Screening

Screening for potential participants began in April 2013. In total, 2118 patients with a tibial fracture were screened over 36 months. Screening data are presented to help assess the generalisability of the results to the overall population. Screening data are presented for all sites, with the exception of the Edinburgh site, which was closed to recruitment early in the main phase of the study.

There were significant differences in the total number of potential participants who were assessed between recruiting sites because of the varying practice at each site; for example, some sites screened children aged < 16 years but other sites were adult-only trauma centres. The wide variation in totals by centre indicates that some sites screened extensively and recorded any patient with a fracture involving the tibia; however, others appear to have only recorded fractures of the distal tibia on the screening logs.

The reasons why screened patients were not eligible (n = 1581) are detailed in Table 27, Appendix 1.

The most common reasons for participant ineligibility were that the fracture did not extend to within 2 Müller squares of the ankle joint (375/1581; 24%), the fracture was open (369/1581; 23%) and the fracture extended into the ankle joint (329/1581; 21%).

The total number of potentially eligible patients for the study was 537. The reasons why potentially eligible patients were not randomised (n = 216) are given in Table 28, Appendix 1.

Of the 537 eligible potential participants identified on screening logs, 40% (216/537) were not randomised. The conversion rate of eligible potential participants to randomised participants was therefore 60%. The most common reasons for non-participation were that there were no research staff available to consent the patient (n = 53; 25%) or that the surgeon had a preference to use an IM nail (n = 54; 25%). The other common reason was that the patient did not want to be part of a research study (n = 25; 12%). There were 18 potential participants for whom the reason for non-participation was ‘other’; these reasons included skin around ankle precluded plate fixation (n = 4), primary amputation (n = 3) and hind foot nail used (n = 2). The complete table of reasons why eligible potential participants were not randomised, by site, is given in Table 28, Appendix 1.

Data on sex and age were available for almost all of the patients screened (2003/2118; 95%). The largest subgroup of screened patients were men aged < 50 years, who accounted for 41% (819/2003) of the screened population.

Table 3 shows the age and sex of potentially eligible participants by randomised status (randomised or not randomised). The ages of both groups were similar and a t-test comparing the means of both groups showed no evidence of a difference in age (mean difference –0.9 years, 95% CI –4.1 to 2.3 years). Likewise, a two-sample test of proportions showed no evidence of a difference in sex distribution between the two groups (p = 0.818).

TABLE 3

Characteristics of eligible participants who were randomised and not randomised

Recruitment

The trial planned to recruit 320 participants in total. Formal recruitment began in April 2013, with the first participant being randomised on 22 April 2013 at the lead site, University Hospitals Coventry and Warwickshire NHS Trust. Recruitment was undertaken at 28 sites over 36 months, with the final participant being randomised on 3 May 2016. For reporting purposes, the final participant has been included in the April 2016 recruitment month for ease of presentation. Details of site names and their opening dates are listed in Table 4.

TABLE 4

The UK FixDT trial research sites

The pilot phase was completed at the end of November 2013, with six sites open to recruitment. The planned recruitment rate was 0.75 participants pcpm. The recruitment rate averaged 0.6 participants pcpm during this phase, so the TSC recommended that the number of sites was increased from 18 to at least 24 to enable the trial to meet the target recruitment of 320 participants within the 30-month recruitment window. This resulted in a revised target of 0.6 participants pcpm.

Actual recruitment rates, by site, are shown in Table 29, Appendix 1. The overall observed recruitment rate for the main phase of the trial was 0.47 participants pcpm. This was somewhat lower than the revised target and so the number of trial sites was further increased to 28. One site was closed after just one patient was recruited, and screening data are not presented for this site.

Figure 2 illustrates the progression of recruitment over time, towards the required sample size target of 320. Recruitment slowed towards the end of the planned recruitment phase, so the recruitment window was extended, although this was still within the original timeline of the trial. Figure 3 shows the Consolidated Standards of Reporting Trials (CONSORT) flow diagram for the UK FixDT trial.

FIGURE 2

Overall recruitment progression and target recruitment during the main phase (January 2014–April 2016) of the UK FixDT trial.

FIGURE 3

The Consolidated Standards of Reporting Trials (CONSORT) flow diagram.

Table 5 describes the follow-up rates achieved during the trial. The expected follow-up rate used for sample size adjustment at 6 months was 80%. This was surpassed at the 6-month time point, with 88% of DRI assessments completed. Similar completion rates of 86% and 80% were achieved at the 3- and 12-month time points, respectively.

TABLE 5

Follow-up status at time points in the UK FixDT trial

Tables 30 and 31, Appendix 1 demonstrate recruitment by stratification variables, namely age group and centre. There is good balance with respect to each factor, indicating that the minimisation procedure used to allocate treatment was implemented successfully.

