Retrograde intramedullary nail fixation compared with fixed-angle plate fixation for fracture of the distal femur: the TrAFFix feasibility RCT

Eligibility criteria

The trial protocol17 aimed to include all patients with fragility fracture. Fragility was initially defined using age as a surrogate marker. After reviewing the screening data during recruitment, the Trial Steering Committee (TSC) recommended that an assessment of the feasibility of using a mechanistic definition of fragility should be made within the trial. The minimum age was therefore revised from 50 to 18 years to ensure that all possible fragility fractures were included. The mechanism of injury was recorded during screening and defined using the World Health Organization (WHO)’s definition as a fall from standing height or less.

Inclusion criteria

Patients were eligible for this study if they:

• were ≥ 18 years

• had a fracture of the femur involving the distal two ‘Müller’ squares18

• would, in the opinion of the attending surgeon, benefit from internal fixation of the fracture.

Many of the patients who sustain this injury have some degree of cognitive impairment, similar to patients sustaining hip fracture. Some have chronic cognitive impairment3 and some develop acute delirium at the time of presentation to hospital. 19 Cognitive status is an important independent predictor of patients’ baseline functional, quality-of-life status and likely outcome in this population. 19,20 This is possibly related to patients’ pre-injury independence, post-injury expectations and ability to participate fully in their rehabilitation. Inclusion of these patients was essential to determine specific effect sizes in this subgroup and so that the sample properly reflected the population sustaining this injury.

Exclusion criteria

Patients were excluded from this trial if they had:

• a loose knee or hip arthroplasty requiring revision

• pre-existing femoral deformity

• an arthroplasty that precluded nail fixation.

Patients who sustained injuries to areas of the body other than the lower limbs, which may affect the primary outcome measure, were still included in the analysis. For patients with bilateral fractures of the distal femur, a rare event, both fractures were recorded but only one fracture was included in the trial. The treating surgeon decided which injury to include in the study prior to randomisation.

All patients who were eligible for inclusion in Trial of Acute Femoral Fracture Fixation (TrAFFix) and their personal consultees, as well as all staff members involved in the research and intervention delivery, were eligible to be approached about participating in qualitative interviews as part of the process evaluation.

Screening and trial recruitment

The clinical care team notified the research team of any potentially eligible patients. Non-identifiable patient details were then used for screening in the emergency department and trauma wards at the trial centres. Reasons for patients’ exclusion or ineligibility were recorded.

Consent to trial participation

The nature of these injuries meant that the great majority of patients were operated on immediately or on the next available trauma operating list, depending on access to an appropriate operating theatre. Some patients were unconscious, all were distracted by the injury to their lower limb and its subsequent treatment and all had large doses of opiates for pain relief, potentially affecting their ability to process information. Similarly, patients’ next of kin, carers and friends were often anxious at this time and had difficulty in weighing the large amount of information that they were given about the injury and plan for treatment. In this emergency situation, the focus was on obtaining consent for surgery (when possible) and informing the patient and any next of kin about immediate clinical care. The consent procedure for this trial reflected that of the surgery, with the clinical team assessing capacity before taking consent for the surgical procedure and this capacity assessment then being used to decide on the proper approach to consenting to the research. The appropriate method, as described below, was used to gain either prospective or retrospective consent from the patient or appropriate consultee by a good clinical practice-trained, appropriately delegated, member of the research team.

Conducting research in this ‘emergency setting’ is regulated by the Mental Capacity Act 2005. 21 As patients may have lacked capacity as described above, and because the urgent nature of the treatment limited access to and appropriate discussion with personal consultees, we proposed to act in accordance with section 32, subsection 9b of the Mental Capacity Act 2005 following a process approved by the relevant Research Ethics Committee (REC): the clinical team would make an assessment of capacity as per usual procedures for obtaining consent for a surgical procedure. The clinical team then provided guidance to the research team as to whether the patient had capacity to consent prospectively or if consultee agreement should be sought.

When the clinical team advised that prospective patient consent was appropriate, this was sought by the research team. If the clinical team advised that prospective patient consent was not appropriate, the research team approached an appropriate consultee. When a personal consultee was available, they were provided with the study information. The personal consultee was given the opportunity to ask questions and discuss the study, after which their written agreement was recorded. When a personal consultee was not available then a nominated consultee was identified to advise the research team. The nominated consultee was the patient’s treating surgeon. If that surgeon was a member of the research team, another independent surgeon was identified. Consent or agreement for further participation into the study after surgery was then sought by the patient themselves or a personal/nominated consultee.

Those patients who were able to consent before their operation were always approached for consent before surgery. For those patients who did not consent prior to surgery, the research associate (RA) provided the patients with all of the study information at the first appropriate time when the patient regained capacity. The patients were given the opportunity to ask questions and discuss the study with their family and friends. They were then asked to provide written consent for continuation in the study. Patients were asked to consent to long-term follow-up and data linkage to routine NHS data sets.

For those patients who did not prospectively consent or who had a nominated consultee give prospective agreement and still lacked capacity after their surgery, a personal consultee was contacted to advise the research team about the patient’s continued participation in the study.

Patients or personal consultees who preferred not to be actively involved in the study follow-up were asked if they were willing to consent to the research team using their routinely collected NHS data for the study.

Responsibility for recording and dating both oral and written informed consent or agreement was with the investigator, or persons delegated by the investigator, who conducted the informed consent discussion. Permission was sought to inform the patient’s general practitioner (GP) of their participation in the study.

Recruitment and consent to process evaluation

As part of the initial consent process, patients and their personal consultees were asked whether or not they could be approached about participating in interviews regarding their views on participating in this trial. Written agreement from participants or their personal consultees to be approached for an interview was sent to the TrAFFix research office. Participants who did not consent to the main trial were also given the opportunity to participate in an interview with a member of the research team.

Participants who agreed to be approached underwent a separate informed consent discussion for the interviews with a member of the research team either in person, by telephone, or by post. Written informed consent was provided when the interviews were face to face and verbal informed consent was recorded if the interviews took place by telephone.

NHS staff were initially invited for an interview by telephone or e-mail, and if they agreed to be approached an informed consent discussion took place with a researcher from the TrAFFix office by telephone or in person.

As part of the process evaluation, conversations between researchers and participants or personal consultees were audio recorded. This included the initial discussion about TrAFFix, before informed consent for the main study has been given. When patients agreed, their verbal consent was recorded at the start of the recording. When patients did not agree, the informed consent conversation continued as usual without being recorded.

Settings

The trial was run in seven NHS hospitals across the UK (Box 1).

Sequence generation

Stratification by centre ensured that any clustering effects related to the centre itself were equally distributed across the trial arms. This would not eliminate centre- or surgeon-specific effects, but, because between 10 and 30 surgeons were likely to be involved in the management of the patient group, any one surgeon would treat only two or three patients enrolled in the trial. Patients were also stratified by chronic cognitive impairment, which was judged by the clinical team on randomisation, to ensure that participants with cognitive impairment were allocated evenly across the treatment groups.

Allocation implementation

Eligible participants were recruited by a trial RA or member of the clinical team using a secure online registration and randomisation system provided by the Oxford Clinical Trials Research Unit (OCTRU). Basic information, including the patient’s initials and age and eligibility checks, was entered. The patient then received a trial identifier (ID) that was used on all non-public-facing trial documentation. The treatment allocation was generated prior to surgery, allowing randomisation outside working hours.

Blinding

Participants were informed of their treatment allocation only at the end of the trial. For this trial, surgical scars were evident, which could have given informed participants knowledge of their allocation. The treating surgeons were not blind to the treatment but took no part in the post-operative assessment of the patients. The functional outcome data were collected and entered onto the trial central database by a research assistant/data clerk in the trial central office. The radiographs collected were reviewed by independent assessors.

Post-randomisation withdrawals

Participants could decline to continue to take part in the trial at any time without prejudice. A decision to decline consent or withdraw did not affect the standard of care the patient received. Once withdrawn, the patients were advised to discuss their further care plan with their surgeon. Data collected up to the point of withdrawal were included in the final analysis. Post-randomisation withdrawals were not replaced.

Retrograde intramedullary nailing

Fixation of the fracture was achieved with a proximally and distally locked nail that spanned the entire diaphysis of the femur. All nails were introduced retrograde through the knee joint. In this pragmatic trial, the details of surgical incision and approach, fracture reduction and supplementary fixation with wires or screws were at the surgeon’s discretion as per their normal clinical practice.

Locking plate fixation

Fixation of the fracture was achieved with anatomical distal femoral locking plate and screws. Locking plates were defined as those in which at least one fixed-angle locking screw was placed distal to the fracture. The operating surgeon determined the length, number and type of additional screws. Additional fixation with lag screws and cerclage wires was at the surgeon’s discretion. In this pragmatic trial, the details of surgical incision and approach, fracture reduction, number and type of other screws and supplementary fixation with wires or screws were at the surgeon’s discretion as per their normal clinical practice.

