Offer of a bandage versus rigid immobilisation in 4- to 15-year-olds with distal radius torus fractures: the FORCE equivalence RCT

Inclusion criteria

Patients were eligible for inclusion in the FORCE (FOrearm fracture Recovery in Children Evaluation) trial if:

• There was radiographical evidence of a torus fracture of the distal radius whereby there was a cortical deformation within the distal third of the radius but no break in the cortex. These could be associated with an ipsilateral fracture to the ulna (the ulna fracture could be buckle, greenstick or otherwise).

• They were aged 4–15 years.

• Randomisation could occur at a recruitment centre that was able to definitively treat the injury (e.g. an ED).

Exclusion criteria

Patients were excluded from this trial if:

• The injury had occurred > 36 hours previously.

• The treating clinician judged that there was a cortical disruption of the radius on radiographs (i.e. a greenstick fracture).

• The patient had sustained an additional fracture at the time of the index fracture (with the exception of ipsilateral ulna fractures). Any child with bilateral torus fractures was therefore excluded.

• There was evidence that the patient and/or parent would be unable to adhere to trial procedures or complete follow-up, such as insufficient English-language comprehension, developmental delay or a developmental abnormality, or no parental access to a mobile phone with internet access.

Consent

Recruitment took place in 23 recruitment centres in England from 21 NHS trusts that treated children with torus fractures of the distal radius. Eligible patients were identified by the clinical team. After introduction of the study concept by the clinical team, a member of the local research team presented the patient and parents with age-appropriate participant information sheets or online study information and verbal explanation of the trial procedures. The patient/parent were then given the opportunity to discuss any issues related to the trial with the local research team and their family and friends. The parent was then asked to sign an electronic informed consent form, and children from the age of 8 years were invited to sign an electronic assent form. Assent was taken where appropriate; however, the absence of assent did not exclude the patient from the study if consent had been obtained from the parent. If a child indicated that they did not want to take part, the child was not included in the study.

The FORCE trial was part of an ongoing NIHR-funded Study Within A Trial (SWAT) [TRials Engagement in Children and Adolescents (TRECA) NIHR Health Services and Delivery Research 14/21/2123] investigating the effects of the mode of information delivery to children and parents. The FORCE trial was one of a number of host trials embedding the TRECA intervention. Recruitment centres were randomised as clusters. All patients and parents received the same content, with information presented differently – either in paper format or through electronic multimedia information. Full details of the SWAT will be published elsewhere. 24

Decline consent and withdrawals

Participants (or their parents) were able to decline consent initially or withdraw consent for the trial at any time without prejudice. A decision to decline consent or withdraw did not affect the standard of care the patient received. Participants (or their parents) could withdraw by contacting the central research team by telephone or e-mail. If a patient withdrew, any data collected up until the time of withdrawal were retained by the research team and included in the final analysis. Withdrawn patients or patients deemed ineligible after randomisation were not replaced.

Blinding

Participants and their parents could not be blind to their treatment. The treating clinician was, of course, not blind to the treatment they were providing. However, the treating clinical team did not take part in the follow-up assessment of the participants. The outcome data were collected directly from the participant and/or their parent.

The offer of a soft bandage and immediate discharge

A simple bandage, such as a gauze bandage or similar, was offered to participants. Whether or not to use, and when to discontinue the use of, the bandage was at the discretion of the child and their parents. For those choosing to use the bandage at the outset, this was applied in the ED. The bandage technique involved application to the wrist from the middle of the forearm to the level of the metacarpophalangeal joints. For those choosing not to use the bandage at the outset, a bandage was offered should they wish to apply this at home. Participants were discharged from the ED with no further planned outpatient follow-up (as per NICE guidance18). It was advised that the child could return to activities as pain allowed, a point of contact for any ongoing concern was provided and no specific restrictions on movement were in place. It was advised that the bandage should not be worn for more than 3 weeks. Details were sought from the patient and/or parent related to the duration that the bandage was worn.

Rigid immobilisation and follow-up as per the protocol of the treating centre

A rigid splint was applied that was either manufactured to conform to the wrist (e.g. futura splints) or was moulded to conform the wrist (e.g. backslab, plaster cast). The study was pragmatic and the exact type of splint was not prescribed to treating clinicians. A record was made of the type of splint used. Treatment advice and follow-up was as per the usual practice of the treating centre. Details were sought from the patient and/or parent related to the duration that the rigid immobilisation was worn.

Rehabilitation

Physiotherapy did not typically form a part in the management of these injuries, and no specific guidelines were offered to clinicians or patients.

