Open reduction internal fixation vs non-operative management in proximal humerus fractures: a prospective, randomized controlled trial protocol

Non-operative management of proximal humerus fractures with a period of immobilization and progressive physiotherapy is a simple, noninvasive and readily available treatment option. In a systematic review of non-operative management, Iyengar et al. [9] evaluated 12 studies (n = 650) [10‐21], with a mean age of 65.0 years and a mean follow-up of 3.8 years (range of 1–10 years). Based on the Neer classification [22], there were 49% undisplaced or one-part (n = 317), 25% two-part (n = 165), 21% three-part (n = 137), and 5% four-part (n = 31) fractures. Although variable, all treatment protocols included a period of sling immobilization followed by progressive mobilization as tolerated. The mean rate of radiographic union was 98% (range 93–100%). Various functional outcome scores were used; with 6 studies (n = 272) [10, 12, 14, 19‐21] showing a weighted mean Constant score of 74 (range 55–81) corresponding to a “fair” outcome. Across all studies, a 13% complication rate was reported, with varus malunion being the most common (n = 44 or 7%). Proximal humerus avascular necrosis was found to be uncommon (n = 13 or 2%) [9]. In the largest included trial, Hanson et al. reported the functional outcomes of non-operative management through a prospective evaluation of 160 patients, with 124 patients having complete 1-year follow-up. Nearly half (53.1%) were undisplaced fractures. The average Constant score was 74.3 with a mean difference between the injured and contralateral shoulder of 8.2. They found an estimated median time to definitive union of 14 weeks, and a 7% risk of delayed or nonunion. Four patients went on to require surgical fixation and 5 patients underwent arthroscopic decompression, with an eventual operation rate of 5.6% [12]. With a large focus on undisplaced fractures, these studies highlight that non-operative management of proximal humerus fractures can lead to satisfactory functional outcomes with modest complication rates. In a report of non-operative management of displaced proximal humerus fractures, Yuksel et al. reported a mean Constant score of 61.3 (n = 18, eight 3-part and ten 4-part; mean age of 68.2 years; mean follow-up of 3.3 years), with nonunion and osteonecrosis detected in 27.8% (n = 5) [23].

Displaced proximal humerus fractures are commonly treated with open locking plate fixation [4‐6, 8, 24‐26]. A systematic review of 514 displaced proximal humerus fractures treated with locking plate fixation (12 included studies, average age of 62, average follow-up of 2.4 years) [27‐38] showed an overall healing rate of 96.6%. The review included 34.0% two-part (n = 175), 44.7% three-part (n = 230), and 21.2% four-part (n = 109) fractures. Nine out of the 12 studies (n = 376) [27‐29, 32, 33, 35‐38] reported an average Constant score of 73.6 when evaluating functional outcome. When stratified for fracture classification, the Constant score was significantly less for 4-part fractures in comparison to the 2-part fractures (p = 0.02). The overall complication rate was 48.8% with a reoperation rate of 13.8%. With the exclusion of varus malunion, the complication rate remained high at 32.6% over the 12 studies analyzed [24]. Two other multicenter studies evaluated locking plate fixation for the treatment of displaced proximal humerus fractures reported similar Constant scores of 70.6 and 72 at a minimum of 1 year follow-up, and overall complication rates of 40 and 45% [25, 26]. The Proximal Fracture of the Humerus Evaluation by Randomisation (PROFHER) trial was another multi-centered randomized controlled trial of 250 patients > age 16 which showed no difference between operative and nonoperative management using the Oxford shoulder score and the Short Form 12 (SF-12). The complications in the surgical and nonsurgical group were reported as 24% and 18% respectively [39].

Despite the lack of superiority demonstrated, there has been a significant increase in surgical fixation of proximal humerus fractures following the introduction of locking plate technology over the last decade [40, 41]. Associated with high complication rates, open reduction and internal fixation of isolated proximal humerus fractures in the elderly has also been found to be an independent risk factor for inpatient adverse events and mortality [42].