Patients received the same standardised written physiotherapy advice detailing the exercises they need to perform for rehabilitation following their injury regardless of which group they were randomised to. All of the patients were advised to move their toes, ankle and knee joints fully within the limits of their comfort. Weight-bearing status was decided by the treating surgeon, with a preference for early weight-bearing mobilisation immediately or as soon as the surgeon felt appropriate. In this pragmatic trial, any other rehabilitation input beyond the written physiotherapy advice (including a formal referral to physiotherapy) was left to the discretion of the treating clinicians. However, a record of any additional rehabilitation input (type of input and number of additional appointments) together with a record of any other investigations/interventions was requested as part of the 6-week and 4-month postal follow-ups and formed part of the trial data set.

Primary

The primary outcome measures for this study were the participant recruitment rate and the completion rate of the EQ-5D-5L at 4 months post surgery.

Other measurements

Other measurements that were made during the trial are summarised in Table 1.

Risks and benefits

The risks associated with this study were predominantly the risks associated with the surgery: infection, bleeding and damage to the adjacent structures, such as nerves, blood vessels and tendons. Participants in both groups underwent surgery and were potentially at risk from any/all of these complications. There were no data to suggest that the risk is greater in one group than another. We believe that the overall risk profile was similar for the two interventions.

Analysis plan

The statistical analysis plan was agreed with the Data Safety Monitoring Committee (DSMC) at the start of the study.

Software

The main analyses were conducted using specialist mixed-effects modelling functions available in the software package R (The R Foundation for Statistical Computing, Vienna, Austria), for which EQ-5D-5L29 data were assumed to be normally distributed, possibly after appropriate variance-stabilising transformation.

Data validation

Data returned to the trial office were entered into an OpenClinica database (version 3.7; OpenClinica LLC, Waltham, MA, USA) and data that were missing or failed to pass automatic validation checks were queried with researchers at participating sites. A random sample of 10% of CRFs were checked against the database to ensure accuracy of data entry.

Missing data

The reasons and patterns of any missing data, loss to follow-up and participant withdrawals were carefully considered and reported.

Final statistical analyses

The primary outcome measures for this feasibility study were the participant recruitment rate and the completion rate of the EQ-5D-5L during follow-up. The overall monthly recruitment rate was estimated using Poisson regression analysis, with 95% CIs to assess the likely range of rates in a future main study. Data completeness for the primary EQ-5D-5L outcome was calculated as the percentage of randomised participants completing the HRQoL at baseline and at 6 weeks and 4 months after operation.

Baseline demographics (e.g. age, sex, cognitive status) were compared between groups to ensure that an approximate balance was achieved. This was a small study so group treatment effects were unlikely to be estimated with much precision and, consequently, inferences were tentative and reported as such.

The main analysis of the effects of the intervention investigated differences in the primary outcome measure, EQ-5D-5L29 score at 4 months, between the two treatment groups (nail and plate) on an intention-to-treat basis. In addition, a per-protocol analysis was also reported and early EQ-5D-5L status assessed and reported at 6 weeks. Differences between groups were based on a normal approximation for EQ-5D-5L. 22,23 Tests were two-sided and considered to provide evidence for a significant difference if p-values were < 0.05 (5% significance level). The stratified randomisation procedure ensured a balance in cognitive impairment and recruiting centre between test treatments. Although generally we had no reason to expect that clustering effects would be important for this study, in reality the data would be hierarchical in nature, with patients naturally clustered into groups by recruiting centre. Therefore, we accounted for this by generalising the conventional linear (fixed-effects) regression approach to a mixed-effects modelling approach, in which participants are naturally grouped by recruiting centres (random effects). This model formally incorporated terms that allowed for possible heterogeneity in responses for patients from different recruiting centres, in addition to the fixed effects of the treatment groups, cognitive impairment and other participant characteristics that might prove to be important moderators of the treatment effect (e.g. age and sex).

Secondary analyses were undertaken using the above strategy for approximately normally distributed outcome measures, such as DRI. 34 For dichotomous outcome variables, such as complications related to the trial interventions, mixed-effects logistic regression analyses were undertaken with results presented as odds ratios (and 95% CIs) between the trial groups.

In order to assess the relative performance and merits of EQ-5D-5L and DEMQoL in the study population, we planned to compare the two measures in the subgroup of study participants who were unable to self-report EQ-5D. Given the relatively small size of this study, and minority of participants who were unable to self-report, we anticipated that there would be insufficient data to undertake formal statistical testing for differences between measures. However, graphical presentation of the distribution and relationship (correlation) between measures for individuals would provide some assessment of the merits and metric properties (e.g. variance) of the two measures. Agreement or moderate to strong correlation between measures would indicate that the instruments measure the same underlying trait, which we take to be the true HRQoL of the study participants.

Implementation

Evaluation of intervention implementation focused on four areas. These were (1) the reach of the interventions, (2) the fidelity of delivering the interventions, (3) the acceptability of the interventions and study procedures and (4) understanding how delivery was achieved.

The reach of the intervention, that is whether or not the intended audience came into contact with the intervention, was assessed through the rates of screening and recruitment at each centre over the 10-month period. The fidelity of delivering the intervention considered the procedures for screening patients and application of the study eligibility criteria, the content of consent discussions, deviations from the study protocol and delivery and compliance with post-operative rehabilitation instructions. In addition to our evaluation of intervention implementation, we examined the implementation of the study processes in order to inform the development of a definitive trial.

Mechanisms of impact

Process evaluations seek to identify how the delivered intervention produces change. From our process evaluation data and patient and public involvement (PPI) workshop, we identified relevant intermediate outcomes that might be associated with the effect of the interventions on the primary outcomes of interest. These were used to develop a logic model for the feasibility study that worked from the hypothesis that nail fixation is associated with better outcomes then locking plate fixation.

Context

Understanding of the context in which the intervention is delivered is important to understand how the outcome was achieved. Our evaluation of context focused on three areas. These were the national context, contextual similarities and differences between the participating centres and patient factors.

Data analyses

Ethics committee approval

TrAFFix was approved by the Wales REC (reference number 16/WA/0225), study-wide NHS approval was given by the Health Research Authority (Integrated Research Application System 206745) and participating NHS trusts provided local approvals. The study protocol has been published. 17

Trial Management Group

The day-to-day management of the trial was the responsibility of the trial co-ordinator, based at Nuffield Department of Orthopaedics, Rheumatology and Musculoskeletal Sciences (NDORMS), and supported by the OCTRU administrative staff. This was overseen by the Trial Management Group, which met monthly to assess progress. It was also the responsibility of the trial co-ordinator to undertake training of the RAs at each of the trial centres. The trial statistician and health economist were closely involved in the setting up of data capture systems and the design of databases and CRFs.

Trial Steering Committee

A TSC, which included independent members, was responsible for monitoring and supervising the progress of TrAFFix. The terms of reference were agreed with the Health Technology Assessment (HTA) programme manager and was drawn up in a TSC charter that outlined its roles and responsibilities. Meetings of the TSC took place at the beginning and end of the feasibility study and once during recruitment.

An outline of the remit of the TSC was to:

• monitor and supervise the progress of the trial towards its interim and overall objectives

• review, at regular intervals, relevant information from other sources

• consider the recommendations of the DSMC

• inform the funding body on the progress of the trial.

The TSC consisted of five independent experts, a lay member and the chief investigator. Membership of the TSC is given in Acknowledgements.

Data Safety and Monitoring Committee

The study DSMC agreed and adopted an appropriate charter, structured in accordance with DAMOCLES principles,45 which defined its terms of reference and operation in relation to oversight of the trial. The DSMC was not asked to perform any formal interim analyses of effectiveness. The committee did, however, review accruing data and summaries of the data presented by treatment group and assessed the screening algorithm against the eligibility criteria. The DSMC reviewed all related SAEs that were reported and could advise the TSC if, in its view, the trial should be stopped for ethics reasons. The DSMC meetings were held once during the recruitment phase of the study and once at the end of the trial. The trial statistician provided data and analyses requested by the DSMC at each of the meetings. Membership of the DSMC is given in Acknowledgements.

Reach

Screening and recruitment

[The quotations throughout this chapter were produced in the course of research and also appear in Griffin et al. 44 © Author(s) (or their employer(s)) 2019. Re-use permitted under CC BY. Published by BMJ. This is an open access article distributed in accordance with the Creative Commons Attribution 4.0 Unported (CC BY 4.0) license, which permits others to copy, redistribute, remix, transform and build upon this work for any purpose, provided the original work is properly cited, a link to the licence is given, and indication of whether changes were made. See: https://creativecommons.org/licenses/by/4.0/.]