The histograms in Figure 4 show the distribution of age in years, separately by sex. The overlaid kernel density estimator (the kdensity option of the histogram function in Stata version 14) shows the smoothed distribution and highlights the differences between sexes, with a noticeable but expected peak in the number of young male patients randomised. The mean difference in age between men and women was 6.8 years, 95% CI 3.2 to 10.4 years.

FIGURE 4

Age distribution of randomised participants by sex. (a) Female; and (b) male.

Primary outcome

Figure 5a shows the trend in group DRI score means over the trial. Figure 5b shows the distribution of the DRI score at each time point, in which the middle bar is the median, the box represents the interquartile range and the whiskers extend to 1.5 times the interquartile range, with observations outside this range presented individually. Figure 5b demonstrates there is a treatment group difference in favour of the IM nail fixation group at 3 months, but this is reduced and not statistically significant at the 6-month time point and decreases further at the 12-month time point. Model assumptions were checked and appeared to be appropriate.

FIGURE 5

The DRI. (a) Box plots of baseline and follow-up scores; and (b) overall trend in means and 95% CIs.

Table 11 shows the treatment estimates modelled using mixed-effects linear regression model, as previously described. The adjusted estimate of the treatment effect for the DRI score, at the 6-month post randomisation time point based on an intention-to-treat analysis, is 4.0 points (95% CI –1.0 to 9.0 points) in favour of the IM nail fixation group compared with the locking plate group. The p-value of 0.114 indicates that there is no evidence for a statistically significant difference in the DRI score between the two treatment groups at the 6-month post-randomisation time point. The prespecified MCID for the DRI is 8 points; therefore, at the 6-month time point we conclude that if there is a difference in the disability outcomes of the two groups, it is likely to be appropriately small as to be clinically unimportant. However, the 95% CI does include the prespecified MCID.

TABLE 11

Disability Rating Index: summary statistics, unadjusted and adjusted treatment effects at all time points

The adjusted estimate of the treatment effect at 3 months is 8.8 (95% CI 4.3 to 13.2) in favour of the IM nail fixation group compared with the locking plate group. The p-value of < 0.001 indicates that there is strong evidence for a statistically significant difference in treatment group means at 3 months. The estimated treatment effect is larger than the prespecified MCID for the DRI of 8 points, so this difference is likely to be clinically important to patients. At the 12-month time point, the p-value of 0.468 indicates that there is no evidence for a statistically significant difference in the DRI score between the two treatment groups at the 12-month post-randomisation time point, and the adjusted treatment difference is 1.9 points (95% CI 3.2 to 6.9 points) in favour of the IM nail fixation group compared with the locking plate group.

Secondary outcomes

Olerud–Molander Ankle Score questionnaire

The adjusted estimate of the treatment effect for the OMAS, at the 6-month post-randomisation time point based on an intention-to-treat analysis, is –6.0 points (95% CI –11.2 to –0.7 points) in favour of the IM nail fixation group compared with the locking plate group (Figure 6 and Table 12). The p-value of 0.026 indicates that there is evidence for a statistically significant difference in the OMAS between the two treatment groups at the 6-month time point.

FIGURE 6

The OMAS questionnaire. (a) Box plots of baseline and follow-up scores; and (b) overall trend in means and 95% CIs.

TABLE 12

Secondary patient-reported outcome measures: summary statistics, unadjusted and adjusted treatment effects at all time points

EuroQol-5 Dimensions index score

The adjusted estimate of the treatment effect for the EQ-5D index score, at the 6-month post-randomisation time point based on an intention-to-treat analysis, is –0.063 points (95% CI –0.12 to –0.01 points) in favour of the IM nail fixation group compared with the locking plate group (Figure 7 and see Table 12). The p-value of 0.033 indicates that there is evidence for a statistically significant difference in the EQ-5D score between the two treatment groups at the 6-month time point.

FIGURE 7

The EQ-5D index. (a) Box plots of baseline and follow-up scores; and (b) overall trend in means and 95% CIs.

EuroQol-5 Dimensions visual analogue scale score

The adjusted estimate of the treatment effect for the EQ-5D VAS score, at the 6-month post-randomisation time point based on an intention-to-treat analysis, is –2.5 (95% CI –6.8 to 1.8) in favour of the IM nail fixation group compared with the locking plate group (Figure 8 and see Table 12). The p-value of 0.247 indicates that there is no evidence of a statistically significant difference in the EQ-5D VAS score between the two treatment groups at the 6-month time point.

FIGURE 8

The EQ-5D VAS. (a) Box plots of EQ-5D VAS baseline and follow-up scores; and (b) overall trend in EQ-5D VAS means and 95% CIs.

Per-treatment analysis

Table 13 shows the results of per-treatment analyses, in which participants were analysed in per-treatment groups, that is, those who received IM nails compared with those who received locking plates. The five participants who did not receive either treatment were excluded from these analyses. An adjusted per-treatment analysis of the DRI scores at 6 months gave an adjusted treatment effect of 4.2 (95% CI –0.9 to 9.2) and a p-value equal to 0.103. Adjusted analysis at 3 and 12 months also gave similar results to the intention-to-treat analysis in Primary outcome.