Previous comparisons of non-operative and locking plate fixation have been conducted with no clear consensus as to the optimal management of osteoporotic proximal humerus fractures. No difference in the quality adjusted life years (QALYs) and societal costs between the two treatment options have been demonstrated [43]. In a randomized control trial evaluating 60 displaced 3-part proximal humerus fractures, Olerud et al. found no statistically significant difference in functional outcomes (Constant, Disability of the Arm Shoulder and Hand (DASH), and health-related quality of life (HRQoL) scores) between non-operative management and locking plate fixation over a 2-year period. One patient (3%) in the non-operative group went on to require surgical intervention and 9 patients (30%) in the locking plate group had a secondary operation with a major complication rate of 13% [44]. Fjalestad and Hole have recently reported on 50 elderly patients with displaced proximal humerus fractures randomized to non-operative or operative management using locking plate fixation. Fracture patterns were categorized based on the AO/OTA classification system, which makes a direct comparison to the study by Olerud et al. difficult. With the Constant score as their primary outcome, Fjalestad and Hole found no significant functional or HRQoL difference over a 2-year follow-up period. Similar to previous studies [25, 26], a 35% overall complication rate with surgical management was reported [45]. Both randomized control trials were limited by a small number of enrolled patients [44, 45].

It is generally accepted that non-operative management is ideal for undisplaced proximal humerus fractures, while displaced four-part fractures can be treated with non-operative management, surgical fixation or arthroplasty options [5, 6]. With no consensus in the literature, the specific management of displaced two and three-part proximal humerus fractures remains highly variable, with non-operative and locking plate surgical fixation the two most common and readily available treatment options. Given the lack of consensus on optimal treatment, conflicting, low quality-of-evidence reports, and higher level of evidence studies beset by various limitations, treatment remains highly controversial. By comparing the functional outcomes of surgical fixation versus non-operative treatment of displaced two, three and four-part proximal humerus fractures in the elderly population, the optimal management strategy for this common injury may be determined. The results of this trial would have the potential to minimize unnecessary complications and provide much needed guidance to orthopedic surgeons striving to maximize patient function and provide quality patient care in an era of rapidly increasing health care costs.

This study is designed as a single-centered prospective, single blind, randomized controlled trial (RCT) with parallel arms comparing nonoperative management with locked plate open reduction and internal fixation (ORIF) of proximal humerus fractures. After enrollment, patients will be randomly allocated to receive either nonoperative or operative fixation of their proximal humerus fractures. This study will abide by the current international research standards and will be reported as per the guidelines in the CONSORT statement [51]. Approval was obtained from the Health Science Network Research Ethics Board of Ottawa and is in compliance with: Ethical Conduct for Research Involving Humans; the International Conference Harmonization Good Clinical Practice: Consolidated Guideline and the provisions of the Personal Health Information Protection Act 2004. All surgical procedures will be performed by fellowship trained shoulder surgeons in a large University-affiliated hospital.

The main purpose of this trial is to determine whether or not a functional difference exists between operative and nonoperative management of low-energy displaced three- and four-part proximal humerus fractures as measured by the Constant, (PROMIS) upper extremity score, the International Physical Activity Questionnaire (IPAQ) for the elderly, and the EuroQol EQ-5D-5 L Health Questionnaire Quality of Life (QoL) functional outcome score [46, 48‐50] over a period of 24 months. We will also aim to determine which method of treatment is associated with a higher incidence of complications as well as the time to union of malunion, nonunion and joint arthrosis. Lastly, we will determine whether initial displacement or quality of surgical reduction has an impact on functional outcome.

We hypothesize that there will be no statistically significant difference in functional outcomes between operative and nonoperative methods. We also hypothesize that there will be a greater incidence of complications in the operative group. Lastly, we hypothesize that fractures with a greater Neer classification and those with a greater degree of initial displacement will have worse functional outcomes.

Patients will be screened in the emergency department of a large University-affiliated hospital and will be enrolled in the fracture clinic. Table 1 lists eligibility criteria for the study. Enrolled patients include males and females > age 60 with acute (< 3 weeks) displaced proximal humerus fractures that fall into the Neer category of 3- or 4- part. Diagnosis will be obtained from radiographs including a true AP (neutral rotation) of the shoulder, a lateral Y-view and an axillary (or trauma axillary) view. The fractures reviewed for inclusion will be independently assessed by two Orthopaedic Surgeons participating in the trial. If there is any disagreement on the classification or inclusion, a third surgeon will be asked to review. The majority consensus will be the implemented inclusion and classification. Eligible patients that have consented to participate in the study will receive informed consent on the two treatment arms prior to randomization. Preoperative baseline functional assessment including the Constant, ASES, EQ-5D-5 L, PROMIS and IPAQ scores will be completed. Patients randomized to the surgical arm will undergo surgery within 7 days of presentation.