Ninety-one patients were screened and recorded on screening logs during the 10-month recruitment period. Of these, 54 were eligible to participate in the study. After review by the TSC, this number seemed too small given the assumptions made in the development of the protocol. The TSC recommended revisiting screening with the sites to confirm the incident fracture population size. Trainee PIs and PIs at each of the sites were asked to revisit their local clinical databases to confirm the number of potentially eligible patients and, when additional patients were identified, to conduct screening based on the clinical records. This enhanced screening found a further 82 unscreened patients with distal femur fractures who were not included on the screening logs. Of these, 31 met the eligibility criteria for inclusion in the study.

The number of patients added from enhanced screening varied by centre, with between 0 and 34 patients added. In four centres, ≥ 40% of distal femoral fractures were not included on the screening logs and were added after enhanced screening. At interview, staff were confident that few patients were eligible for the study, suggesting that they were unaware that potentially eligible patients were being missed. Some staff appeared confident that they were not missing patients, even in centres that were subsequently found to have missed many patients. Some research teams were unable to provide a fully integrated daily provision to the clinical teams:

With the TrAFFix study, well we just never had the numbers come through and I think the only two that we actually missed were during weekends and I think one was when I was on annual leave and obviously, research nurses don’t work 7 days a week and that was the issue for recruiting.

Staff (RA) 19

Differences in the number of patients missed from screening logs may relate to the experience of the research teams and the research culture within the centres, which are discussed further in the sections Procedures for screening patients and Contextual similarities and differences between the participating centres. For example, in some centres, not all surgeons within the team screened and identified eligible patients and this may have contributed to patients being missed:

Generally it is mainly the PI that screens for them and that would be identifying them. We do if we find them on the trauma list but it would be mainly the PI that would be identifying if the patient was eligible for us and obviously the PI wants to have people in the trial.

Staff (RA) 4

In addition, patients deemed ineligible might not have been recorded on the screening logs in some centres. One surgeon felt he had ‘a very big screening log to actual recruitment’ (staff, surgeon 18) despite having very few patients recorded on the log.

Figure 2 shows the number of patients screened, the number of patients deemed eligible for inclusion in the study and the reasons for exclusion. Of the 85 eligible patients, 23 were included in the study. Thirty-nine eligible patients were excluded because of surgeon preference, two declined to participate, five were missed as a result of staffing (e.g. the PI was not present to confirm eligibility, or the patient was admitted and operated on over the weekend) and one was missed owing to failure of the randomisation programme. The reasons for missing the remaining 15 patients are unknown. Further details of screening are reported by site in Appendix 3.

FIGURE 2.

Summary of screening. RRAMP, Registration/Randomisation and Management of Product. Adapted from Griffin et al. 44 © Author(s) (or their employer(s)) 2019. Re-use permitted under CC BY. Published by BMJ. This is an open access article distributed in accordance with the Creative Commons Attribution 4.0 Unported (CC BY 4.0) license, which permits others to copy, redistribute, remix, transform and build upon this work for any purpose, provided the original work is properly cited, a link to the licence is given, and indication of whether changes were made. See: https://creativecommons.org/licenses/by/4.0/.

Adapted from Griffin et al. 44 © Author(s) (or their employer(s)) 2019. Re-use permitted under CC BY. Published by BMJ. This is an open access article distributed in accordance with the Creative Commons Attribution 4.0 Unported (CC BY 4.0) license, which permits others to copy, redistribute, remix, transform and build upon this work for any purpose, provided the original work is properly cited, a link to the licence is given, and indication of whether changes were made. See: https://creativecommons.org/licenses/by/4.0/.

After the TSC recommended that the age limit be lowered, a further 69 patients were screened. In this group, eligibility increased from 44% to 60% but this did not translate into an increase in successful recruitment into the study. Only 18 out of the 88 ineligible patients were excluded because of age. After the change in eligibility criteria, only two participants were recruited who were < 50 years of age. Only 3 of the 69 patients screened after the change sustained their fracture from an injury involving more energy transfer than a fall from standing height.

The reasons for ineligibility and missed patients varied between sites. Of the 39 patients treated non-operatively, 18 were from a single site, whereas another site did not report any patients treated non-operatively. Similarly, the proportion of eligible patients excluded because of surgeon preference ranged from 0–10 to 15–19 between sites, although two sites did not completely report the reasons why patients were not included as some patients were screened retrospectively during monitoring activities after the study had closed.

Two centres did not record reasons for missing the majority of patients eligible for inclusion in the study. In the remaining five centres, surgeon preference was the most common reason for eligible patients not being included. The proportion of eligible patients excluded because of surgeon unwillingness to randomise in these five centres varied from 43% (three out of seven eligible patients) to 79% (15 out of 19 eligible patients). In three centres, < 50% of eligible patients were excluded because of surgeon preference, and in two centres > 70% of eligible patients were excluded because of surgeon preference. Several factors may have contributed to this variation. Differences in surgeons’ interpretation of the eligibility criteria were evident at interview and are discussed fully in Applying the study eligibility criteria. Some surgeons felt that patients could be ‘borderline’ or not appropriate for both of the treatments and this might have influenced the number of patients they excluded. Research culture and the experience of the research teams may have also influenced the number of patients excluded because of surgeon preference. In addition, of these five centres, the three that were major trauma centres excluded a smaller proportion of eligible patients because of surgeon preference, perhaps reflecting a greater willingness among specialist trauma surgeons to use both technologies.

Combining the baseline demographic data collected for all screened patients with the hospital data supplied from TARN demonstrated that the screened population was over two-thirds women, one-fifth of whom had cognitive impairment and were aged approximately 70 years. Although a little younger than the hip fracture population, which has a mean age in the UK of 83 years,47 this cohort was similar to a hip fracture population.

Fidelity

Acceptability of the interventions to staff and patients

Acceptability of study procedures

Twenty-five patients were invited to participate in the study. Two declined to participate when approached prior to surgery and two declined when approached for retrospective consent to continue in the study after surgery. One participant who consented to participate prior to surgery withdrew from the study after surgery.

Overall, participants were accepting of the study procedures. They expressed altruistic motivations for participation, wanting to help future patients or science. The acceptability of the following study procedures are discussed in turn: (1) nominated consultee agreement, (2) randomisation, (3) blinding and (4) completing clinical reporting forms.

How was delivery achieved?

The overarching concept of facilitating trial recruitment was identified from the interview data collected from participants and staff as part of this process evaluation. This was developed from a phenomenological approach and a thematic analysis (see Chapter 2). Facilitating trial recruitment was defined as the proactive process of navigating patient and family involvement, knowing that it is the right decision to include patients in the trial, and making it work by juggling an array of activities in order to fit in with normal everyday practice. This overarching concept and the three themes it encompasses highlight facilitators of, and barriers to, delivering a complex trauma trial, such as TrAFFix, and aid understanding of how delivery was achieved. Table 8 presents the definitions of the themes and the categories encompassed within the themes along with one example finding for each category.

Existing factors

Patients’ age and health prior to their fracture might contribute to their recovery. The mean age of the study participants was 74 years, and only six were able to move about freely without aids prior to their fracture. Nine participants were classified as frail [either mildly (n = 3), moderately (n = 2), severely (n = 3) or very severely (n = 1)] at baseline. During their interview, some participants described limited mobility prior to their fracture and other health concerns that they felt may have an impact on their recovery:

I could get out on my scooter and I could get to town. I have got a three-wheeler as well and I used to take that on the bus to town.

Participant 5

I have got no feeling in the right leg at all and I am not sure whether that actually let me down or whether I slipped. I just went out the back door across to put some recycling in the bin and next thing I am on the floor. As I say I have got no feeling in the ride side, right leg at all from the outside it is numb so I am assuming that I didn’t pick it up high enough and it just gave way.

OK, so you have had trouble with that leg for a while?

Yes, a long while. I have had a knee replacement that went wrong about 15 years ago something like that.

Very slowly, because I had bronchitis when I came out and they thought it was perhaps due to the heat but my legs, both my legs swelled up very much and so I couldn’t get around an awful lot, but um I took water tablets and they have more or less gone down now but I realised that my left leg, which I broke, the fracture was near where I had osteoarthritis in the knee and I had that before I broke the leg and that is what is holding me back now I think.

Participant 5

With him losing the ability to walk it has really hindered him every which way but I don’t think that’s because of the operation I don’t think that’s because of the treatment he got I just think it’s his age, dementia and just having that time not walking I think it affected his walking ability.

Participant 11

Two surgeons identified factors that may result in patients being treated non-operatively. These were existing comorbidities that cause a patient to be unfit for surgery and a strong preference from patients for a non-operative treatment:

There will be some people who are unfit for surgery for one reason or another, sometimes it may be cardiovascular problems or sometimes it may be that they have skin issues and sometimes they are simply too obese or have other comorbidities, including dementia. Whether to operate or not operate is a decision made on the balance based on the entire patient and not based on the fracture.

