Recruitment
A total of 375 patients screened were found to be ineligible for the trial. The most common reason for being ineligible (n = 155) was presenting > 14 days after the Achilles tendon injury. Late presentation after an Achilles tendon rupture is not uncommon. Although a patient may have severe pain when the tendon ruptures, this acute pain usually settles quickly and the patient can often put weight through their leg, albeit without being able to walk normally. Some patients feel that they would not be able to walk at all if they had suffered a ‘serious’ injury and therefore continue to try to mobilise on their leg before eventually seeking treatment some time later, when their ability has not improved. The cut-off point of 14 days used to define an ‘acute’ rupture is somewhat arbitrary but in keeping with definitions used in previous research into this injury. Patients presenting at later times may have partial or complete healing of the tendon but often with tendon lengthening, which restricts their function. As the treatment of late presentation injuries is not straightforward and often requires surgical intervention, these patients were excluded from this trial of acute non-operative management.
The second common reason why patients were ineligible (n = 120) was that they chose to have surgery. This number is perhaps smaller than had been anticipated when UKSTAR was designed and, to some degree, accounts for the faster than expected rate of recruitment. However, it is in keeping with the worldwide trend towards non-operative treatment of acute rupture of the Achilles tendon, as per the recent evidence base, which suggests little functional advantage of surgery.8,9
Only 46 patients were excluded because they were unable to adhere to trial procedures or complete questionnaires, most commonly because they could not read written English, which was used in the follow-up questionnaires. Thirty-seven patients were excluded because they had suffered a previous Achilles tendon injury, which was likely to have affected their baseline, pre-injury function. Achilles tendon rupture is very rare in children and hence it is not surprising that only three patients were excluded because they were aged < 16 years. The remaining 14 patients were excluded by recruitment centres under the heading ‘other’; we did not record details of these individual cases.
Of the 1076 potentially eligible patients screened across the 39 recruitment centres, 540 consented to enter the trial. Ninety-nine patients were not approached about the trial because no research associate was available to discuss the trial (n = 97), usually because the patient had presented at the weekend or because no functional brace was available at the time of presentation (n = 2). A further 50 patients were not offered the opportunity to take part in the trial because of a clinician decision. In some cases, specific reasons were given for this decision, for example ‘active treatment of a local skin lesions precluding the use of a cast’, but in other cases the clinician did not provide a reason. Therefore, a total of 149 potentially eligible patients were never offered the opportunity to take part in the trial. This reduced the number of participants but is unlikely to have caused selection bias.
Of the 927 patients who had the opportunity to take part, 385 declined. Patients may decline to take part in a trial for various reasons. Those who do not want to be part of any research – often because of the perception, or indeed the reality, of filling out extra, onerous questionnaires – are unlikely to adversely affect the trial in terms of the difference between treatment groups. However, those who decline because they have a preference for one treatment over another do create selection bias. In this trial, it is reassuring that 540 out of the 927 patients who were offered the opportunity to take part in the trial (58%) agreed to participate.
Participants and interventions
The two groups of participants were well balanced in terms of baseline pre-injury characteristics.
In keeping with the epidemiology of Achilles tendon rupture, more men were aged 30–40 years, whereas women were a little older at the time of their injury, with the age range of 40–60 years being most common. Sports accounted for the great majority of ruptures (70.4%) and relatively few patients declared predisposing risk factors; only 4.3% of participants had diabetes mellitus and 3.9% were taking steroids, these being associated with an increased risk of tendon rupture. Only 3.7% declared a pre-existing Achilles tendinopathy, despite the fact that histological studies indicate that degenerative changes are almost always present on biopsies taken from tendons immediately after an injury.4,5
The median pre-injury ATRS score was 100 in both groups, indicating normal Achilles function. Similarly, the median EQ-5D utility score was 1, indicting perfect health. This, and the fact that the large majority of participants were employed or self-employed, suggests that most Achilles ruptures affect working-age people with good pre-injury health. In studies of acute injury, it is necessary to collect functional and QoL data by recall, in this case as part of the baseline post-injury clinical reporting form. Although some recall bias/response shift is inevitable, there is no alternative in this sort of trial. In clinical areas in which pre-injury QoL is much more variable than for patients with Achilles tendon rupture, we have demonstrated that recall estimates of QoL still agree with age- and sex-matched controls.50 However, the characteristics of the participants reflect that Achilles rupture affects all age groups, with both men and women in their 80s represented in the trial. Overall, the participants in the trial are representative in terms of demographics of previously reported patients with this injury.
