do not allow assessment of the extent to which improvements in symptoms could be attributed to surgery, or of whether patients with particular neurophysiological abnormalities benefitted more from surgical treatment. Because of limited statistical power and the possibility of uncontrolled residual confounding, we cannot draw firm conclusions on this second question. However, our results support the hypothesis that abnormal median SNC, as we have defined it, distinguishes hands that are more likely to benefit from surgical decompression of the carpal tunnel. They also suggest that surgery is unlikely to be of value when numbness and tingling do not affect any parts of the hand that lie exclusively within the sensory distribution of the median nerve. In contrast, when both Tinel’s and Phalen’s tests were negative, there was still an apparent benefit from surgical treatment.

Conclusions

We have already demonstrated that our proposed definition of abnormal median SNC distinguished disease with distinctive risk factors[4]. That it appears also to predict benefit from surgical treatment gives further support to its validity as a diagnostic criterion in epidemiological research.