TABLE 13

Means and SDs of outcomes at all time points and estimated treatment effects after adjustment, using a per-treatment analysis approach

Adjusted per-treatment analysis of the secondary outcome measures are also given in Table 13. All analyses yielded similar results to the equivalent intention-to-treat analysis.

Missing outcome data

The follow-up questionnaire completion rate was very good at all time points of the UK FixDT study. The expected loss to follow-up rate at the 6-month primary outcome time point was 20%; the actual rate was 12%. A summary of status at each follow-up time point is given in Table 5. In total, 284 out of 321 participants (88.4%) completed a DRI at the 6-month time point. The item-level missingness for the DRI score, and the secondary outcome measures, is given in Tables 32–34, Appendix 2. There were very few missing items in the DRI questionnaire at baseline (one item from 3816 items; < 0.1%), 3 months (1/3312; < 0.1%), 6 months (2/3408; < 0.1%) and 12 months (5/2640; 0.2%). Similar levels of missing data were seen across the OMAS questionnaire and the EQ-5D, with item-level missingness below 0.2% for all scales.

Table 14 shows the impact that including partial completion of the DRI using 11 or fewer items had on calculating the overall DRI score. There is very little difference in the adjusted treatment differences compared with the results given in Primary outcome based on fully completed questionnaires only. The largest differences are seen in the 12-month questionnaire data but this is because there are fewer observations and marginally more missing items than at 3- and 6-month data collection, resulting in a slightly larger treatment difference but being still in favour of IM nail fixation, as per the results in Primary outcome.

TABLE 14

Means and SDs of outcome at all time points and estimated treatment effects after adjustment per treatment, using all values

Subgroup analysis

At 6 months, the p-value of 0.516 indicates that there is no evidence for a significant interaction effect between age group and treatment group. The results of this analysis, and the subgroup analyses performed at 3 and 12 months, are given in Table 15.

TABLE 15

Results of subgroup analysis of the DRI score showing means and SDs of outcome at all time points and estimated treatment effects after adjustment

Neither was there any evidence for a significant interaction effect between sex and treatment group (p-value of the interaction term = 0.384).

Exploratory analyses

Local complications related to the fracture or its treatment

Complications were assessed by the research team at the 6-week time point for almost all randomised participants (n = 314; 99% of participants who had a surgical procedure) and, again, in the patient reported questionnaires at 3, 6 and 12 months. Table 16 shows these complications by allocated treatment group.

TABLE 16

Local complications

The number of patients with symptoms related to infection was higher in the locking plate group (n = 32; 20%) than in the nail fixation group (n = 20; 13%), although this was not statistically significant (Fisher’s exact test p = 0.094). Twenty-one (13%) patients in the locking plate group and 14 (9%) in the IM nail fixation group were treated with antibiotics.

Other local complications were rare in both groups.

Data on ability/inability to bear weight on the injured leg were also collected at the 6-week visit in Table 16. In total, 33% of those allocated to nail fixation reported being fully weight-bearing at the 6-week visit, compared with only 14% of those allocated to locking plates. The difference in proportions is 0.19, 95% CI 0.10 to 0.41. The p-value for the difference is these proportions is p < 0.001, indicating strong evidence of a difference in the proportion of participants weight-bearing at 6 weeks.

Further surgery related to the fracture or its treatment is shown in Table 17. Where patients reported any further surgery, full details of the operation were requested and provided from the treating centre. There was more further surgery in the locking plate group (n = 19; 12%) than in the nail fixation group (n = 14; 9%), although this was not statistically significant (Fisher’s exact test p-value = 0.363).

TABLE 17

Further related surgery within 12 months of injury

Radiographic outcomes

An independent orthopaedic surgeon assessed the radiographic images, when available. The results of these independent assessments are given in Tables 18 and 19 for 6-week and 12-month images, respectively.

TABLE 18

Imaging outcomes at the 6-week post-surgery radiograph

TABLE 19

Imaging outcomes at the 12-month post-surgery radiograph

In addition to the independently assessed radiographs, the principal investigator at each recruiting site assessed key radiographic outcomes and reported these on the 6-week postoperative form. Comparisons between these site assessments and independent assessments were investigated, and there were good levels of agreement between assessors.

Systemic complications potentially related to the fracture or its treatment

Information regarding complications was also given through the SAE reporting pathway over the full 12 months of the trial. Data from related SAEs have been included in these complication profiles, with cross-referencing to ensure that appropriate counts are taken using multiple data sources. Not all of the venous thromboses or infections of the chest, urinary system, etc., will be related to the treatment but we have included all ‘potentially’ related events here for completeness. These data are summarised by treatment group in Tables 17 and 20.

TABLE 20

Systemic complications