We anticipated some crossover between the treatment groups following the random allocation, but in fact this was relatively uncommon. Only one participant decided to have a cast after having been allocated a functional brace. Thirteen participants decided to change to a functional brace after having been allocated a cast, which may reflect the perception that the brace made mobilisation easier. However, the numbers were small and we did not formally investigate the qualitative aspects of the decision to change treatment. Two further participants withdrew from the trial immediately after randomisation and seven others crossed over treatment groups for what were described as clinical or unknown reasons. Given the small number of crossovers at baseline, these are very unlikely to have influenced the results of the trial.
In terms of compliance with treatment once implemented, the trial protocol stipulated that patients would be deemed compliant if they maintained their allocated treatment for a minimum of 6 weeks. The choice of 6 weeks reflects the fact that this was the time at which weight-bearing would usually be permitted for those patients in a cast, those in a functional brace generally being fully weight-bearing from the outset. Overall, 88% of participants were fully compliant. However, compliance was higher in the functional brace group (97%) than in the plaster cast group (80%). This may reflect participants’ desire to have the cast removed as soon as possible (most of these participants used a functional brace for a further 2 weeks or longer), but we did not interview participants about the reasons why they changed treatment after 6 weeks.
Some patients did, of course, change treatment before 6 weeks, having initially accepted their allocated intervention. This was more common in the plaster cast group; 11.3% changed to a functional brace, compared with the 0.4% allocated a functional brace who changed to a plaster cast. This may also suggest that functional brace was preferred but, although we asked these participants if they or the clinician treating them chose or recommended changing treatment, we were not able to formally explore the reasons behind decisions to change treatment. An additional 3% of participants chose to have surgery before 6 weeks (1.9% in the plaster cast group and 1.1% in the functional brace group). In some cases, these participants described another fall or injury to their tendon. However, we have not reported these as ‘re-ruptures’ of the tendon on the basis that the tendon was unlikely to have healed before 6 weeks.
One other notable element of participants’ treatment beyond the treatment allocation was the use of VTE prophylaxis. Patients with Achilles tendon rupture are at increased risk of VTE, as the injury defunctions the triceps surae muscles, which are an important part of the calf muscle pump that helps to return venous blood to the heart. In the plaster cast group, 70% of patients had VTE prophylaxis, most commonly self-administered low-molecular-weight heparin injections. Fewer patients (58%) had VTE prophylaxis in the functional brace group. This difference may reflect the belief that patients who are able to fully weight bear in a functional brace are at lower risk of VTE than those with restricted weight-bearing in a cast, but this trial was not designed to address questions related to the management of VTE.
Results
In total, 93.3% of participants completed the primary outcome measure 9 months after their Achilles tendon rupture: 91.7% in the plaster cast group and 94.9% in the functional brace group. Therefore, loss to follow-up was considerably lower than the 20% accounted for in the trial design, which, alongside the fact that the trial was able to recruit more patients than the minimum of 330 required by the sample size calculation, ensures that the trial had considerably more than 90% power.
Follow-up was also good at other time points, with 88%, 88% and 86% of participants completing questionnaires at 8 weeks, 3 months and 6 months, respectively.
Primary outcome
The adjusted ITT analysis showed no statistically significant difference in ATRS between the two treatment groups at the primary end point of 9 months post injury (–1.38, 95% CI –4.9 to 2.1).
There was a statistically significant difference in ATRS at 8 weeks in favour of the functional brace group (5.53, 95% CI 2.0 to 9.1), although this is of borderline clinical importance. However, any benefit of functional brace was not evident later in the participants’ recovery, with very similar ATRS scores at 3 and 6 months, as well as at 9 months.
As expected, given the relatively small number of patients who were non-compliant with treatment, the secondary sensitivity analysis, using adjusted CACE, showed the same pattern. There was no evidence of a difference at 9 months post injury (–1.17, 95% CI –4.5 to 2.1), nor was there any evidence of a difference on the other prespecified analysis of overall ATRS scores (AUC) over the full period of follow-up.
