Important treatment trials
Modern evidence based medicine is heavily dependent on formal trials of treatment used to determine which treatments are most effective for any given condition, something which can be surprisingly difficult in a condition like CTS which can improve even without any treatment. Some trial methodologies are considered to produce more reliable results than others, ie they employ means of reducing bias. The available trials of treatment in CTS do not answer all the questions one might ask about treatment but some important high quality ones are summarised here:
Treated 30 patients with a steroid injection (methylprednisolone 40mg + lignocaine 10mg - a local anaesthetic) and 30 patients with an injection of lignocaine only. Neither patients nor those assessing the outcomes knew which type of injection was used (double blinding) and outcomes were studied at 1 month after injection at which time 23 of the patients given steroids had improved compared to 6 of the patients who did not receive steroids. Up to this point this trial was of very high methodological quality and it has been extensively quoted but at one month the investigators broke the blinding and offered the 24 patients who had received the lignocaine only injection, and whose symptoms had not resolved, a further injection with steroids. They then went on to follow all 60 patients up for one year but the fact that the patient assessments beyond 1 month were not blind to treatment allocation has led most reviewers to disregard the longer term outcome results in this study. For what it is worth, the 1 year follow-up results were that 2 patients of the 30 in the placebo (lignocaine only) group had not required further treatment for their CTS while 15 of the 30 patients who originally received steroids had not needed any further intervention.
One of the more useful studies of different doses of injected steroid. 132 patients with mild to moderate CTS were randomised to receive 20mg, 40mg or 60 mg of methyprednisolone and were followed up for either a year, or until relapse of symptoms. A significantly higher proportion of patients given 60mg were still in remission after 6 months (32/44 = 73%) compared to those given 40mg or 20mg (48/88 = 54%). However, by the one year follow-up assessment there was no longer a significant difference between the groups. This paper is also one of the only studies to tell us anything at all about second injections in that, on relapse, the patients were offered the option of a second injection or surgery. 54/68 (79%) chose to have a second injection and 39 of these (72%) had not required further treatment by the time the paper was written, though it is not reported what the mean length of follow-up for these patients was.
Randomised 80 hands to surgery and 83 hands to a somewhat unusual steroid injection regime in which one injection of 20mg paramethasone acetonide was given, followed by a second injection two weeks later if night-time tingling had not completely resolved (69 of the hands in the injection group received this second injection). The patients, obviously, were not blind to which treatment they received and the outcome measurement used - improvement in a visual analogue scale for severity of symptoms, was an unusual one. At 12 months 70% of the injected patients still had a satisfactory result compared to 75% of the surgical cases - a difference which was not statistically significant.
Randomised 25 patients (only one hand from each patient was studied) to a single injection with 20mg methylprednisolone and 25 patients to surgery. Outcomes were evaluated at 20 weeks after treatment and measured using a ‘global severity scale’ on which patients were asked to score the overall severity of their symptoms from 0 (no symptoms) to 50. The mean change in this global severity scale between the pre-treatment assessment and the 20 week assessment was reported and it was clear that the surgical group showed a greater change in overall symptoms, 24.2 scale units compared to 8.7 scale units in the injection group. This however does not tell us how individual patients fared. Other data in the paper suggests that 6 of the surgically treated patients and 21 of the injected patients still had significant symptoms at 20 weeks. The trial is described as ‘single blind’ but as the primary outcome measure was a subjective evaluation of symptom severity by the patient and the patients were clearly not blind to treatment allocation this seems rather a dubious claim.
Randomly allocated 20 patients with clinically diagnosed CTS to surgery and 20 to injection with 40mg methylprednisolone evaluating the results with the same global symptom scale (GSS) used in the study by Hui et al at 2, 4 and 12 weeks after treatment. Improvements in mean GSS were identical at 2 and 4 weeks but by 12 weeks symptoms had started to recur in the injection group and the average scores were significantly better in the surgical group. This is therefore a very similar study to Hui et al with a similar conclusion. This study tells us even less about what actually happened to the injected patients than Hui et al and we cannot tell whether all of the injected patients had relapsed by 12 weeks or only some of them - thus reducing the mean improvement in GSS. It is also perhaps notable that in this study one surgical patient suffered from 'reflex sympathetic dystrophy' (complex regional pain syndrome) after surgery - about the average complication rate for surgery of 5%. One injected patient had an episode of cellulitis but this settled with treatment.