Primary outcome

The primary outcome measure for this study was the Wong–Baker Scale,28 which is a validated self-reported tool. It is an ordinal assessment of pain using a series of six facial expressions to illustrate the degree of pain intensity. A numerical rating is assigned to each face (from 0, ‘no hurt’, to 10, ‘hurts worst’). It has been validated for use among children aged > 3 years, including in the paediatric ED,29 with its use being most established in those aged > 5 years. 6,30 It has been identified as an excellent measure of pain when estimating the effect of treatments in the ED, and is highly correlated with the visual analogue scale (VAS) (r = 0.90; p < 0.001). 29 The test–retest reliability is excellent (r = 0.90; p < 0.001). 31 The Wong–Baker Scale is widely used in clinical practice, forming part of the Royal College of Emergency Medicine ‘composite tool for the assessment of pain in children’ produced in 2013 as part of a best practice guideline,18 and was recently specifically highlighted for use by the NICE major trauma guidelines. 32

Secondary outcomes

The secondary outcome measures in this trial were as follows.

Sample size

The primary outcome was the six-category Wong–Baker Scale score at 3 days. 30 The Wong–Baker Scale has demonstrated a very high correlation with a standard 0–100 mm VAS. 29 Each face equates to 2 points on the six-category Wong–Baker Scale. The minimally clinical important difference on the Wong–Baker Scale is one face (i.e. 2 points), which was determined in the setting of the paediatric ED. 29 This trial was designed to investigate equivalence of the offer of a bandage to the use of rigid immobilisation, assessing the difference in means on the Wong–Baker Scale at 3 days post randomisation. Assuming an equivalence margin of 1 point (half of the minimally clinical important difference), 90% power, conducting two one-sided tests at 2.5% significance and a standard deviation (SD) of 2.3 (based on results from a feasibility study37), 278 patients (139 per group) with primary outcome data were required to show equivalence.

The Wong–Baker Scale is a categorical outcome that may behave non-linearly in some instances (i.e. the magnitude of pain within the intervals is not uniform), with non-linearity most likely in younger age groups, tending to linearity in those aged > 8 years. 38 Therefore, the trial was powered for equivalence separately in the two subpopulations (i.e. 4–7 years and 8–15 years), which is also important for the secondary outcomes.

Allowing for 20% loss to follow-up inflated the sample size to 348 in each of the subpopulations (174 per group). Given that the primary outcome was measured at 3 days post randomisation, the loss to follow-up inflation could be readily adjusted to ensure that the study recruited effectively and efficiently. Sample size calculations were performed in PASS [PASS 13 Power Analysis and Sample Size Software (2014); NCSS, LLC, Kaysville, UT, USA; www.ncss.com/software/pass].

We planned on collecting primary outcome data for a minimum of 556 patients, a minimum of 278 in the 4–7 years age group and a minimum of 278 in the 8–15 years age group. Allowing for 20% loss to follow-up, we anticipated recruiting 696 patients in total (348 patients in each group).

Analysis plan

The statistical and health economic analysis plan for this trial has been published previously. 20 The methods used for the statistical analysis are summarised here.

Costing of the treatments

Some of the assumptions made when cleaning, analysing or costing the data included the following:

• If a patient answered ‘no’ to a prompt question about resource use, the frequency of service use for this category of resources was equal to zero.

• If the drug use box was checked at all time points, that is days 1, 3 and 7, drug use was considered continuous up to the last checked period.

• Treatments were delivered by either accident and emergency junior doctors or emergency nurse practitioners.

Valuation of resource use

Unit costs for each resource item associated with the trial were sourced from the latest national sources, such as the NHS Supply Chain Catalogue 2018/1942 and NHS Reference Costs 2015–16. 43

The unit costs of the different forms of immobilisation applied (i.e. futura-type splint, backslab, soft cast or hard cast) and the soft bandage were sourced from the latest NHS Supply Chain Catalogue 2018/19. 42

As the injury was primarily managed within the ED, any potential cost of hospitalisation, as defined by the Healthcare Resource Group code, was expected to be the same between the treatment groups.

The unit costs of direct medical costs that were not part of the trial treatments, such as outpatient care and community care, were sourced from the latest available NHS Reference Costs43 and Unit Costs of Health and Social Care. 44

The unit cost of medication related to wrist injury was sourced from the latest available British National Formulary (BNF)45 based on the assumed daily dose using BNF recommendations.

Calculation of utilities and quality-adjusted life-years

The HRQoL of participants was estimated using the EQ-5D-Y, a child-friendly version of the EuroQol-5 Dimensions (EQ-5D), at baseline, 3 days, 7 days, 3 weeks and 6 weeks post randomisation. The EQ-5D-Y instrument estimates a respondent’s HRQoL (sometimes referred to informally as a ‘utility’) – in this context a preference-based valuation placed on an individual’s particular health outcome. The EQ-5D consists of five health state dimensions (i.e. mobility, self-care, usual activity, pain/discomfort and anxiety/depression). There are three levels of health status to choose from: no problems, some problems and a lot of problems. Each participant or their proxy reported the participant’s present health at the date of questionnaire completion. In addition, they self-rated their health at the time of survey completion using a VAS, a non-preference-based measure.

As there is no validated tariff for estimating EQ-5D utility based on the EQ-5D-Y, we used the UK time trade-off tariff for the adult version of the EQ-5D questionnaire. 46 A recent review of patterns and trends of measurement and valuation of childhood health utilities47 found that 78.7% of the studies employing the EQ-5D-Y used general adult-derived tariffs, with the rest providing no information about the tariff used.