Staff (surgeon) 18

Very occasionally, very occasionally they’ll listen to you and say they just really don’t want an operation and I accept the risks that come with that and the potential benefits, I just don’t want an operation. They’re generally very elderly patients or patients who have had bad experiences of operations or things where their spouses have died during an operation and they just want to have a plaster and go home and I get that.

Staff (surgeon) 13

Injury and treatment

Several factors relating to the patients’ injury and the treatment they receive in hospital could have an impact on their recovery after a distal femur fracture. These relate to the injury event, surgery, rehabilitation and care in hospital.

Patients’ recovery might be influenced by the severity of their fracture, the mechanisms of injury and whether or not they sustained other injuries at the time of their fracture. For example, two participants from this study sustained other injuries at the time of their fracture:

But they are very pleased from the hospital anyway, the consultant is pleased with how it is healing, you know what I mean, because they said it was going to be a really tricky operation because I had smashed it in several places. When I came home and looked on the internet you normally only smash it, break your femur if you have a really bad car accident so the thing was as well, I have had a knee replacement in that knee . . . I have had a knee replacement on both my knees prior to this happening.

Participant 7

The time between a patient’s injury and surgery might be important. Two relatives who were interviewed described a delay between their relatives’ injury and surgery. One participant was not taken to hospital until 3 days after he fell, as paramedics did not suspect he had sustained a fracture, and the other participant waited several days for an operating theatre to become available as more emergencies were prioritised:

She had the accident on a Thursday and she was on standby for the theatre all the time until they could operate but obviously as more emergencies came in she kept going down the list so she was actually in hospital from the Thursday to the Monday waiting to operate on . . . well the first problem is because of her age, her skin is very thin and it has caused quite bad pressure sores because obviously they had to put her in traction. The other thing is it has caused a lot of problems with being constipated and things because she didn’t eat hardly anything in that period of time she was on standby to go to the theatre. We have had quite a lot of problems you know and also she lost weight because she was very small anyway and she sort of lost weight because you know she hadn’t eaten a lot before she had the operation.

Participant 6

In addition to the intervention that patients were randomised to, several other factors relating to surgery could be important. This includes the experience of the surgeon, for example the grade of the supervising surgeon, the number of trauma lists the surgeon routinely supervises and the number of distal femoral fractures the surgeon has fixed. In this study, 15 participants were operated on by a consultant, six by a specialist trainee (level 3) and two by a staff associate specialist (SAS).

Surgeons’ preferences for one intervention over the other could also influence patients’ outcomes. Interviews with surgeons revealed that some may not be confident with both techniques, with preferences often based on surgical skill and experience. These fractures are not common and were described by surgeons as difficult to fix. Two surgeons explained that this operation, unlike many others, is not performed from implicit memory. It may, therefore, be more difficult to use their less preferred technique:

Distal femoral fractures are not common and so there are only about 10% of fractures of the femur overall and so no-one is doing lots unless you put your hand up to do them.

Staff (surgeon) 13

Distal femur fractures are in that group of it’s a bit more complicated than average and can really be quite difficult.

Staff (surgeon) 23

It’s not like a routine thing like an ankle fracture that you don’t really think about, it’s not like a brain stem reflex.

Staff (surgeon) 10

Surgeons’ beliefs about rehabilitation and the rehabilitation instructions that they give to patients could influence how quickly patients mobilise following surgery. As described in Protocol deviations, surgeons held different views about the most appropriate rehabilitation instructions for these patients. Although early weight-bearing was favoured by many, some surgeons restricted weight-bearing for some patients:

Personally, I do bed to chair for the first 4 weeks and I don’t allow them to fully weight-bear if they’ve had a plate, which I know is different to a lot of people. It depends a little on what their bone quality is like at surgery and nails will be weight-bear as tolerated. A lot of them will stay in for quite a long time and they rehabilitate pretty slowly on the whole.

Staff (surgeon) 3

In the elderly group I think you’ve got to let them walk straightaway. I just don’t think it’s meaningful to ask a 90-year-old man or woman to walk around on crutches and so I just get the physios [physiotherapists] to get them walking.

Staff (surgeon) 13

Many participants described receiving support from physiotherapists while in hospital. Several participants spoke of the exercises and support they were given to walk enthusiastically:

They gave me the confidence to do it. I don’t know what I would have done if they didn’t . . . if I hadn’t had a good team of physios [physiotherapists], they were really great as I say I was shaking from head to foot because I knew I had got, that this bad leg was going to have to be the leg that I was going to have to use to support me . . . they have got some bigger [walking frames] you know where you put your arms on to and I used those first off and originally they said I could bring one home because I was terrified to go on a normal [walking frame] and they had said I could bring one home, not the one they were using in the hospital but an older model but eventually they got me on to a [walking frame], they gave me the confidence they were absolutely great.

Participant 7

I had to do exercises this morning and they were painful but you know once they got too painful you just stop and then I went downstairs to the gym and with the help . . . of one of those pushers I was able to walk using my wonky leg and my good leg. I couldn’t put too much weight on that one and I walked the length of the gym and back again and they were very pleased with me. I was pleased with myself.

Participant 3

For some participants, their time in hospital was difficult. One participant described being moved to another hospital the day after surgery as other patients were waiting for a bed:

Well I didn’t like it. I wanted out because I don’t like hospitals and so I wanted out.

Participant 8

It was a horrendous feeling for my daughter and for me, you know not knowing where we were going and they just came on the wards and started packing my stuff up and said we want this bed, you have got to go to the [hospital name] and there was no warning, no nothing, no it was awful and when I went to the [hospital name] they more or less said we weren’t expecting you, so they had got no pads or anything.

Participant 9

In contrast, others described being cared for and were pleased with their experience in hospital:

You know nothing is too much for them no matter what time of the night or day because I have got a [call nurse button] . . . So as I say I am quite pleased with them. You know they don’t treat you as an idiot you know they consider that you have got some intelligence left.

Participant 3

Post-discharge factors

Several factors were identified that might influence a patient’s recovery once they were discharged from hospital. These were post-discharge living arrangements, access to carers, loss of independence, family support, access to physiotherapy, psychological factors, such as resilience, self-efficacy and confidence, and pain.

For some participants, recovery was slow and frustrating. At interview, several participants described spending a long time in hospitals, rehabilitation units or care homes before returning to their pre-fracture residence. Ten participants were discharged to their own home or sheltered housing, nine were discharged to rehabilitation units, one to a residential care home and one to an acute hospital. At the 4-month follow-up, 8 out of the 11 participants who completed the health economics questionnaire were living in their own home or sheltered accommodation and three were living in residential care:

I was, I was in there what nearly 6 weeks and then I had the enablement team from the hospital for about 2 or 3 weeks for a couple of weeks or so and now, I have got carers full you know full time.

Participant 2

I am back at home with my bed downstairs. Obviously, I haven’t been upstairs since they actually did the operation.

Participant 7

At the interview, participants described trying to ‘muddle through’ at home but often needed support from family, carers and cleaners to manage. The majority needed help with household chores and a few also needed help washing, dressing and cooking. Their loss of independence could be frustrating and they did not like to be reliant on others or ask for help:

To be honest I don’t want to be where I can’t do things for myself. It’s very frustrating when you have to keep asking people and you know they are thinking what now but they don’t say so half the time and I want to be able to get back up and go shopping if I want to you know . . . well just do what I always did.

Participant 6

Yes, I am back at home, I have a carer in the morning because I can’t stand and wash myself yet and I am incontinent so that makes it difficult because I am not walking on me leg yet. I am walking with a frame just like toe touch toe touch sort of thing with a frame but most of the time I just sit in a wheelchair.

Participant 10

Several participants found ways to contribute to the housework and overcome the restrictions their injury imposed, suggesting resilience. For example, as they struggled to stand for long without support, they ironed sitting down, hoovered from their wheelchair or washed the dishes from a stool:

So I am very pleased with what has happened at home so hopefully I can start . . . I mean I can walk with a frame but you can’t carry a cup of tea or you can’t carry a meal to where you sit to have your meal. So I have to more or less have to stay in the wheelchair whereas the physiotherapist wants me walking everywhere if you know what I mean, to make my legs stronger but I just find it easier to skim around in my wheelchair. I even do the hoovering in the wheel chair, well it’s only a little granny annex so it’s not a big carpet space to do but . . . and I sit and do my own washing up and everything on my perching stool so yea I am quite happy at home.

Participant 9

I have got a cleaner that comes in three times a week for an hour Monday and Tuesday and Thursday and she does the house cleaning because I can’t bend down and get up again. I do my own washing but I am going to try and do my own ironing. It’s the fact that I can’t stand for a long time but my daughter has given me one of those table tops that you can sit down on a chair and iron on it so I am going to try that.