The sample size will be 160 patients. The minimum clinically important difference for the Constant score is 12 [52, 53], with standard deviation of 23.1 from measurements in previous studies of similar patient populations. For the primary outcome, to achieve 80% power to detect a clinically meaningful difference of 12 points on the Constant score, with standard deviation of 23.1, and alpha of 0.05, a sample size of approximately 116 people would be necessary (58 per arm). The overall sample size is increased from 116 to 129 to account for an expected 5% crossover from the non-operative study arm to the operative study arm. An additional 31 patients were added for a conservative sample size adjustment accounting for 20% loss-to-follow-up over the two years of follow-up, for a total sample size of 160 patients.

Figure 1 shows the flow of procedures in the trial. Study group allocations will have been pre-determined from an online randomization generator and catalogued in sealed envelopes by personnel independent of the study. The allocation will be in a 1:1 ratio, stratified by fracture type (3 and 4 part): 3- and 4-part fractures will be stratified using permuted blocks of variable length (4 to 6). Randomization and allocation to treatment will be determined on the day of the first assessment in the plaster room clinic within 7 days post injury or at the time of the emergency room visit if the patient is unable to return home. The research coordinator will open the sealed envelope containing the study allocation and will inform the surgeon of the patient’s assigned treatment: non-operative or ORIF treatment. Patients assigned to the ORIF group will have surgery scheduled within 7 days following injury.

The research coordinator will enter all necessary patient information into a password protected electronic database, however due to the nature of the trial design, it is not possible for the surgeon (or the patient) to remain blinded to treatment allocation. The research coordinator will carry out the follow-up assessments and will remain blinded to the patient’s treatment allocation and will not have access to the participant’s chart or radiographs prior to or during the assessment. This will minimize the potential for biases introduced by the examiner when performing the physical assessment and recording data. To help reduce the potential for observer bias, the physical examination and the administration of study questionnaires are standardized. A trained musculoskeletal radiologist (AS) will perform the radiological assessments.

Outcome measures are separated into primary and secondary as outlines below. All functional assessment scores will be completed at baseline and 3, 6, 12 and 24 months post injury. Complications and adverse events will be recorded on standardized case-report forms that will be completed immediately following the completion of the surgical procedure for patients randomized to the surgical treatment arm. At follow-up visits, clinical evaluation will be conducted to monitor for complications including infection, nerve injury, and hematoma formation or hospital re-admission.

Crossovers are expected in a clinical trial of this nature. If crossovers occur, patients will be analyzed in the in the group to which they were initially assigned in keeping with the intention-to-treat principle.

We will take the following measures to aid in completion of follow-up: Patients will normally reside within 90 min travelling time of their surgical center. Patients are called by the research assistant two to four days before their appointment and those who do not attend their appointment will be rescheduled. If the patient misses their re-booking then the surgeon will phone the patient to encourage their attendance. In the event that a patient chooses not to return for a follow-up, the questionnaires are mailed out with a stamped return envelope.

Patient characteristics will be summarized with descriptive statistics. Comparative analyses will be based on the full trial cohort, using the intention-to-treat (ITT) principle (i.e., based on the participant’s randomized allocation). Reasons for missing data will be examined, and appropriate imputation methods will be used to address missing follow-up data, retaining the entire cohort for the ITT analysis. Interim analyses will occur for primary outcome measure when 50% of recruited patients finish their 1-year follow-up. Using the O’Brien-Fleming criteria for sequential tests, the significance level will be 0.005 for the interim analysis and 0.049 for the final analysis. The O’Brien-Fleming boundaries, as well as clinical judgment concerning adverse events, will be used as guidelines for stopping the study early. However, this interim analysis will only occur if randomization is not yet complete.

In addition to the primary and secondary analyses outlined above, several other analyses are planned, though the study is not primarily powered for these statistical analyses. We will perform a subgroup analysis on the above outcome measures to assess whether there is a difference between 3- and 4- part proximal humerus fractures. This will primarily include a subgroup analysis of the Constant score, within groups of fracture type: 3- and 4- part fractures. The second will be a multivariable assessment of the progression of osteoarthritis between pre- and post-operative shoulders. A multivariable regression analysis will also be conducted of possible factors associated with progression of shoulder osteoarthritis, including demographic variables and fracture type. The third will be a multivariable regression analysis to determine which factors may be associated with functional outcome: with the Constant score as the dependent variable, demographic and radiographic factors (including osteoporosis) will be analyzed to determine if prognostic factors exist that may assist in planning treatment.