Mean ATRS was imputed pro rata when no more than 50% of the items were missing (at least five questions answered). Although this method may underestimate the variance, the number of ATRS items imputed was small (see Chapter 2, Descriptive analysis).
Secondary outcomes
The analysis of patient-reported HRQoL (EQ-5D utility score) provides powerful corroborating evidence in support of the findings using ATRS. There was a statistically significant and clinically relevant difference in favour of functional brace at 8 weeks (0.069, 95% CI 0.03 to 0.1). A breakdown of the EQ-5D by domain of health showed that this difference at 8 weeks lies in ‘self-care’ ability and ‘usual activities’. This difference in EQ-5D utility scores was of borderline statistical significance (0.035, 95% CI 0 to 0.07) at 3 months, but there was no evidence of a difference at any subsequent time point. There was no evidence of a difference in EQ-5D VAS scores.
The trial was designed to compare patient-centred outcomes between participants randomly allocated to a plaster cast and those allocated to a functional brace. However, the safety profile of the functional brace was another important consideration. Specifically, if the risk of re-rupture of the tendon was higher in those patients allowed to fully weight-bear in a functional brace, this would influence decision-making in this area, even when PROs were similar. Interestingly, the risk of re-rupture was generally lower than that reported in the literature, with a total of 17 (6.4%) cases in the plaster cast group and 13 (4.7%) cases in the functional brace group. None of the re-ruptures occurred > 6 months after the injury.
There was no evidence of an association between the treatment group and any other type of complication, with the exception of non-injurious falls, which were more common in the plaster cast group (p = 0.015).
Health economics evaluation
The mean direct intervention costs were £36 for the plaster cast group and £109 for the functional brace group. The higher upfront cost of the functional brace (mean difference £73) was statistically significant. However, by 8 weeks this difference had reversed, such that the mean total NHS and PSS costs were significantly lower in the functional brace group. The difference was driven mostly by the greater number of outpatient appointments required in the plaster cast group.
This is an important finding, as it will reassure the finance teams in trauma and orthopaedic departments that, despite the extra initial cost of a functional brace, they will reduce their overall costs when treating patients with an Achilles tendon rupture.
The mean total NHS and PSS costs during the entire follow-up period were £1183 for the plaster cast group and £1018 for the functional brace group. Although functional brace was marginally cheaper, the mean between-group cost difference of £164 was not statistically significant.
In terms of HRQoL, the mean QALY value was, on average, marginally higher for the functional brace group among complete cases and in the sensitivity analyses, although this mean QALY difference was not statistically significant.
Therefore, as the functional brace group incurred slightly lower costs and achieved slightly better QoL over the course of the study, in health economic terms, functional brace is the dominant intervention.
In summary, the health economic evaluation indicates that functional brace is very likely to be cost-effective.
Limitations
A concern at the start of the study was that patients would not be willing to take part in a trial comparing two interventions that needed to be worn for a prolonged period of time. Some of the 385 patients who declined did so because they did not want to be part of a research project. These patients, although undoubtedly affecting the external validity of the trial, are unlikely to create selection bias when comparing the two interventions. By contrast, those who declined because they had a preference for one treatment over another do create selection bias. However, in total, 540 of the 927 patients who were offered the opportunity to take part in the trial (58%) agreed to participate, so we can be confident that the participants in the trial are broadly representative of the population of patients having non-operative treatment for acute rupture of the Achilles tendon.
A further anticipated limitation was crossover from the allocated treatment and, indeed, 14 patients did not receive their allocated intervention after being randomised. There were also some cases of incomplete compliance with treatment. The ability to bear weight immediately using a functional brace may have triggered a desire to change treatment, given that the majority changed from the plaster cast group to the functional brace group. However, the overall number is small for a trial of this size and the CACE analysis, that is the analysis adjusted for incomplete compliance, confirmed the result of the primary analysis (i.e. there was no evidence of a difference between the two groups of participants at 9 months post injury).
Loss to follow-up is another potential limitation. However, > 93% of participants provided primary outcome data at 9 months, which is considerably higher than the 80% assumed in the trial design. Therefore, given that the trial also exceeded the minimum sample size by some margin, we can be confident that the conclusions are robust and the risk of type II error is very low.