Explanation for the choice of comparators

This trial will compare closed reduction prior to surgery in displaced distal radial fractures to no closed reduction prior to surgery. Currently, the Dutch guidelines state there is a knowledge gap concerning the efficacy of closed reduction prior to surgery. Therefore, either outcome of this study would lead to an update of said guideline. In daily practice, both treatments are used at the moment of writing this protocol.

Intervention description

Closed reduction of distal radial fractures is performed with axial traction on the wrist with Chinese finger traps connected to weights, with manipulation by the caretaker of a combination of both. In most cases, a local anesthetic will be administered in the form of hematoma block with lidocaine between fracture fragments [7].

Criteria for discontinuing or modifying allocated interventions

If patients allocated to the no closed reduction group develop neurovascular symptoms, the attending physician can opt to perform closed reduction regardless to relieve said symptoms.

Strategies to improve adherence to interventions

Intervention will or will not be performed in the emergency department, so no additional strategies are required.

Relevant concomitant care permitted or prohibited during the trial

Apart from the closed reduction, patients will receive standard care and have no trial specific limitations.

Provisions for post-trial care

Standard test subjects’ insurance has been taken out for all participants.

Outcomes

The primary outcome is the average of the pre-operative visual analog scale for pain (VAS) score, reported on a daily basis from the ED visit until surgery.

Secondary outcomes are wrist function measured with the Patient-Rated Wrist Evaluation (PRWE) score after 6 weeks and 3, 6, and 12 months, length of stay in the ED, type and quantity of used pain medication, patient satisfaction, quality of life, and complications. Furthermore, a cost-effectiveness analysis will be performed using the Medical Consumption Questionnaire (iMCQ) and Productivity Cost Questionnaire (iPCQ) [26, 27], and the ability to assess CT scans of unreduced fractures and reduced fractures will be compared.

Participant timeline

The participant timeline is shown in Table 1.

Table 1 Overview questionnairesSample size

The sample size calculation is based on our primary outcome parameter: the VAS score for pain reported on a daily basis from the visit to the ED until the day of surgery. To find out if the pre-operative pain scores of patients that are refrained from CR are comparable with the scores of the CR group, we will use a non-inferiority test. This requires a non-inferiority margin, which is the biggest difference between the two groups in favor of the CR group, without a statistical difference [28]. The European Medicines Agency advices to base the non-inferiority margin on a difference that is not clinically important [29]. Therefore, our non-inferiority margin is 50% of the minimal clinical important difference (MCID) of the VAS [28]. Based on a systematic review concerning the MCID of VAS scores, a MCID is a context-specific and methodological dependent value [30]. To the best of our knowledge, the MCID of the VAS score for DRFs is unknown. Therefore, we used the MCID of a study that included trauma patients with isolated acute extremity pain [31]. For the description of the MCID of the VAS score, the VAS scale is seen as a 100 mm line. With an MCID of 19.3 mm, our non-inferiority margin will be 9.7 mm (= 50%). As only Bird et al. provide a standard deviation (SD) around the MCID, we will use that value of 15 mm for our sample size calculation [24]. To correct for a cluster-effect we use an intra-cluster correlation (ICC) coefficient between the different hospitals of 0.06, which is generally reported in the literature for hospital processes. To calculate the required sample size, two simulations were run both assuming a fixed number of six clusters (hospitals). In the first simulation, an equal number of subjects per hospital was assumed. Hospitals were randomized to start with treatment A (3 centers) or B (3 centers) in a 1:1 fashion and crossed over after half of the patients were included. For total sample sizes varying from 12 to 502 with increments of 12, 5000 simulations were run for each sample size considered. In each simulation, random data was generated under the specifications described above. A linear mixed model was fitted with a random intercept for cluster and treatment group as fixed effect. A 95% confidence interval was calculated based on the t-distribution with degrees of freedom estimated by the Satterthwaite method. The power was calculated as the percentage of simulations in which the upper limit of the 95% confidence interval was smaller than the non-inferiority margin. From the first simulation, it followed that, using a two-sided alpha of 5%, n = 80 patients total would be required in total to demonstrate non-inferiority with a power of 80% (Fig. 1).

Fig. 1

Power curve for simulation 1 assuming of balanced clusters (equal number of patients included per center)

However, in practice, unequal number of patients is expected to be included by each center. Also, seasonal effects should be accounted for in the analysis. Therefore, a second simulation was conducted in a similar way. The number of patients included per center was (roughly) based on the number of eligible patients in the past year (known for 4 hospitals) divided by 2 (assuming that 50% of eligible patients would participate). Inclusion rates for the two hospitals for which this was not known were set equal to the hospital with lowest inclusion rate, resulting in monthly inclusions of 13, 7, 7, 1, 1, 1. Seasonal effects were included by a monthly effect, which were assumed to follow a sine function with a period of 12 months. The SD of the seasonal effects over a 12-month period was set equal to the between cluster SD. The half period (x = π) was set equal to the moment of cross-over. For sample sizes varying from 60 (two months of inclusion) to 360 (12 months of inclusion), again 5000 simulations were run for each sample size considered. Linear mixed models were fitted including center as random effect and month and treatment as fixed effects. Using a two-sided alpha of 5%, n = 120 patients total (rounded to above) would be required in total to demonstrate non-inferiority with a power of 80% (Fig. 2). Accounting for a 10% loss to follow-up, a total of n = 134 patients will be included. It should be noted that in the simulations, it was assumed that only one VAS pain will be collected for each patient, while in reality multiple VAS scores (one per day) will be obtained during the period between the visit to the emergency department and the operation. An average VAS pain for each patient will be estimated by a multilevel linear mixed model using random intercept for patient nested within center. Including this aspect in the simulation studies was deemed too complicated, since no reliable estimates for the correlation between daily VAS pain scores in this setting are available. However, following the reasoning that multiple measures will lead to a more precise estimate than a single measure, it can be expected that in practice a higher power will be achieved compared to the simulations and including this is not strictly necessary (i.e., the simulations can be regarded as conservative in this respect).

Fig. 2

Power curve for simulation 2 assuming unbalanced clusters (unequal number of patients included per center) and adjustment for seasonal effects

In the Maasstad Hospital, approximately 590 adult patients are diagnosed with a DRF per year. According to Brogren and colleagues, approximately 350 (60%) will be displaced [5].

However, not all of these patients will have CR followed by surgical treatment. Assuming that 60% of all patients with a displaced DRF will meet the inclusion criteria, combined with a rejection rate of 30%, we estimate to require 13 months for inclusion in case of a single-center design. Assuming that the average amount of inclusions per month in the other participating centers equals the amount in the Maasstad Hospital, we can divide the required time by the amount of centers. With a total of six participating centers, we estimate to require 3 months to complete the inclusion. Combined with a follow-up time of 12 months and 3-month margin, the total required time for the clinical part of the trial will be 18 months.

Recruitment

Eligible patients at the ED, fit to the inclusion and exclusion criteria, will be informed about the study and invited to participate by the treating physician. The patient will also receive written information on the study and will have the opportunity to ask questions on before making the decision to participate or not. However, this needs to be done in a short period of time (approximately 30 min) because of the acute problem presentation and setting. If the patient fits the inclusion and exclusion criteria and signs the informed consent form, baseline information will be recorded.