Participant 3

In contrast, some participants became isolated and depressed after their fracture:

Well she was actually getting quite depressed because obviously she can’t get to the front door and obviously I used to take her round in her wheelchair to see her sister and things like that and she wouldn’t do it because she is frightened of . . . You know she has had nobody since she got home she was doing quite well to start with but then as I say we were expecting a physiotherapist to come round but then gradually because nobody has come to assess her or . . . she was getting quite depressed about not being able to get up and things like that and not being able to move round like she could before.

Participant 4

Our interview data highlighted the role of family and carers in supporting patients’ recovery. We also assessed patients’ social support using the Medical Outcomes Study, Social Support Survey. 27 We calculated an overall support score by averaging the scores for the 19 items included in the survey. This provides a score between 1 and 5, with higher scores indicating more support available. Within our sample, the mean overall support score (the average of the 19 items in the Medical Outcomes Survey scale) was 4.18 (SD 0.76). In addition to providing an overall support score, the survey comprises four separate social support subscales: (1) tangible support, (2) emotional/informational support, (3) affectionate support and (4) positive social interaction. The mean scores for the subscales were (1) 4.28 (SD 0.66) for tangible support, (2) 4.22 (SD 0.86) for emotional/informational support, (3) 4.28 (SD 0.75) for affectionate support and (4) 3.87 (SD 1.10) for positive social interaction. These data show that, within the sample, levels of social support varied but tended to be high.

As described in Patient compliance with rehabilitation instructions, some participants experienced a loss of confidence after their fracture or struggled with pain and needed help to move forward with their rehabilitation:

She has also lost confidence after several falls and although she has been able to stand up on the patient turner and even holding onto the gutter frame, she has not been able to move either of her feet when standing.

Participant 10

Just if I walk too far, that’s when I feel it you know, I need to sit down sort of thing give it a rest, at night-time, that’s when I get the pain, at night-time.

Participant 8

Self-efficacy may be an important factor in patients’ recovery as low self-efficacy may result in patients being reluctant to mobilise following surgery. We assessed self-efficacy using an adapted five-item version of the General Self Efficacy scale, which gives respondents a score between 5 and 20, with higher scores indicating more self-efficacy. Within our sample, patients’ levels of self-efficacy varied. The mean self-efficacy score was 13.4 (SD 4.1).

Of the 12 participants who completed the health economics questionnaire, which asked about services they had received, four reported receiving physiotherapy after discharge from hospital. Some families advocated for support for their injured relatives, particularly during the transition from hospital to home. They sought access to physiotherapy, which was not easily available for all patients owing to long waiting lists or in one case because it was initially believed that the patient would not be able to weight-bear again:

No. They haven’t done nothing [sic.] to be honest. Because they were supposed to send me for therapy but I never had nothing.

Participant 8

After the fall which caused the present fracture to her right femur, the physios [physiotherapists] initially said that she wouldn’t be able to weight-bear again and would probably need to be hoisted everywhere for the rest of her life. I objected and said that I planned to have her come to live in my home but would need her to be able to stand up using a patient turner. From that point, efforts were made to help mum to learn to stand and she was allowed to spend a few weeks in an NHS-funded bed for rehabilitation.

Participant 10

Participants also described the difficulty in accessing transport for rehabilitation activities:

I missed the last [physiotherapy] appointment because the ambulance didn’t turn up to take me. Other than that, I have been going to hydrotherapy. I did a 4-week course at hydrotherapy and they offered me to go to the self-help hydrotherapy thing and then they told me that if I go to that they will probably ban me from the ambulance because you are not allowed to have an ambulance for that and if you can get to that without an ambulance then you can get to any other appointment without an ambulance.

Participant 1

There were many challenges for the TrAFFix study that might have affected the mechanism of impact of the intervention, highlighted in Box 3.

National context

To provide a national view of the occurrence of this injury and differing treatments, a brief questionnaire to elicit surgeons’ views was administered. The findings suggest that surgeons perceive distal femoral fractures to be uncommon and hold different preferences for their treatment. Seventy-nine surgeons responded to the questionnaire (details of the questionnaire are reported in Appendix 2), of whom 57% (n = 45) worked in major trauma centres and 76% (n = 60) defined themselves as specialist trauma surgeons; 82% (n = 65) reported that they personally operate on < 10 distal femoral fractures per year (Table 10). Distal femoral fractures are less common than some other fractures and individual surgeons are unlikely to regularly operate on them. This may therefore influence their experience and confidence to perform both interventions noted in the interview data.

Seventy-eight respondents answered the question ‘do you have a preference for operative fixation of most distal femoral fractures?’, with the option of answering anatomical locking plate, retrograde intramedullary nail or no preference. Just over half (54%, n = 42) of the respondents expressed a preference for one method of fixation. Of those who expressed a preference, 62% (95% CI 46% to 76%, n = 26) preferred anatomical locking plates.

Based on limited data from this questionnaire and our interviews with surgeons, there is a sense that anatomical locking plates may be preferred, although most surgeons say that they do not have a preference.

Contextual similarities and differences between the participating centres

Contextual similarities and differences between the participating centres can influence how the study is delivered and how patients recover. These will be discussed in turn.

Seven centres participated in this feasibility study. Four major trauma centres were included. As demonstrated in Table 10, they differed in the size of population served and number of consultants and mix of full-time equivalent staff. This demonstrates that this feasibility study was conducted in a range of settings.

Contextual differences between the participating centres that may have led to differences in how the study was implemented were evident within our evaluation of intervention implementation. These relate to the degree of development of a research culture identified through (1) surgeons’ willingness to randomise patients, (2) a range of ongoing research studies, occurring at the same time as TrAFFix, (3) the frequency of occurrence of injury and (4) the research teams.

Surgeons’ willingness to randomise patients, shaped by their surgical skill, equipoise and beliefs about the appropriateness of research participation for their patients as described in acceptability of study interventions, could influence recruitment. Strong surgeon preference, or in some centres a group preference, for one of the two interventions could lead to some centres including different patients to others. This is discussed in How was delivery achieved?. Centres where there was a range of surgical expertise may have been better equipped to avoid potential protocol deviations or missed patients.

Surgeons’ willingness to randomise patients was also influenced by the emergency nature of trauma work. The fast pace, nature of trauma work and complexity of major trauma meant that, although a patient may have met the eligibility criteria, not all surgeons were prepared to include them. For example, patients with open fractures require surgery urgently to minimise the risk of infection:

There’s been a couple [of eligible patients not included], I remember a lady she had an open tibial fracture and the distal femur as well and I’d spoken to the PI about it and he felt she was eligible but I think that the surgeon who was doing it . . . thought he had the two big injuries to deal with at the same time and it was all very time pressured because she’d got an open fracture and to go to theatre straight away.

Staff (RA) 7

The level of surgeon input required for a study such as TrAFFix is challenging. Some RAs contrasted recruitment to TrAFFix with that of less challenging studies, for example dressing studies, which they believed surgeons were more amenable to as they require less input:

I do feel that at different levels we’ve got quite an open environment to research and they’re very happy us doing it so long as it doesn’t really involve them. So all these new studies that are the more interventional studies that you have to get more of an input from the surgeon, it’s harder. Today we’ve got two who are eligible and it was a struggle to get them to say they were eligible and then a struggle for the surgeon to say yes I’ll do it. So the surgeon will do it as long as someone else will do all the decision-making, does all the paperwork behind him and just literally at the final hour tells him which bit to do because he hasn’t got capacity in his brain to take on the responsibility.

Staff (RA) 5

You find the consultant response is variable, they always seem much more amenable with something like WHIST [Wound Healing in Surgery for Trauma], which is just a dressing and not interfering with what they’re doing whatsoever and they really couldn’t care less and anything you find physio [therapy] orientated never received a huge amount of objection.

Staff (RA) 7

The research culture varied across sites in the degree to which research was embedded in daily trauma activity. A range of studies facilitated recruitment by increasing opportunities for engagement. As described in How was delivery achieved?, surgeons working in departments where several trials are running may be more engaged in research and inclined to participate:

So we’ve got surgeons who are doing a couple of other hand studies and so they would be involved and they’re keen to do them partly because they are doing their own bits of research and they want to get involved with it and partly because we’re doing the WHiTE studies and all of a sudden it is like we’re a team, an orthopaedic research team so I think that’s moving forward.

Staff (RA) 4

The infrequent occurrence of distal femoral fractures inhibited the study from fitting in with the existing research culture. Patients who were eligible to participate in TrAFFix did not present frequently enough for the clinical team to keep the study in mind and the study did not become routine, as described in Procedures for screening patients:

We’ve only been getting one every month or 6 weeks. For example, 6 weeks goes past and the patient comes in and they’ve forgotten about the study because it’s not in the front of their minds and then I have had to pretty much ring every time to speak to the consultant to say have you considered this study, if not why.