Randomised 57 patients (one hand studied) to surgery and 59 patients to a complex combination of non-surgical measures including splinting, non-steroidal anti-inflammatory drugs, hand therapy (including ligament stretching and nerve and tendon gliding exercises), ergonomic interventions in the workplace and therapeutic ultrasound. All of the patients had already failed a trial of splinting for their symptoms before entry. Surgery in this trial was carried out by a variety of surgeons with different techniques including both traditional open and endoscopic (keyhole) operations. The outcomes were assessed after 12 months, primarily using an ‘intention to treat’ method - ie the results were measured for the two groups of patients at 12 months whether or not they actually received the randomly allocated treatment. Understanding this is vital to interpreting the results of this study because only 42 of the 57 patients randomised to surgery actually had the operation in the first three months of the trial, 6 of the patients allocated to conservative treatment never received any, and 19 of the 59 supposedly conservatively treated hands actually had surgery before the 12 month follow-up point. Outcomes were measured by looking for changes in the Boston/Levine questionnaire scores and showed a statistically significantly greater improvement in the ‘surgical’ group. The function score improved from a mean of 2.47 to 1.74 ( better by 0.73) in the surgical group and from 2.53 to 2.17 in the non-surgical group (better by 0.36). For the symptom severity score the change was from 2.95 to 1.74 (better by 1.21) for the surgical group and 3.01 to 2.07 (better by 0.94) for the non-surgical group. The difference was felt by the authors to be ‘modest’ but remember that a fair proportion of the non-surgical group had had surgery and a significant number of the ‘surgical’ group had not! A retrospective analysis was also carried out dividing the patients by the treatment actually received however and confirmed the difference between surgery and the conservative regime used here. Note that local steroid injection was not a part of the non-surgical protocol in this trial and only one patient received a steroid injection during the study. Overall the results of this trial are not unexpected - the interventions used in the non-surgical arm include many treatments for which there was little prior evidence of efficacy and the patients were preselected for unresponsiveness to splinting - the only intervention used here which is supported by previous trial evidence - so the better results from surgery are not surprising. What is perhaps surprising is that the results from surgery are not more different than they turned out to be compared to not operating.
Randomly allocated 73 patients to 4 different oral medications - 16 to placebo, 16 to a diuretic (trichlormethiazide 2mg daily), 18 to a non-steroidal anti-inflammatory drug (tenoxicam 20 mg daily) and 23 to a steroid (prednisolone, 20 mg daily for first 2 weeks 10mg daily for second 2 weeks), each taken for 4 weeks. Symptoms were assessed using the 0-50 point global symptom scale at 2 and 4 weeks. The mean GSS score improved from 27.9 to 10 in the steroid group but did not change in any of the other three groups. This remains the only good quality trial comparing oral drug treatments.
One of the best trials of traditional (open) vs endoscopic (keyhole) carpal tunnel surgery, notable for the use of very good outcome measures and very long (5 years) and complete (all but 2 patients who had died) follow-up. 65 patients/hands (one hand per patient) were randomised to open surgery and 63 to endoscopic. One patient allocated to endoscopic surgery ended up having an open operation for technical reasons but essentially all patients received the planned treatment (cf Jarvik 2009 above). The short term results (3 month) showed less pain in the endoscopic group but by 5 years there were no measurable differences between the two groups in any respect. The results of the study are equally interesting in what they tell us about the long term outcomes of carpal tunnel surgery in general, regardless of the technique used. Only 47/126 patients reported a symptom severity score of 1 (no symptoms at all) at 5 years. It is thus only a minority of patients who are completely ‘cured’ by carpal tunnel surgery in a strict sense. Most of the residual symptoms however were fairly trivial with a further 62 patients having symptom scores <2.0. Five patients in total had to have their carpal tunnel surgery repeated because of recurrent symptoms. The authors seem to feel that all of these were probably true recurrences rather than primary failures. Eight patients (6.3%) expressed themselves dissatisfied with the results of surgery at 5 years.
