A Combined Randomised and Observational Study of Surgery for Fractures In the distal Radius in the Elderly (CROSSFIRE): a statistical analysis plan

Fractures of the distal radius are the most common fractures presenting to emergency departments and orthopaedic surgeons [1]. These fractures are more common in older people (due to osteoporosis and increased risk of falls) and the incidence in older people is increasing [1]. Considerable practice variation exists in the management of distal radius fractures in the elderly in Australia [2], with two common methods being closed reduction (manipulation of the arm to realign the fracture) with cast immobilisation, and open reduction (surgical exposure and direct realignment of the fracture) and fixation with plate and screw. Open reduction and (volar locking) plate fixation has become the most common form of surgical treatment provided [3]. While there is evidence showing no significant advantage for some forms of surgical fixation over closed treatment, and no difference between different surgical techniques [4‐16], there is a lack of evidence comparing the two most common treatments used in Australia, initial treatment with volar locking plate fixation versus closed reduction and cast immobilisation. Surgical management of these fractures involves significant costs (implant costs, medical costs, hospital costs) and risks (infection, implant failure, general surgical risks) compared to non-surgical management (closed reduction and cast immobilisation in the emergency department). Therefore, high-level evidence comparing the current treatment alternatives (plate fixation versus casting) is required in order to address practice variation, justify or avoid costs, and to provide the best clinical outcome for patients with these common fractures.

The CROSSFIRE study protocol was published in 2017 [17]. This statistical analysis plan (SAP) was prepared in accordance with published guidelines [18] on the content of SAPs and was written with input from members of The CROSSFIRE Study Group.

The primary and secondary aims of the study are to determine whether surgical treatment (volar-locking-plate fixation) is superior to non-surgical treatment (closed reduction and cast immobilisation) with respect to both effectiveness and safety for adults aged 60 years and older with dorsally displaced distal radius fractures.

The results of the trial will be published in a peer-reviewed journal and will be disseminated via various forms of media. The results will be incorporated in clinical practice guidelines produced by professional bodies.

CROSSFIRE is a prospective, multi-centre study with 19 sites in Australia and New Zealand. The study includes a randomised, comparative-effectiveness trial and a parallel observational study; eligible patients were invited to participate in the randomised trial in which their treatment was randomised to surgical or non-surgical treatment. Eligible patients who declined to participate in the randomised study were invited to join the observational study. Participants in the observational study received standard care (one of the two treatment arms in the randomised trial) according to patient and surgeon preference and had the same outcomes measured at the same timepoints as those participating in the randomised trial.

We included an observational ‘preference’ arm that followed non-randomised patients in a similar manner to randomised participants to investigate potential selection bias and provide information on the generalisability of the study [19]. This study design has been used in surgical trials [20] and has been recommended as a model for trials of surgery versus non-surgical treatment where recruitment rates are expected to be lower than for other randomised controlled trials (RCTs) [21].

Three hundred participants provided consent and were recruited to CROSSFIRE from 19 participating hospital sites: 164 participants in the randomised trial and 136 in the observational study. No patients declined participation in the observational study.

Surgical fixation using a volar locking plate was performed within 2 weeks of the initial injury according to the usual protocol of the participating institution, with an orthopaedic surgeon in attendance. This is a commonly performed procedure. Surgical technique and type of plate (make and length) was as per surgeon preference. A plaster cast was applied post-operatively but for no longer than 2 weeks. Active finger movement was encouraged post-operatively. Participants were reviewed 2 weeks (10–17 days) after surgery; the wound was reviewed and sutures removed where necessary. Participants were provided with a post-operative home-exercise programme (printed information). Referral for outpatient rehabilitation was not routinely provided but was permitted.

Participants in this group were treated with a closed-reduction and cast immobilisation, avoiding wrist flexion, within 2 weeks of the initial injury. This method of casting is consistent with standard casting practice in Australia. Immobilisation of a distal radius fracture in flexion has been associated with an increased risk of fracture displacement as well as finger and metacarpophalangeal joint stiffness [23]. Also, immobilisation in a cast that is too restrictive and excessively flexed has been associated with an increased risk of complex regional pain syndrome (CRPS) [24, 25]. The reduction was performed in the emergency department under sedation and local anaesthetic infiltration into the fracture (haematoma block) where possible or in an operating room (if not possible in the emergency department). The procedure was performed by an orthopaedic surgeon or registrar. Post-reduction radiographs were taken to assess the fracture alignment. The best reduction achievable was accepted. The cast was removed at 6 (± − 1) weeks from the initial reduction. Active finger movement and light use of the hand was encouraged immediately. Participants were provided with a home-exercise programme (written information). Referral for outpatient rehabilitation was not routinely provided but was permitted.