Quality-adjusted life-years (QALYs) were calculated as the area under the utility curve of the EQ-5D utility scores using baseline data and data obtained 3 days, 7 days, 3 weeks and 6 weeks post randomisation using the trapezoidal rule. 48

Missing data

Imputation and estimation was conducted according to good practice guidance using the multiple imputation framework within Stata. 49 Multiple imputation provides unbiased estimates of treatment effects if data are missing at random (i.e. causes of missingness are explained by observed variables). This assumption was explored in the data using logistic regression of the missingness of costs and QALYs against baseline variables. 50 Imputation models used fully conditional multiple imputation by chained equations methods, which are appropriate when correlation occurs between variables. Each multiple imputation ‘draw’ provided a complete data set, which probabilistically reflected the distributions and correlations between variables. The imputation process was partitioned to run independently for the two treatment groups. Predictive mean matching drawn from the five nearest neighbours (knn = 5) was used to enhance the plausibility and robustness of imputed values, as normality may not be assumed. An analysis of multiple draws was conducted with Stata’s multiple imputation framework providing estimation adjusted for Rubin’s rule. 51 Within the imputation, missing costs and EQ-5D-Y scores were imputed for each period of follow-up and aggregated to overall patient costs and QALYs for each draw. All imputed variables acted as predictive variables, supplemented by trial baseline variables if significant and plausible predictors of missingness. Multiple imputation estimation models were bootstrapped to provide non-parametric estimates. Initially, the imputation model employed 10 draws, reflecting the proportion of missing data. To minimise the information loss of finite imputation sampling, the fraction of missing information was assessed, ensuring that the number of draws exceeded the fraction of missing information percentage.

Cost-effectiveness analysis

Using ITT principles, the incremental cost-effectiveness ratio (ICER) was estimated comparing (1) the offer of a soft bandage and immediate discharge (as the ‘new’ treatment) with (2) treatment with rigid immobilisation (‘current’ practice) with follow-up as per the protocol of the treating centre. The ICER and CI were estimated using bivariate analysis (complete-case analysis) or multiple imputed bivariate analysis (base-case analysis and other sensitivity analyses). Bootstrapped models were used to report summary estimates (median and percentiles) of group costs and QALYs and to graphically visualise the ICER plane, net monetary benefit (NMB), cost-effectiveness acceptability curve and expected value of perfect information. Value for money was determined by comparing the ICER with several willingness-to-pay thresholds, using an upper threshold for NICE ‘regular’ approvals of £30,000 per QALY,52 a central value of £20,000 per QALY and a lower value of £15,000 per QALY, which reflects the uncertainty about the true value appropriate to the NHS. 53

To assess the robustness of findings, base-case assumptions were explored using a range of supportive sensitivity analyses. A planned subgroup analysis explored the interaction of age group (i.e. 4–7 years or 8–15 years) with the findings. Further planned sensitivity analyses included a complete-case analysis (without imputation and assuming ‘missing completely at random’) and a broader societal perspective (including productivity losses and loss of earnings). Some participants were recorded in recruitment centre-reported data as having attended hospital but did not self-report these visits. To investigate the effect on the cost evaluation, a post hoc sensitivity analysis assumed that participants attended hospital at least as often as reported in the recruitment centre-reported data (i.e. if a participant reported no additional hospital visits, yet the recruitment centre had recorded one additional hospital visit for that participant, then ‘1’ was assumed to be the correct assignment). It was unclear if such visits occurred in outpatient departments or EDs; therefore, an average cost of hospital services was applied to any additional visits. The sensitivity analysis recalculated total NHS costs and explored the impact on cost-effectiveness.

Analyses and modelling were undertaken using Stata 16 and reporting follows the Consolidated Health Economic Evaluation Reporting Standards (CHEERS) statement. 54

Ethics committee approval

The National Research Ethics Committee approved this study on 16 November 2018 (18/WM/0324).

Trial Management Group

The day-to-day management of the trial was the responsibility of the clinical trial manager, based at Nuffield Department of Orthopaedics, Rheumatology and Musculoskeletal Sciences, and supported by the OCTRU staff. This was overseen by the Trial Management Group (TMG), which met monthly to assess overall trial progress. It was the responsibility of the trial manager to train the research teams at each of the recruitment centres. The trial statistician and health economist were closely involved in setting up data capture systems, design of databases and design of clinical reporting forms.

Trial Steering Committee

The TSC, which included independent members, provided overall supervision of the trial on behalf of the funder. Its terms of reference were agreed with the HTA programme and were drawn up in a TSC charter that outlined its roles and responsibilities. Meetings of the TSC took place at least once per year during the recruitment period. The responsibilities of the TSC included monitoring and supervising the progress of the trial towards its interim and overall objectives, reviewing relevant information at regular intervals from other sources (i.e. similar studies/trials newly reported in the literature), considering the recommendations of the Data Safety and Monitoring Committee (DSMC) and informing the funding body on the progress of the trial.