Dr Atroshi has also carried out a placebo controlled trial of local steroid injection with a 1 year blind follow-up. (Flondell et al 2010). The first report of the outcome of this study appeared as an abstract to be presented at a meeting of the American College of Rheumatology in November 2012. It is however quite difficult to find that abstract on the ACR website (it is number 2648!) so I have copied the text here:
Patients with idiopathic carpal tunnel syndrome (CTS) are commonly treated with local steroid injection but there is currently no evidence from placebo-controlled trials supporting efficacy beyond 1 month.
We conducted a randomized triple-blind placebo-controlled trial of first-time steroid injection into the carpal tunnel in patients with moderately severe CTS. Patients aged 18 to 70 years with primary idiopathic CTS, no severe sensory loss or muscle atrophy, and no previous steroid injection for CTS were randomized to 1 of 3 groups (37 patients in each): 80 mg Methylprednisolone, 40 mg Methylprednisolone, or saline (each also containing 10 mg Lidocaine). Patient-reported outcomes (CTS symptom severity scale, QuickDASH, SF-36 bodily pain, and SF-6D) were obtained and physical examination by a blinded assessor was performed at baseline and at 10, 24 and 52 weeks after injection. The primary end points were change in CTS symptom severity score at 10 weeks and rate of surgery at 52 weeks. Data from all patients were analyzed.
During 1 year after injection surgery was carried out on 27 patients (73%) in the 80 mg Methylprednisolone group, on 30 patients (81%) in the 40 mg Methylprednisoline group, and on 34 patients (92%) in the placebo group. Patients who received placebo were significantly more likely to have surgery during 1 year after injection than patients who received Methylprednisolone; adjusted hazard ratio 2.0 (95% confidence interval 1.3-3.2, P<0.01). The change in the CTS symptom severity score from baseline to 10 weeks was significantly larger in both Methylprednisolone groups than in the placebo group.
In patients with moderately severe carpal tunnel syndrome first-time steroid injection into the carpal tunnel has a significant benefit in relieving symptoms up to 10 weeks and in reducing the need for surgery up to 1 year after treatment.
The full version of this trial has now been published in Annals of Internal Medicine so we can now consider what the results mean
This trial is not only the first to compare different initial doses of steroid but also the first to maintain blinded follow-up of the patient cohort for one year. 111 patients were randomly assigned to receive a placebo injection, 40 mg or 80 mg of methyprednisolone (37 patients in each group). All three injections contained added lidocaine. Patients were followed up for one year and allowed to opt for surgery when they wished. By one year 92% of the placebo group, 81% of the 40 mg methylprednisolone group and 73% of the 80mg methyprednisolone group had opted for surgery. Both of the steroid groups were significantly less likely to have had surgery during the one year follow-up than the placebo group - but the difference is fairly modest and for the 40mg group it was only significant when analysed using a particular statistical approach taking time to surgery into account. The study was very well conducted - only one hand per patient was analysed, randomisation and blinding were very good and follow-up was almost complete (only one or two patients missed some of the intermediate follow-up assessments - all were seen at 1 year)
This trial has immediately been seized upon for comment by surgeons, who argue that it means injection is always temporary and that everyone will require surgery in the end, and by exponents of injection who point out that some of the patients did not require surgery. My own view is that it is a little more complex than it at first might appear.
The first thing one always has to consider when looking at any treatment study is - how were the patients recruited? In this study the subjects had been referred from primary care to an orthopaedic clinic. I do not think one should under-estimate the ability of primary care physicians to selectively refer patients who are likely to require surgery to surgical clinics. Secondly these patients had all already failed an 8 week trial of treatment with splinting again selecting out more treatment resistant cases for the trial. This trial is therefore likely to have started out with a cohort of patients who were intrinsically more likely to require surgery than the average CTS patient. They were also attending the clinic of a surgeon with a longstanding interest in CTS and documented excellent surgical results. It is perhaps not surprising that patients in such a clinic are more likely to opt for surgery.