Staff (RA) 11

TrAFFix hasn’t been regular enough for it to be routine to some extent.

Staff (RA) 11

An effective research culture required all clinical staff to be aware of the study. If patients were told that they would receive one intervention before they were informed of the study, it could be difficult for them to accept being randomised to one of two different treatments:

I think the only challenge we have is when someone’s already spoken to them about a treatment option and then you suddenly come saying you might do something different . . . If they’ve been told something different to what you then tell them in terms of nail and plate and don’t notice any difference they say the last surgeon they saw said it was definitely going to be better with a plate it’s then very difficult to convince them otherwise.

Staff (surgeon) 3

The research culture was facilitated by experienced RAs, who could facilitate research activity and make things easier for clinical staff. Differences in the size of the research team, their workload and working hours could influence the implementation of the interventions. For example, as described under implementation, the presence of RAs in the trauma meeting could facilitate screening but required RAs to start work before 08.00, which was not possible in every centre. Such structural issues are substantial barriers to the integration of the clinical and research teams, development of a strong research culture and effective delivery of these studies. Meanwhile, some centres relied on one RA or part-time RAs. This might result in patients being missed (at consent or follow-up) either when the RA was away or because they were struggling with their workload:

Yes so for instance the Friday before there were two patients coming in but there was nobody available because either the surgeon was in theatre, the surgeon was on annual leave, the registrar was on annual leave who normally helps with clinic, who helps recruit and so yes that does happen but not all the time. There was nobody in this Friday just gone and it can happen because I’ve got to be everywhere.

Staff (RA) 15

I did once have nine patients in a morning clinic and they were seeing two different consultants so it was trying to be with one consultant and patient doing a follow-up and my other patient I might have missed because they were at the same time with simultaneous appointments.

Staff (RA) 15

For RAs, familiarity with the paperwork was an important factor. For centres that were unfamiliar with Oxford trauma studies,51–54 some staff thought that paperwork could be difficult to get to grips with. Not only was the paperwork considered lengthy, but there were also multiple versions of information sheets and questionnaires. Although complex, many RAs commented on the logical and familiar structure of the paperwork, which facilitated data collection:

I mean I think from a paperwork point of view we’re very well versed in Oxford paperwork so you’ve got the prospective PIS, prospective CIS, retrospective and then you’ve got all the consent forms that go with all the various options and then you’ve got the dementia self-reporting, dementia non-reporting. But we’re quite familiar with that so that’s not an issue and I don’t think it would have been an issue, maybe for some centres that don’t do as many Oxford studies, the level of paperwork and different categories of patients might be a little bit overwhelming, it hasn’t been for us.

Staff (RA) 11

The development of a research culture appears crucial to the success of a study such as TrAFFix. This varied across the centres, with some sites further on than others in their development of a research culture.

Contextual factors relating to the participating centres also had potential to influence patient outcomes after a distal femoral fracture. For example, surgeons’ views about rehabilitation instructions for these patients and the instructions they gave differed between the centres, as described in Protocol deviations. The National Institute for Health and Care Excellence (NICE) recommends that health-care professionals should consider advising patients who have had a distal femoral fracture to immediately weight-bear as tolerated following surgery. 55 Differences in the delivery of this post-operative component could have an impact on patients’ recovery and quality of life after their fracture, as discussed in Chapter 2, Mechanisms of impact:

My whole argument about why I do the operation is because you want to get them to walk, you could treat them with a plaster and you can walk on that so I think if your concept is to treat the patient in a way that allows them to walk immediately then you end up allowing them to walk immediately and I’m strongly of the view that that is what we’re trying to do and so I’m happy to take the risk that some will fail because I let them walk.

Staff (surgeon) 13

Further contextual factors relating to the participating centres that might influence patient outcomes are discussed in Chapter 2, Mechanisms of impact.

Patient factors

Contextual factors relating to the group of patients who were eligible to participate in this study could have the potential to inhibit recruitment and also influence patients’ outcomes after a distal femoral fracture. These factors include frailty, attitudes towards research and access to support.

Patient frailty was the main contextual factor for this population. Despite reducing the minimum age for patients eligible to participate to 18 years, participants were typically older (mean age of 74.1 years). Many were frail, compounded by the shock of injury and surgery, and often experienced some confusion or memory loss following their fracture, as described in Chapter 2, Mechanisms of impact:

No, I don’t [remember] because it was when I was coming through you know being up there and I was trying so hard to be normal.

Participant 3

I probably wasn’t in a fit state to ask anyway but I don’t remember wanting to know anything.

Participant 5

They understand the main idea however . . . they forget easily. They can remember the plate or the nail but it is difficult to remember everything.

Staff (surgeon) 17

Interviews with staff and participants revealed that patients were typically accepting of research. Many participants expressed altruistic motivations for participation and trusted their surgeon to fix them:

They don’t have any idea about what you’re talking about and nor do they particularly care. I think they just think that you’re recommending an operation, they buy into that idea and then you’ll just do a technically sound operation. So when you say well there’s these two different ways of fixing your fracture and one is a nail and one is a plate and we’re not a 100% sure which is better and so we’re doing a study comparing the two where you might end up with either one and I can do both and its reasonable we could do either then they just say yes fine.

Staff (surgeon) 13

Although accepting of research, some participants demonstrated confusion over randomisation and few could describe the study in their own words:

Yes, yes they had to decide because they know what they are doing and I don’t so, they have done hundreds haven’t they.

Participant 5

Well because it helps other people and obviously it would obviously help her the best way you know . . . obviously because of her age her bones weren’t going to mend properly and the doctors did explain and they said that they would give her the best which they thought would help her.

Participant 4

These factors highlight the need for consideration of the information given to this group of patients and how best to approach and engage them in research. In the participating centres, RAs were typically experienced at navigating consent with this group of patients. They understood the complexity of approaching these patients about research participation and sought strategies to mitigate these factors as described in How was delivery achieved?.

Several participants described a loss of confidence following their fracture. This, along with their frailty and limited mobility prior to their injury, as described in Chapter 2, Mechanisms of impact, may make recovery difficult for this group:

When she was in hospital she was actually getting on very well with walking while she was in hospital but as soon as she came home she was frightened of doing it on her own without . . . you know she needed somebody like the lady who came today [name] was very good because she started to get her doing things.

Participant 4

Yes but I did have a walking stick . . . and if I go down town say to [a shop] or anywhere I use three wheels to walk with because I can’t do it for long on my own natural sources if you know what I mean. I have arthritis in both knees and I have got arthritis in my lumber region, so walking isn’t my best thing but when you have to do it you have to find the easiest way don’t you. I think the three wheels have been all right for me up to now but I still haven’t tried them yet [since the surgery], I am still trying with the frame.

Participant 9

These participants required support to aid their recovery and allow them to live independently. As described in Chapter 2, Mechanisms of impact, participants received support from their families and from paid carers. Of those who completed the health economics questionnaire, eight reported receiving unpaid care from relatives or friends, most frequently their adult children (n = 5). Several participants also received resources from charities to enable them to manage at home. As described in Chapter 2, Mechanisms of impact, access to physiotherapy was difficult. Several participants explained that they did not receive physiotherapy after discharge from hospital while two personal consultees described advocating for their relative to ensure that they received physiotherapy. It would seem that access to support from physiotherapists, family and carers is important to these patients during recovery as their frailty may cause them to struggle to manage alone:

Obviously it would have been a lot better if we’d have had the physiotherapists out straight away, which is what I thought was going to happen. But we have actually had to have the doctor out to her because she was getting the bed sores as well due to the not moving round enough, which is why I kept chasing the physiotherapist to come out and sort of assess her and things which is as I say what has happened today so I feel a lot happier today.

Participant 4

The contextual challenges are summarised in Box 4.

Collection of resource use

This study collected data on the use of services within the health-care system, as well as on broader costs to society, in an attempt to capture relevant components of the overall cost of treating fractures of the distal femur. Key outcomes included data completeness and the main cost drivers of the interventions under investigation.

Valuation of resource use

Estimation of unit costs followed recent NICE guidelines on costing health and social care services. 59 All costs presented are in 2017/18 Great British pounds. For unit costs obtained before 2017, prices were inflated to 2017/18 prices as necessary using the NHS Hospital and Community Health Services (HCHS) index for health service resources. 60

The cost of the distal femur fracture surgery was assessed using the latest NHS Reference Costs 2015 to 201661 and Healthcare Resource Group code for ‘Major Knee Procedures for Trauma’ (Table 19). If a patient experienced a length of stay greater than the average length of the stay (i.e. 5 days) reported from the NHS reference costs, then the inpatient excess day cost from the NHS reference cost was used to compute the direct medical cost associated with the trial. 61

The NHS reference cost is the average unit secondary health-care cost to the NHS and would in fact include the implant cost but we have included unit costs of major implant cost components here for future trials that might utilise microcosting (or bottom-up costing). These unit costs were obtained from the 2018 NHS Supply Chain Catalogue62 (see Table 19). Similarly, use of additional implants for the operative treatment of distal femur fractures (e.g. number of screws, wires or cables) was reported in the trial CRFs.