Data Safety and Monitoring Committee

The DSMC adopted a DAMOCLES charter, which defined its terms of reference and operation in relation to oversight of the trial. It did not perform any formal interim analyses of effectiveness. However, it reviewed summaries of data accrued to date by treatment group and assessed the screening algorithm against the eligibility criteria. It also considered emerging evidence from other related trials or research and reviewed related SAEs that had been reported.

Patient and public involvement

The trial was co-produced with families from the outset, including in the development of the funding application.

We initially planned a trial to address the NICE research recommendation, comparing ‘no immobilisation’ with ‘bandages or splinting’. However, this proposal was discussed with the Generation R Young Persons Advisory Group and at the Parents and Carers Forum at Alder Hey Hospital. Both groups were clear that ‘no treatment’ was not an acceptable treatment to families, favouring a ‘bandage’, even if young people or their parents chose not to use the bandage. ‘No immobilisation’ was therefore adapted, with guidance from young people and parents, to ‘the offer of a bandage’.

Throughout the study design, families helped determine the primary outcome tool and the primary outcome time point. Early pain, rather than function, was a key concern for families. How well that pain was resolved should be measured on several occasions during the first few days, and it was agreed that the primary outcome measure would be pain recorded on the third day following the injury. To determine the outcome tool, parents and children were shown a number of different pain tools with similar scientific properties (i.e. the Wong–Baker Scale and the FACES Pain Scale Revised)55 to ascertain which they preferred. The Wong–Baker Scale was the tool preferred by parents and children.

Parents also helped determine the method of collection of outcome data. Parents preferred to respond to periodic text messages rather than telephone or ‘diary’ responses. The timing of the data collection was chosen to coincide with school closure (i.e. 16.00), with reminders closer to the child’s bedtime (19.00).

Once funded, a competition was held using an online design agency to design the study logo. The process yielded around 10 excellent designs. To select the winning design, over 100 children in schools and hospitals voted for the winner. Alongside this, the trial team was engaged in undertaking school assemblies and broader ‘research’ education to engage the public in what was being undertaken.

To ensure ongoing patient and public involvement, a parent representative (PG) was actively involved in the day-to-day management of the trial. This parent representative offered regular insights into the trial from a parent perspective, and regularly discussed the trial progress at the Parents and Carers Forum at Alder Hey Hospital. In addition, a further independent parent representative was a member of the TSC.

One of the largest pieces of engagement at the beginning of the trial was the development of trial recruitment materials. This involved an iterative process of developing an ‘explainer video’. A script was co-produced by the trial team, and this was shared widely to ensure that the text was accessible to families, with the treatments presented in a ‘balanced’ way. The parent representative (PG) shared these with a Parents and Carers Forum, as well as families with children, to seek to ensure that the objectives of the team were met. Iterative changes were made to the text, and then an explainer was produced by a professional design company. Families again reviewed the animation produced. In the original animation, the rigid immobilisation was multicoloured, and children commented that they ‘prefer the rigid immobilisation’ because of the coloured splint. Based on this, the video was revised to ensure that children identified rigid immobilisation to be as equally appealing as the bandage.

A dissemination video that is closely aligned to the study explainer animation (i.e. using the same cartoon characters) has been produced. The text and animation in the video underwent a similar development process to the explainer animation to ensure that the trial result is broadly accessible to the public. The video will be shared through social media, and will be made available to hospitals to ensure that it is available to be played in radiology department and ED waiting rooms (i.e. the places it will be most relevant and impactful to affected families).

Stratification factors by treatment groups

The stratification factors (i.e. recruitment centre and age group) are summarised by treatment group and overall using numbers and percentages in Table 4. There were around twice as many recruits in the older age group (i.e. 8–15 years) as in the younger age group (i.e. 4–7 years).

Recruitment by month

A seasonal pattern to recruitment was anticipated in this trial, with substantially more recruits in the summer months than in the winter months. The trend in terms of monthly recruitment rates is summarised in Figure 3. The rate of recruitment is calculated as the total number of recruits divided by the number of recruitment centres open.

FIGURE 3.

Recruits per recruitment centre per month. Note that the dashed line indicates 23 March 2020, when restrictions on activity were imposed in the UK owing to the lockdown associated with the COVID-19 pandemic.

Treatment allocation

Participants who did and participants who did not receive their allocated treatments at baseline are summarised in Table 5, along with details of the type of immobilisation used and whether a bandage was applied or given. The average duration of treatment, as well the number of patients still being treated, at 3 weeks post randomisation and the number who changed from their allocated treatment prior to 3 weeks post randomisation, is also summarised in Table 5. In addition, details of all hospital-initiated immobilisation changes are summarised by time of change and overall. This includes instances where a participant changed to another treatment within the same arm (e.g. from a splint to a cast).