The Consumer Advisory Group of ANZMUSC (the Australian and New Zealand Musculoskeletal Clinical Trials Network) reviewed the protocol and the study was endorsed by ANZMUSC. Separately, three elderly patients with wrist fracture (who were not study participants) were interviewed and provided feedback on what post-treatment information was most relevant and important to older patients with wrist fracture.

Randomisation occurred immediately after consent was gained by the recruiting orthopaedic team, within 1 week of the date of the injury. The orthopaedic team member contacted a central computer-based randomisation service by telephone. Participants were randomised using the method of minimisation. Randomisation was stratified by site, and minimisation, adjusting for gender and age (60–74 years and > 74 years), was employed as recommended by the National Health and Medical Research Centre (NHMRC) Clinical Trials Centre which provided the randomisation service.

Given the nature of the comparisons (surgery versus no surgery), it was not possible to blind the surgeon (study) investigators or participants. While this increases the risk of performance and detection bias, every effort was made to ensure that treatment, other than the interventions under study was otherwise identical in both groups. The primary outcome (PRWE score at 12 months) was collected from participants by blinded researchers, by telephone.

The statistician and the investigators conducting the analysis will remain blinded to the treatment groups. The dataset will be de-identified and treatment allocation will be masked using dummy group names (for example, Group A and Group B). De-identification and masking will be performed by an independent investigator who is not associated with the analysis. Masking of treatment allocation will be maintained until the statistical analysis and interpretation has been completed and two versions of the manuscript have been prepared and agreed to by all authors.

Outcomes are collected at four timepoints; baseline, 3 months, 12 months and 24 months. The timepoints and corresponding outcomes are listed in Table 1. Baseline data is collected by the recruiting clinicians following consent. Baseline data includes age, gender, fracture type (AO/OTA 23A OR 23C), radiographic features of fracture, fracture-healing risk factors (see co-morbidities in Tables 2 and 3) and treatment preference. Paper-based data collection is forwarded to the study coordinator for direct electronic data entry into a central electronic database – REDCap™ [26].

Participants are followed up at 3 months post initial procedure (± 1 week), 12 months post initial procedure (± 1 month) and 24 months post initial procedure (± 1 month). There is an intention to collect outcomes at 5 and 10 years post treatment.

Participant follow-up at 3-month, 12-month and 24-month timeframes is conducted by telephone. The outcomes for the randomised arm of the study are collected by a blinded investigator. On rare occasions where data collection by telephone is not practical or possible (hearing impairment, English language proficiency), paper questionnaires are sent by post or by email to the participant to be completed and sent back to the research institution by reply-paid post. Completed telephone and postal questionnaires are entered into the password-protected central electronic database and paper-based records are collected and securely stored within the administering institution.

The primary outcome is the patient-rated wrist evaluation (PRWE) questionnaire at 12 (± 1) months post injury. The PRWE is a 15-item patient-reported measure of pain and function, specific to the wrist. It is a continuous score on a scale from 0 to 100 with higher scores indicating poorer outcomes [27]. It is commonly used, was developed with patient input and has been validated for use in patients with distal radius fractures [28].

Secondary outcomes include:

Adverse events are defined as symptomatic fracture non-union (three of four cortices not united radiographically at a minimum of 6 months), infection (local infection requiring any treatment), neuropathy, tendon irritation (requiring treatment), tendon rupture or complex regional pain syndrome (diagnosed according to the International Association for the Study of Pain (IASP) clinical diagnosis criteria [31]).

The primary hypothesis is that patients aged 60 years and older with displaced fractures of the distal radius managed surgically using volar-locking-plate fixation will have a clinically important superior patient-rated pain and function (PRWE score) at 12 months post injury compared to those managed non-surgically with closed reduction and plaster casting. A difference of greater than 14 points will be considered to be clinically important [32].