Data on further resource use were collected within the trial at 4 months post surgery via self-reported patient questionnaires and appropriate proxies (when necessary). 29,63 Questionnaires captured the frequency of use of community-based health and social care services (classified as general practitioner, practice nurse, community nurse, physiotherapist, occupational therapist), number and duration of admissions to inpatient wards, number of diagnostic tests, use of outpatient services [classified as orthopaedics (injured bone), orthopaedics (any other bones), rehabilitation unit, physiotherapy, emergency department], medication use and equipment provided, indirect costs borne by partners and carers as a result of attending hospital visits, as well as direct non-medical costs (including travel expenses) attributable to the participants’ health state. Unit cost data were obtained from the latest available national databases, such as the NHS Reference Costs 2015 to 2016,61 NHS Supply Chain Catalogue62 and Personal Social Services Research Unit (PSSRU) Unit Costs of Health and Social Care,64,65 or other sources, such as North Yorkshire County Council’s Paying for Care at Home,66 Independent People Homecare Services’ The Cost of Live in Care67 and the Office for National Statistics’68 Annual Survey of Hours and Earnings: 2017 Provisional and 2016 Revised Results (Table 20), and used to compute the direct medical costs that are not directly related to the trial.

For medications, the defined daily dose for each medication was obtained from the World Health Organization Collaborating Centre for Drug Statistics Methodology website70 using the relevant Anatomic Therapeutic Chemical code. Unit cost of each medication was then sourced using the British National Formulary71 and either tablet or capsule form (if tablet form was not available) of the medication was assumed.

The human capital approach was adopted to estimate the cost of informal caregiving. As the age group of the caregivers was not collected in the CRFs, we assumed the wage to differ by sex only.

Missing data

The reasons for and patterns of any missing data, loss to follow-up and participant withdrawals have been carefully considered and reported, with particular emphasis on how these may impinge on any future trial.

Analysis

A data analysis was conducted on an ITT basis and aimed to determine the main cost drivers of the trial from the NHS and Personal Social Services perspectives. The mean cost and SD were reported for each resource item by treatment allocation.

Response rate and data completeness

Of the 23 participants who were randomised, 52% (n = 12) responded to the health resource use questionnaires. Participants’ perception of the resource use questionnaires in terms of their content and length is reported in Chapter 3, Results of the process evaluation. The quality of response to the resource use questionnaires was acceptable with no missing items in most of the resource categories (Tables 21–23).

Estimation of resource use and costs

The mean usage and cost of the major implant cost components of each intervention per patient are provided in Table 24. The mean total implant cost per patient of the locking plate intervention was higher than that of the intramedullary nail intervention.

The total inpatient cost of intramedullary nail fixation seemed to be higher than that of locking plate fixation as one participant had a much longer length of stay (of 233 days) than the rest of the participants in the same group (length of stay of ≤ 25 days) (see Table 21). Removing this participant would result in a mean length of stay of 14.8 days (SD 7.4 days) and a total surgery cost of £31,000.62 (SD £2722.85).

In terms of direct medical cost that was not directly related to the trial (i.e. broader NHS and Personal Social Services resource use utilisation) (see Table 22), inpatient care following readmission constituted the largest component of direct medical cost for both interventions. Two participants reported the use of outpatient care, such as rehabilitation and attendance at emergency departments, not related to their injury. The costs of formal social care was also a very substantial cost driver compared with the other relatively small costs incurred.

Table 23 depicts the direct non-medical (home adaptations not funded by NHS) and indirect costs (informal caregiving and income loss) borne by the study caregivers and participants. The largest single-cost item associated with the interventions other than the surgery cost was informal caregiving from the participants’ relatives. One participant in the nail group reported that a relative was providing informal care for 24 hours per week for 17 weeks. Removing this participant’s response in a sensitivity analysis would result in a mean cost of £5750.92 (SD £4513.19).

The resource item that had the most number of missing responses was home adaptations. This could be due to the order of the mobility aids and home adaptations questions in the questionnaire. As the mobility aids and home adaptations questions were not in sequence, two participants had reported their home adaptations under other response categories.

Screening

A total of 173 patients were screened during the course of the feasibility study, 85 of whom were potentially eligible for inclusion in the study. This very closely matched our estimate of approximately 1.5 eligible patients per centre per month,3 which predicted a total eligible population of approximately 80 patients. Initial screening estimates of the incident population fell short of this number. The process evaluation revealed that in some centres screening was being conducted by only a subgroup of the clinicians, and often eligibility assessment and screening were confused. Centres with a stronger research culture and a more embedded research team that was organic to the trauma and orthopaedic department provided more complete screening data.

The distribution of baseline demographic characteristics, including mechanism of injury, sex and degree of cognitive impairment, in the screened population was similar to that observed in the hip fracture population. The assumptions underpinning the development of the protocol were in part based on the more highly investigated hip fracture population.

As expected, there were a small number of patients in whom a pre-existing femoral deformity precluded fixation. Nineteen per cent of the screened population were ineligible owing to the presence of an ipsilateral femoral component of a hip arthroplasty. We are not aware of any existing data that describe the proportion of patients with a distal femoral fracture who have an ipsilateral femoral component. More surprising was the finding that a further 18% were excluded as they were managed non-operatively. Not only is this not the practice described in multicentre audits,3 but there is evidence in favour of operative treatment for fractures of the distal femur. 7

A total of 85 patients were eligible for inclusion in the study, which greatly exceeded our specified definition of feasibility of 1.0 participants per centre per month. The most important reason for missing eligible patients was a lack of willingness among surgeons to enter patients into the trial, accounting for 39 out of the 60 missed patients. The reasons for 15 missed patients could not be determined but it is likely that the majority of these were also lost because of surgeon-related factors. Therefore, between 54% and 65% of all eligible patients were lost because of surgeon-related factors: principally, a lack of confidence with both interventions.

The screening data demonstrated very marked variation in practice across the centres. This was most apparent in the variation in surgical decision-making. For example, in one centre 18 patients were managed non-operatively whereas in another there were no patients who were managed non-operatively. In two centres, > 70% of those missed were because of surgeon preference whereas in another two centres < 40% of those missed were because of surgeon preference.

The process evaluation identified that the interaction at surgeon level between a lack of individual equipoise and an imbalance in their confidence in the use of the two technologies posed a very significant barrier for surgeons to participate fully in a trial investigating what is perceived to be a technically demanding procedure. An integrated qualitative recruitment intervention based on the theme of community equipoise, to support surgeons in their decision-making to recruit participants, could substantially overcome this major barrier to the successful delivery of a definitive trial.

If all the participants who were missed because of lack of surgeon equipoise had in fact been recruited, which is a reasonable assumption as so few patients declined to participate, then the recruitment rate would have risen to 1.1 participants per centre per month, exceeding the prespecified definition of feasibility.

During the recruitment phase of the study, the TSC took the decision to widen the eligibility criteria to all adults in an effort to understand whether or not age as a proxy for fragility was leading to a substantial loss of eligible patients at screening. Thirty-one out of 88 ineligible patients were excluded because of their age. The eligibility change did lead to a change in eligibility rate from 44% to 60%, although this did not translate into an increase in the recruitment rate. Throughout the duration of the study, only two participants < 50 years of age were recruited.

The mechanistic criterion reported in the screening process of a fall from standing height (defined as < 2 m) successfully identified fragility fractures and was the commonest cause of injury. Only 3 out of the 69 eligible participants for whom these data were recorded sustained their fracture from higher-energy injuries. This definition is concordant with the WHO’s definition of fragility fracture and may be an eligibility criterion for a future definitive trial. We suggest eligibility criteria that include both surrogates for fragility.

Recruitment

As expected, recruitment in the lead centre outstripped that in other centres, exceeding twice the rate of recruitment in the best-performing other centres. Approximately half of all the participants were recruited in the lead centre.

One of the centres failed to recruit any participants and four centres recruited only three participants, despite screening larger numbers of eligible patients. Further exploration of the reasons for this in the process evaluation showed that the surgeons’ personal preferences were the principal barrier to recruitment. Recruitment was also hindered by the procedures for identifying and screening patients; our findings suggest that a strong research culture and integrated clinical and research teams within the centres may facilitate the identification of potential patients and recruitment.

We found that gaining confirmation from the treating surgeons that their patient was eligible for inclusion and that they were willing to randomise them was the greatest barrier to delivering this study. As demonstrated throughout the results of our process evaluation, although patients tended to be accepting of the study interventions and procedures and research teams were successful at recruiting eligible patients, surgeons held strong views about the two interventions and their appropriateness for individual patients. When integrating these data, it was apparent that recruitment is enhanced in specialist trauma centres. This is probably because of an interaction of greater confidence with both technologies among specialist trauma surgeons and more fully integrated research teams with a more complete understanding of community equipoise at these sites.