Participants were considered to have crossed over if they changed from their allocated treatment on or before the 3-day follow-up time point. In total, there were 36 crossovers in the offer of a bandage group (7.4%) and one crossover in the rigid immobilisation group (0.2%). Crossovers are summarised by treatment group in Table 6. Participants who changed treatment after day 3 are also summarised. The reasons for changes are summarised, and the crossovers and later treatment changes are also summarised separately by age group.

Available data

Follow-up was completed between 17 January 2019 and 27 August 2020. Availability of each of the primary and secondary outcomes at each follow-up time point from day 1 to week 6 post randomisation by allocated treatment and by age group is summarised in Appendix 4, Table 34. Overall follow-up rates were high: approximately 94% at the 3-day follow-up time point.

Withdrawals and protocol deviations

Only five participants withdrew during the trial. Withdrawals and reasons for these are reported by treatment group in Table 7. The time in days from randomisation to withdrawal is also summarised in Table 7.

Twenty protocol deviations were recorded during the trial. These are summarised by treatment group in Table 8, along with the reasons for the deviation. Four protocol deviations relating to the diagnosis were recorded by recruitment centres. Owing to the pragmatic approach to diagnosis of torus fractures taken in this trial (i.e. the diagnosis was made by the treating clinician, thus emulating routine care), these were not considered eligibility errors. These deviations were recorded as ‘query about diagnostic criteria; participant remains eligible’ and participants were not excluded from the PP population for this reason.

Baseline participant characteristics

The baseline characteristics are summarised by treatment group, both overall and for each of the age groups separately, in Table 9. The treatment groups appear to be well balanced in terms of these characteristics at baseline. The proportion of female participants was larger in the younger age group (50.3%) than in the older age group (34.3%), but, otherwise, the two age groups were similar.

Patient-reported outcome measures at baseline

Baseline data were collected from the participant or parent after consent was obtained and before randomisation occurred. The patient-reported outcome measures (PROMs) (i.e. the Wong–Baker Scale, PROMIS and EQ-5D-Y) are summarised by treatment group overall and for each age group separately in Table 10. The treatment groups appear to be well balanced at baseline.

Wong–Baker Scale scores at day 3

The primary outcome in the FORCE trial was the Wong–Baker Scale score at 3 days post randomisation. The Wong–Baker Scale scores at day 3 were similar in the offer of a bandage group (mean = 3.21, SD = 2.08) and the rigid immobilisation group (mean = 3.14, SD = 2.11). The trial was designed to test the equivalence of the two treatments in terms of the Wong–Baker Scale score, with a prespecified equivalence margin of 1 point. Comparisons between the two groups in both the ITT population (adjusted difference = –0.10, 95% CI –0.37 to 0.17) and the PP population (adjusted difference = –0.06, 95% CI –0.34 to 0.21) indicated that any difference between the two treatments was less than the prespecified 1 point on the Wong–Baker Scale and, therefore, equivalence was concluded (Table 11). The trial was also powered to separately assess equivalence in the two age groups. The results of these comparisons are also presented in Table 11 and indicate equivalence of the two treatments in both of these subgroups. Similarly, the prespecified PP analyses also demonstrate effect estimates that are well within the equivalence margins. These results are presented graphically in Figure 4, where they are compared with the equivalence margin.

FIGURE 4.

Day 3 Wong–Baker Scale score treatment effects compared with equivalence margin (dashed lines).

Wong–Baker Scale score from day 1 to week 6

In addition to the scores at 3 days post randomisation, the Wong–Baker Scale score was also collected at baseline and at days 1 and 7 and weeks 3 and 6. Scores are summarised by treatment group in Table 12. The Wong–Baker Scale scores over time are summarised separately for the two age groups in Figure 5, which demonstrates that, in the first few weeks after the injury, pain scores are slightly higher in the older age group (regardless of treatment received); however, this difference has disappeared by week 6. Adjusted differences between the two groups at each time point are also provided in Table 12, along with associated 95% CIs. All of the effect estimates and 95% CIs lie within the prespecified equivalence margin (1 point). AUCs up to day 3 post randomisation and up to week 6 post randomisation were also calculated for each group and compared. A statistically significant difference between the two groups in terms of AUC up to 3 days post randomisation is identified; however, it is clear that this value is not clinically significant.

FIGURE 5.

Wong–Baker Scale score by treatment group and age group from baseline to 6 weeks post randomisation (ITT population).

Patient-reported outcome measures

Patient-reported outcomes measurement system scores from baseline to week 6 for the ITT population are summarised by treatment group overall and for each age group in Table 13. Trends over time are summarised graphically for each treatment group in each age group in Figure 6. In all groups, the scores increase over time, with a particularly marked increase between day 7 and week 3. Scores were, in general, slightly higher in the older age group than in the younger one. Comparisons between the two treatment groups at each time point overall and for each age group are presented in Table 13. No significant differences between the two groups were identified. This analysis was also repeated for the PP population (see Appendix 4, Table 35).

FIGURE 6.

The PROMIS scores by treatment group and age group from baseline to 6 weeks post randomisation (ITT population).