The secondary hypothesis is that patients aged 60 years and older with displaced fractures of the distal radius managed surgically using volar-locking-plate fixation compared to those managed non-surgically with closed reduction and plaster casting, will have:

We consider 14 points on PRWE to be the minimum clinical difference necessary to justify the additional costs of surgery compared to non-surgical treatment. This is the MCID determined by Sorensen [32] and larger than the 11.5-point estimate determined by Walenkamp [38].

A total of 128 patients (64 in each group) will provide 90% power to detect a difference of 14 points on the PRWE scale at a significance level of 0.05. We aim to recruit 160 patients to allow for 20% loss to follow-up. Two previous RCTs that had each published results at the time of our sample size calculation reported loss-to-follow-up rates of 19% [5, 39]. Given that this is a multi-centre trial, the closing date for recruitment (31 December 2018) was set based on expected recruitment of 160 randomised participants. However, by the time recruitment closed, 164 participants had been recruited.

The observational arm includes all eligible patients not consenting to randomisation. We recruited 136 participants to the observational arm to the trial.

In the analysis, missing data will be dealt with according to the instructions on the use of the outcome tools (PRWE, DASH and EQ-5D-5L). The need to impute will be assessed and confirmed at the time of review of blind data. If greater than 10% of the primary outcome data is missing from the randomised sample, then missing primary outcome data will be imputed using multiple imputations based on baseline data including age, gender and fracture type as well as PRWE score at 3 months. A completed case analysis will be performed as a sensitivity analysis.

The flow of participants through the study will be displayed according to Consolidated Standards of Reporting Trials (CONSORT) (see Fig. 1 for example), including numbers of participants for enrolment, allocation, follow-up and analysis.

Participant characteristics will be displayed according to their baseline characteristics for each group. Characteristics will be displayed for both the randomised and the observational arms of the study and displayed for ease of comparison (see Tables 2 and 3 for examples).

The primary analysis will follow the intention-to-treat (ITT) principle. The primary outcome is the PRWE score at 12 months for the participants recruited to the randomised trial. The MD in PRWE scores between the surgical and non-surgical groups will be tested using a two-sample t test. The mean values, SDs, MDs and 95% CIs will be displayed (see Table 6 for example). The medians and inter-quartile ranges will be represented in a box plot (see Fig. 2 for example).

Outcomes for the randomised group and the observational group will be analysed and represented separately. Outcomes gathered at 3 months will be displayed for the randomised group (see Table 4) and for the observation group (see Table 5). Outcomes gathered at 12 months will be displayed for the randomised group (see Table 6) and for the observational group (see Table 7). Continuous outcomes will be reported by mean values as well as the MDs and 95% CIs. Categorical outcomes will be reported by frequency, incidence and odds ratios.

The MDs of PRWE measured at 3 months between the surgical and non-surgical groups will be tested using two-sample t tests. The MDs of DASH measured at 12 months between the surgical and non-surgical groups will be tested using two-sample t tests.

Physical therapy utilisation up to and at 3 months will be analysed. If surgical treatment is shown to be superior, a cost-effectiveness analysis will be conducted incorporating therapy utilisation and published separately to this analysis. For the purpose of this analysis, therapy utilisation up to 3 months (Yes/No) and ongoing therapy utilisation at 3 months (Yes/No) will each be reported in terms of frequency, incidence and odds ratio between each treatment group. See Tables 3 and 4 as examples.

Complications will be reported in terms of frequency, incidence and the odds ratio between each treatment group (see Tables 8 and 9 as examples). Odds ratio will be calculated based on the number of participants for whom data is available.

Repeated measures analysis will be used for the following:

The observational group will be used to investigate potential selection bias. Baseline characteristics will be compared between the observational group and the randomised group (see Table 10 for example).

The 2-, 5- and 10-year outcomes will be analysed separately to this analysis and will be published at a later date.

In order to investigate the impact of missing data, a completed case analysis of the primary and secondary outcomes at 12 months will be performed and compared with the analysis using imputed data (if required); see Table 11 for example.

In order to investigate the impact of non-compliance with the protocol, an as-treated analysis and a per-protocol analysis of the primary and secondary outcomes at 12 months will be performed and the results will be compared with the ITT analysis; see Tables 12 and 13 as examples. Non-operative treatment will be defined as a minimum of 28 days in the plaster splint for the purposes of the per-protocol analysis.