The development of a research culture appears crucial to the success of a study such as TrAFFix. TrAFFix was particularly difficult to implement because of the comparatively uncommon number of injuries seen in any one centre by each treating surgeon. RAs commonly reported that the infrequent presentation of eligible patients challenged their established structures for screening and recruitment. The data from the process evaluation suggest that a supportive and experienced research team may contribute to the success of delivering difficult studies. Experienced RAs may make things easier for surgeons by undertaking activities that limit the amount of research-related activity required and prompt recruitment through attendance in the daily trauma meeting and understanding of the eligibility criteria. Furthermore, in this study, by understanding how best to involve this group of patients in research, RAs enabled the majority of patients who were approached to participate.

Overall, the estimate of recruitment rate for the definitive trial was 0.42 [95% confidence interval (CI) 0.27 to 0.62] participants per centre per month. This estimate falls short of the prespecified definition of feasibility, which was 1.0 participants per centre per month. However, it very closely mirrors the recruitment rate observed in a successfully completed trial of similar technologies used in the tibia. 72,73 It is therefore likely that the recruitment rate observed here could be sufficient to deliver a successful definitive study given modifications in the design of that future trial.

For the future success of a definitive trial, it is important that surgeons are confident to recruit participants. This is facilitated by sharing the existing community equipoise found in this study with individual surgeons who may not have a sense of equipoise. Furthermore, the perceived operative complexity of these fracture patterns is real and recruitment could be substantially improved if an integrated qualitative recruitment intervention was adopted.

Declined to participate

Only five patients declined to participate or withdrew from the study. This suggests that the patient and consultee materials and the consent discussions conducted by the RAs were effective. However, patients often recalled the consent discussions and details of the study incompletely.

Treatment according to allocation

Twenty-one out of the 23 participants received their allocation treatment according to protocol. Although the randomised numbers were small, this is a reassuring proportion and suggests that once allocation had been assigned the majority of participants and surgeons were happy to proceed with the study intervention.

Completion rates

Seventy per cent of the participants were available at the primary outcome time point but not all follow-up questionnaires were complete. Some research staff reported that for this group of patients the length of the CRFs could be burdensome. Research staff stated that the burden of questions at baseline exceeded what could be easily completed during a single consultation and felt this may deter patients from completing the CRFs at follow-up. The majority of the participants were followed up in hospital with clinical and radiographic assessments. This is distinctly different from the hip fracture population, among whom patients are infrequently followed up in outpatient settings. Therefore, collection of face-to-face instruments and radiographs has been demonstrated in this study to be feasible, but the number of the data to be collected should be reduced to a minimum data set.

Baseline characteristics

Generally, baseline data were completed adequately, but research staff commented that measurement of some of the effect modifiers, such as social support, was time-consuming, complex and intrusive. It is likely that in a larger trial a perceived burden of entry into the trial by both researchers and participants could be a barrier to successful delivery. We would therefore recommend reducing the number of additional measurements included at baseline.

Measures of clinical effectiveness

Staff reported that they perceived a duplication between the EQ-5D-5L and DRI, instruments that are designed to measure quality of life and function, respectively; however, participants did not often recall the detail of their contacts with research staff.

Insufficient numbers of participants with chronic cognitive impairment were available in the recruited sample to draw any inferences about the relative merits of DEMQoL versus EQ-5D-5L in determining HRQoL in this group. It is likely that any future trial would not include participants with chronic cognitive impairment owing to the very low rate of operative treatment in this group and, therefore, such considerations would no longer be relevant.

The collection of radiographical outcomes and AEs was uncomplicated with good data completeness. It is not anticipated that the collection of these data need be removed from a definitive trial.

Measures of cost-effectiveness

Approximately half of all the participants provided complete responses to the resource use questionnaires. Researchers reported that they found the length of the questionnaires to be excessive for this group of patients. It is likely that the main drivers of cost within a definitive analysis would reduce the questionnaires to a much smaller subset of questions so that the burden of follow-up could be reduced.

Readmission to hospital for inpatient care and use of social care services in the community dwarfed all of the other costs collected in the study. The majority of the resource use categories, such as pathology services, yielded costs that were negligible in comparison. This information could be used to reduce the size of questionnaires and burden of follow-up and promote better data completeness without compromising the validity of a definitive health economic evaluation.

The estimation of health states using EQ-5D-5L was very successful, with an overall attrition rate of only 30%. The requirement to collect a separate instrument in the cognitively impaired subgroup is likely not to be relevant in any future definitive trial.

Implementation

Some surgeons had strong treatment preferences, which could be considered a barrier to future intervention implementation. Surgeons’ willingness to implement their less preferred intervention in their practice, should it be found more effective in a definitive trial, could be questioned. However, the results of a definitive trial may overcome surgeons’ concerns about the appropriateness of both interventions for these fractures. Furthermore, surgeons have been found to change their practice in response to the results of musculoskeletal trauma trials. 75

Mechanisms of impact

A variety of factors may interact with the interventions to influence patient outcomes after a distal femoral fracture. Most notably, we found that, among our sample of participants, recovery was difficult. In some cases, recovery was inhibited by limited mobility prior to the fracture, a loss of confidence after the fracture and limited access to physiotherapy. Support from relatives or carers enabled some participants to manage at home.

Context

Contextual factors may influence the implementation of the interventions and may affect intervention outcomes. Important factors include the frailty of patients who sustain distal femoral fractures. These patients need support to rehabilitate and to manage after surgery and post discharge.

Logic model

We developed a logic model that summarised the problem posed by fractures of the distal femur. The logic model highlights key factors, in addition to the surgical and rehabilitative interventions, that may have an impact on patients’ HRQoL after distal femoral fractures. These include mediators such as self-efficacy and social support, as well as intermediate outcomes such as being pain free or returning to a pre-fracture residence.

Design

A RCT reporting a two-sided test for superiority with an integrated qualitative recruitment intervention based on the theme of community equipoise. The trial should be made up of two phases: an internal pilot to test recruitment assumptions and a main recruitment study. Clear stop–go criteria should be defined for the pilot phase, with a mean recruitment cut-off point specified at 0.5 participants per centre per month.

Population

We have modified the inclusion criteria to accommodate a clinical definition of fragility fracture. Patients are potentially eligible if they:

• are aged ≥ 60 years or

• are aged ≥ 18 years or have sustained a fragility fracture defined as a fall from standing height and

• have a fracture of the femur involving the distal two ‘Müller’ squares18 that would, in the opinion of the treating surgeon, benefit from internal fixation of the fracture.

We have modified the exclusion criteria to exclude patients with cognitive impairment. This reflects the findings that few such patients undergo operative treatments. Patients will be excluded if they:

• have a loose knee or hip arthroplasty requiring revision

• have a pre-existing femoral deformity

• have an arthroplasty that precludes nail fixation

• are unable to adhere to trial procedures, for example as a result of chronic cognitive impairment.

Interventions

Outcomes

The primary outcome should be a patient-reported functional outcome, such as DRI, assessed at 4 months. Our patient and public representatives at the workshop reported that this instrument had suitable face validity; it is a general measure of lower limb function capturing important global domains of function34 and has performed well in previous HTA studies in lower limb trauma. 72,73 We found that patients with chronic cognitive impairment were under-represented in the recruited sample as surgeons tend not to treat them operatively. Hence, there is no requirement for a proxy-reported instrument in the proposed definitive trial. Secondary outcomes would include HRQoL measured using the EQ-5D-5L, AEs and resource use questionnaires focused on key drivers of cost, such as readmission to hospital and use of social care services.

Sample size

The best available evidence for what constitutes a minimally important difference for DRI comes from a recent large trial. 72 The minimally important difference used in the FixDT trial was 8 points on a 100-point scale. On an individual patient level, a difference of 8 points represents the ability to climb stairs with ‘some difficulty’ versus with ‘great difficulty’. Using an estimate of the SD of 21 points from other studies,72 this suggests a standardised effect size of approximately 0.38, which is a ‘moderate effect’ based on Cohen’s criteria. 12

Assuming that the DRI at 4 months post surgery has an approximate normal distribution (which our data suggest is reasonable), a 1 : 1 allocation ratio and an attrition rate of 30%, then, if the true difference between the experimental and control group EQ-5D-5L means is 8 points, 211 participants would need to be recruited in each group (422 in total) to be able to reject the null hypothesis that the population means are equal with probability (power) 0.9 and type I error rate of 5% (significance).

Recruitment

Given a sample size of 422 and a plausible recruitment rate, based on the findings here, of approximately 0.5 participants per centre per month, a recruitment window of approximately 2.5 years would be required across 28 centres.