The EQ-5D-Y utility scores from baseline to week 6 for the ITT population are summarised by treatment group overall and for each age group in Table 13. Trends over time are summarised graphically for each treatment group in each age group in Figure 7. In all groups, the scores increased over time. There is little difference between the two age subgroups. Comparisons between the two treatment groups at each time point overall and for each age group are presented in Table 13. No significant differences between the two groups were identified. This analysis was also repeated for the PP population (see Appendix 4, Table 35).

FIGURE 7.

The EQ-5D-Y utility scores by treatment group and age group from baseline to 6 weeks post randomisation (ITT population).

The EQ-5D-Y VAS scores from baseline to week 6 for the ITT population are summarised by treatment group overall and for each age group in Table 13. Trends over time are summarised graphically for each treatment group in each age group in Figure 8. In all groups, the scores increased over time. Scores were consistently higher in the younger age group than in the older age group. Comparisons between the two treatment groups at each time point overall and for each age group are presented in Table 13. No significant differences between the two groups were identified. This analysis was also repeated for the PP population (see Appendix 4, Table 35).

FIGURE 8.

The EQ-5D-Y VAS scores by treatment group and age group from baseline to 6 weeks post randomisation (ITT population).

Satisfaction scores at day 1 and at week 6 are summarised separately by treatment group for the ITT population in Table 13. Overall satisfaction with the treatment received was very high, with a score of 1 representing ‘extremely satisfied’ and a score of 2 representing ‘very satisfied’. Mann–Whitney U-tests were used to compare the two groups at each time point. At day 1, satisfaction was significantly higher in the rigid immobilisation group than in the offer of a bandage group; however, by week 6 there was no significant difference between the groups (see Table 13). These analyses were repeated for the PP population and the results were similar (see Appendix 4, Table 35).

Receipt of pain medication

The numbers and percentages of participants who, at days 1, 3 and 7 post randomisation, had received pain medication in the previous 24 hours are summarised by treatment group in Table 14. The proportion of participants receiving pain medication decreased from 80% in the first 24 hours to only 25% by day 7. The proportion of participants receiving pain medication was larger in the offer of a bandage group than in the rigid immobilisation group at each time point. Comparisons between the two groups are reported at each time point (see Table 14). The difference between the two groups of the ITT population was significant at day 1 (OR 0.53, 95% CI 0.28 to 0.98; p = 0.04). No significant differences between the two groups were identified in the PP population (see Appendix 4, Table 36).

The proportion of participants receiving pain medication is also summarised separately for the two age groups in Table 14. More participants in the older age group than in the younger age group were receiving pain medication at each time point. Comparisons between the two treatment groups are also reported for each age group (see Table 14). No significant differences between the two treatment groups were identified for the younger age group; however, in the older age group, more participants in the offer of a bandage group than in the rigid immobilisation group were receiving pain medication at days 1 and 3.

The types of pain medication received at each time point are reported for the overall population in Table 15. Almost all the medication received was either paracetamol (approximately 80% of participants receiving pain medication at each time point) or ibuprofen (approximately 50% of participants receiving pain medication at each time point). Participants could report more than one type of medication at each time point.

Wong–Baker Scale scores by pain medication status

Wong–Baker Scale scores at day 3 are summarised separately by treatment group for those who did and did not report taking pain medication in the preceding 24 hours in Table 16. Pain scores were higher among those who reported taking pain medication in the past 24 hours than among those who did not. The treatment effect estimates and associated 95% CIs all lie within the prespecified equivalence margin of 1 point. Summaries by treatment group and effect estimates and 95% CIs are also presented separately for the two age groups in Table 16. All effect estimates and CIs lie within the equivalence margin.

Complications

Overall, only eight participants experienced a complication during the trial, and these were relatively evenly split between the two treatment groups (Table 17). Seven of the complications were due to an alternative fracture being identified, with one caused by another reason. Details of the types of alternative fractures are also provided in Table 17.

School absence

The number and proportion of participants reporting school absence up to week 3 are summarised by treatment group in Table 18 and Appendix 4, Table 37. Around 24% of participants missed some school in the first 3 weeks post randomisation. Comparisons between the two treatment groups identified no significant differences in the proportions missing school. For participants who missed school, the average number of days missed was similar in both treatment groups at 1.5 days. These analyses were performed for both the ITT (see Table 18) and PP (see Appendix 4, Table 37) populations with similar results.

Audit results

Fracture diagnoses performed during the screening of randomised patients were compared with the formal medical report. Twelve recruitment centres completed the audit, including 218 randomised participants. In 84% (95% CI 80% to 89%) of participants, the clinician confirming eligibility and the reporting radiologist agreed that there was a torus fracture. There was disagreement in 16% of cases: 7% (95% CI 4% to 10%) were reported to the radiologist to have ‘no fracture’, 7% (95% CI 4% to 10%) were reported as a greenstick fracture, 1% (95% CI 0% to 3%) were reported as a Salter–Harris II fracture, and 0.5% (95% CI 0% to 1%) were reported to have an unspecified ‘fracture’.