Secondly one has to look at exclusions. The mean age of the patients treated was about 48 years - somewhat younger than average for CTS. I think this reflects two of the inclusion/exclusion criteria for the study. Patients over the age of 70 were automatically excluded from consideration anyway. In common with most studies of either diagnosis or treatment for CTS, patients with any other coincident problems - diabetes, neuropathy, thyroid disease, inflammatory disease, severe illness, drug or alcohol abuse - were also excluded. Such coincident morbidities are commoner in older patients and this too will have tended to reduce the average age of eligible patients. More than half of the patients originally screened for eligibility (181/323) were excluded by the above criteria (a further 31 declined to participate).
One other exclusion criterion is important - patients with obviously very severe CTS - wasting of the thenar muscles or marked objective sensory loss - were also excluded. This will have, to some extent, mitigated the effect of recruiting in a surgical clinic by taking out of the study the small group of advanced CTS cases who are unlikely to respond as well to any kind of treatment.
The results of this study should therefore not be casually generalised to all CTS patients. It tells us about the chance that a patient <70 years old with mild to moderate CTS resistant to splinting will need surgery during the first year after an injection. It should be noted that patients were not allowed a repeat injection - once their symptoms relapsed after the first one their only option was surgery.
Side effects are well documented in this study for the injections, less so for surgery. There were no serious ones but the patients receiving methyprednisolone reported more pain after injection than the placebo group - "mild or moderate in most patients" and lasting about 2 days. 10 patients reported swelling which resolved within 2 weeks (2 of them in the placebo group). I would have liked to see a more detailed presentation of the surgical results in the 91 operated patients. In my experience any group of 91 patients operated on for CTS is likely to include at least 1 who is unimpressed with the result of surgery but the Swedish experience may be better.
How then should this be interpreted overall. Dr Atroshi and collegues themselves make two comments in the paper which I will quote:
1) "Patients who received 80mg methyprednisolone had a modest but statistically significant lower likelihood of requiring surgery within 1 year than those who received placebo. It is uncertain whether this relatively small difference could justify steroid injection as routine treatment of idiopathic CTS. However, considering the large number of surgical procedures, even a small reduction may be valuable"
2) "Methyprednisolone reduces symptoms and rate of surgery.... Future research should explore how to obtain a consistent durable effect"
My own view has not been greatly altered by this result but I am glad to see this addition to the evidence base telling us what we can expect from steroid injection. It essentially confirms that a single injection is effective at treating CTS but very likely to be followed by relapse within a year. The relapse rate in this study is somewhat higher than we see in East Kent but that is likely to be due to differences in patient selection. It remains the case that injection is markedly safer, cheaper and more convenient than surgery and one has to balance those factors against the greater probability of lasting benefit, but also greater probability of lasting harm, resulting from surgery.
It does provide one very useful result. The higher steroid dose did not produce a dramatically superior effect and it is probably not worth changing our routine injection policy at present. A more fruitful approach may be to explore the avenue of repeat steroid injection depending on response to treatment.
Randomised 68 patients to surgery and 79 to neutral angle splinting for at least 6 weeks. The results need to be read with a little caution. As with Jarvik 2009 the main results are presented as an intention to treat analysis. At one month following randomisation outcomes were actually better in the splinting group but most of the surgical patients had not actually been operated by that stage, surgery being delayed by a mean of 35 days after randomisation. At 18 months treatment success rates were 95% for the surgical group and 75% for the splinting group - but 41% of the splinting group had in fact had surgery by this time. The intention to treat analysis also underlies the curious finding of reported ‘complications’ in the splinting group, which turn out to be exclusively the complications of surgery - Total surgical complications in the 100 or so cases eventually operated were - 53 painful or hypertrophic scars, 24 patients with wrist stiffness, 10 wound haematomas, 5 wound infections, 2 patients with persistent severe wrist pain and one case of reflex sympathetic dystrophy (complex regional pain syndrome)
These authors have since published a sub-group analysis of the patients who were treated with splinting, looking for patient characteristics which predicted success with splinting only (Gerritsen 2003). The overall success rate of splinting was 31%. Patients with a short duration of symptoms before treatment and those with less severe night time paraesthesiae were more likely to respond to splinting alone with a 62% success rate if both factors were present.
Revision date - 7th September 2014