Use of health resource and data completeness

Health resource by treatment group is shown for the period from treatment to 3 weeks post randomisation in Table 19 and for the period from 3 to 6 weeks post randomisation in Table 20. Table 21 shows the response rate for resource use and EQ-5D-Y by follow-up points and treatment group.

Health-care resource use and costs

Information on the type of interventions and the time to apply them was collected by a survey sent to recruitment centres.

Cost-effectiveness results

The ICER plane in Figure 9 shows the joint distribution of incremental cost and QALYs for the base-case analysis. Patients allocated to the offer of a bandage treatment experienced a marginally larger average quality of life gain (0.0013 QALYs, 95% CI –0.004 to 0.003 QALYs) and incurred lower average health costs (–£12.55, 95% CI –£19.51 to –£5.30) than those in the rigid immobilisation group (Table 31). The probability of the offer of a bandage being cost-effective was > 95% at each of the cost-effectiveness thresholds of £15,000, £20,000 and £30,000 per QALY (Figure 10 and see Table 31). The NMB associated with the offer of a bandage was found to be positive and increased with willingness to pay (Figure 11). Therefore, the base-case analysis indicates that the offer of a bandage rather than rigid immobilisation is likely to be cost-effective in the studied population.

FIGURE 9.

Incremental cost-effectiveness plane with 95% credible region, showing the base-case analysis of bandage compared with rigid immobilisation.

FIGURE 10.

Base-case analysis of the offer of a bandage compared with rigid immobilisation: cost-effectiveness acceptability curve.

FIGURE 11.

Base-case analysis of the offer of a bandage compared with rigid immobilisation: NMB.

Sensitivity analyses showed similar results that supported the base-case finding: in each case, the offer of a bandage proved the dominant strategy. The ICER, when the complete-case analysis was implemented, was –£10,680 per QALY and for the societal perspective the ICER was –£9890 per QALY. The probability that use of a bandage was cost-effective when compared with rigid immobilisation within the studied population was generally > 95% in both sensitivity analyses.

The planned subgroup analysis of age found that, in the offer of a bandage group, costs were higher and quality-of-life gains smaller in the 4–7 years age group than in the 8–15 years age group, as shown in Table 31.

When the assumption was made that participants attended hospital at least as often as was identified by the recruitment centre-reported data, the results again supported the base-case analysis. The ICER was –£9432 per QALY. The probability that use of a bandage was cost-effective when compared with rigid immobilisation within the studied population was > 95% in this sensitivity analysis.

The expected value of perfect information per patient in the base-case analysis was about £1 at a willingness-to-pay threshold of £30,000 per QALY (Figure 12). Given that there are about 60,000 emergency attendances for torus fractures of the distal radius in children per year in Great Britain,2,3 realistic technology time horizons of 10–20 years suggest that further research to reduce uncertainty is unlikely to be appropriate.

FIGURE 12.

Base-case analysis of the offer of a bandage compared with rigid immobilisation: EVPI. EVPI, expected value of perfect information.

Primary outcome

This trial showed equivalence in the Wong–Baker Scale scores at 3 days post randomisation between the offer of a bandage group and the rigid immobilisation group in the management of torus fractures of the distal radius in children aged 4–15 years. Both the ITT analysis (analysis by treatment randomised) and the PP analysis (analysis of participants who received their allocated treatment) confirmed equivalence. Furthermore, the trial was powered to separately assess equivalence in each of the age subgroups, and equivalence was confirmed for both groups.

The number of missing data at the primary end point was very small (approximately 5%); therefore, as per the statistical analysis plan,20 no attempt was made to account for missing data.

Secondary outcomes

In keeping with the primary analysis of pain at 3 days, this trial found no evidence of a difference between the two treatment groups at any of the time points up to the final 6-week follow-up, with the exception of day 1. For the day 1 follow-up, the ITT analysis demonstrated a small but statistically significant difference in pain scores (difference –0.36, 95% CI –0.61 to –0.12) favouring the rigid immobilisation group, but this difference was notably smaller than both the prespecified equivalence margin of 1 point and the minimal clinically important difference of 2 points. This finding was not significant in the PP analysis (difference –0.22, 95% –0.47 to 0.03). Interestingly, the number of participants receiving analgesia was similarly slightly larger in the offer of a bandage group than in the rigid immobilisation group, with approximately 5% more children receiving analgesia at each time point during the first 7 days (day 1, 83% vs. 78%; day 3, 57% vs. 51%; day 7, 25% vs. 23%). The analgesia used was, almost universally, simple ‘over-the-counter’ analgesia.

In keeping with the outcome of pain, the secondary outcomes of upper extremity function or quality of life identified no evidence of a difference between the two treatment groups at any of the time points during follow-up. Parental satisfaction was slightly better (extremely satisfied vs. very satisfied) at day 1 among those treated with a rigid immobilisation than among those receiving the offer of a bandage, but there was no difference at the completion of follow-up.

Although differences did not exist between the treatment groups, there were small differences between the age groups. Children in the older age group generally reported slightly higher pain and poorer quality of life at each time point than those in the younger group, but, conversely, this group also reported more rapid functional recovery. Although this difference is small, it may reflect a difference between self-reporting and proxy reporting, which alters the interpretation of the experience.

School attendance was similar, with participants in both groups, missing an average of 1.5 days of school.

The size of this study allowed particular consideration of complications, of which refracture and worsening deformity requiring intervention are the key concerns of families and clinicians alike. Of the 965 children, none was found to have a worsened deformity. One fracture was identified to have a refracture; the patient was initially randomised to the offer of a bandage group but crossed over to the rigid immobilisation group in week 1 because of pain and was treated with a splint. This patient then experienced refracture at around 3 weeks, following a fall. In total, only eight complications were reported, seven of which related to an alternative type than that originally diagnosed by the treating clinician – all of which were treated with cast immobilisation. Owing to the pragmatic nature of the study, these were not considered protocol deviations, as the treating clinician acted in accordance with their usual practice, and there is subjectivity in making the diagnosis. However, as these alternative fracture patterns inevitably were present from the outset, they amount to errors of radiographical interpretation and, therefore, of study eligibility.

Diagnosis audit

Given the unexpected age distribution of participants (i.e. imbalance between the younger and older patient groups) the FORCE trial TSC recommended that the trial team audit the diagnostic agreement between treating clinicians and reporting radiologists to ensure the validity of the diagnoses. The audit took place when the first 250 patients had been recruited to the trial. The decision was made to use the ‘reporting radiologist’ as a baseline, but it should be noted that the radiologist’s interpretation of the radiograph is also prone to misinterpretation, as there is no clear reference standard to follow.

In total, 12 recruitment centres participated in the audit, contributing 212 fractures. There was agreement between the treating clinician and the reporting radiologist in 85% of cases. Seven per cent of cases were reported by the radiologist to be less severe (i.e. no fracture) and 8% of cases were reported to be more severe (i.e. greenstick, growth plate injury or a complete fracture) than the report by the treating clinician. There was, therefore, broad consistency in the diagnosis of torus fractures, but there was some diagnostic uncertainty. Although, in a few cases, this resulted in crossover between treatment groups, the majority, after review by the treating emergency clinicians, continued to be treated as torus fractures. The absence of worsened deformity, irrespective of the fracture pattern, indicates that the treatment approach is likely to be appropriate even in the presence of diagnostic debate among expert clinicians.

The addition of the posters detailing the inclusion parameters for the trial (see Appendix 3, Figure 13) may have improved the diagnostic accuracy beyond that seen in usual care. However, 281 clinicians from 23 recruitment centres recruited patients to the trial, demonstrating the generalisability of the study findings.

Trial management team

Mr Daniel Perry, chief investigator; Dr Ruth Knight, trial statistician; Associate Professor Susan Dutton, senior trial statistician; Dr Juul Achten, research manager; Professor Matthew Costa, mentor to chief investigator; Dr Damian Roland, PERUKI executive member; Dr Shrouk Messahel, PERUKI executive member; Mr James Widnall, lead for trainee and engagement; Ms Jennifer Preston, patient and public involvement manager; Dr Duncan Appelbe, senior information specialist; Mrs Louise Spoors, trial manager; Ms Amender Juss, trial co-ordinator; Mrs Amrita Athwal, senior trial manager; Dr Melina Dritsaki, trial health economist; Professor James Mason, senior trial health economist; Mrs Phoebe Gibson, public member; Miss Gurneet Sur, Miss Kinzah Abbasi and Mr Hugo Strachwitz, trial administrative co-ordinators; and Mrs Moe Byrne and Dr Peter Knapp, TRECA SWAT team.

Trial applicants

Daniel Perry, chief investigator.

Co-applicants: Juul Achten, Matthew Costa, Susan Dutton, Melina Dritsaki, James Mason, Damian Roland, Shrouk Messahel, Phoebe Gibson, James Widnall and Jennifer Preston.

Trial Steering Committee

Dr Catriona McDaid, Methodologist (chairperson and independent member, and a member of the HTA and Efficacy and Mechanism Evaluation editorial board); Professor Deborah Eastwood, Consultant Paediatric Orthopaedic Surgeon (independent member); Mr Fergal Monsell, Consultant Paediatric Orthopaedic Surgeon (non-independent member); Dr Emma Morely and Mrs Amy Moscrop (lay independent members); and Mr Daniel Perry, chief investigator.

Data Monitoring Committee

Professor Richard Body, Professor of Emergency Medicine (chairperson and independent member); Professor Steve Turner, Professor of Paediatrics (independent member); and Mr Charlie Welch, Statistician (independent member).

Statisticians

Dr Ruth Knight and Associate Professor Susan Dutton.

Health economists

Dr Melina Dritsaki and Professor James Mason.

Programming team

Duncan Appelbe, Patrick Julier, Lucy Eldridge and Malcolm Hart.