Introduction

Radial head fractures occur in approximately 4 % of all fractures and 33 % of elbow fractures [1]. A retrospective epidemiological study from the Netherlands noted that the incidence of radial head fractures was 2.8 per 10,000 inhabitants per year [2]. The modified Mason classification system is commonly used to describe radial head fractures and guide the clinical treatment. The types of Mason I and II fractures are treated with nonoperative therapy or by ORIF [3, 4]. The modified Mason type III radial head fractures are comminuted fracture, and the modified Mason type IV are radial head fractures with dislocation. These two kinds of fractures are difficult to treat, and the prognosis is unsatisfactory [5].

Proper treatment is essential in order to regain the stability and functional activities of the elbow and minimize the sequelae. The treatment included traditional resection, ORIF and RHA. Resection is an option in Mason type III without valgus instability, but some studies showed resection had less satisfactory joint motion, strength and function than ORIF and RHA, so it was not recommended [6].

Another treatment ORIF is utilized to perform operations of the displaced, noncomminuted fractures and part of the complex, unstable fracture dislocations. The goals of ORIF treatment are to obtain a anatomic reduction and stable fixation [7]. Bone nonunion and internal fixation failure are the most common complications which can lead to chronic pain and dysfunction of the elbow joint [8]. While RHA is made use of managing comminuted displaced fractures or other failed treatments [9]. The main complications are implant loosening, overstuffed, etc. [5]. Each treatment has its own advantages and disadvantages. Therefore, management of radial head fractures is still a matter of debate.

The clinical outcomes of ORIF and RHA have been compared in a few studies. Although a meta-analysis was earlier performed by Li [10], there are some new studies in recent years. Therefore, we performed a meta-analysis to evaluate the clinical outcomes and complications of ORIF and RHA in patients with radial head fractures (modified Mason type III or IV).

Materials and methods

Our meta-analysis was conducted following the guidelines provided by the Cochrane Handbook and according to the PRISMA guideline.

Study selection

Studies were included if they met the following criteria: (1) randomized controlled trials (RCT) or nonrandomized controlled trials (nRCT) comparing ORIF versus RHA for the treatment of radial head fractures, and (2) studies reporting at least one of the following outcomes: elbow functional evaluation criteria by Broberg and Morrey, elbow score (Broberg and Morrey), MEPS, QuickDASH score, VAS, range of motion (ROM), operation time, etc. The exclusion criteria were reviews, case series and cadaver studies.

Literature search

A computerized search of five electronic databases (OVID, PubMed, EMBASE, Cochrane Central Register of Controlled Trials and MANTIS) was conducted from the inception to July 2015. Search terms were selected in order to maximize the search sensitivity and specificity. The key words of searching included “radial head fractures,” “terrible triad injuries,” “radial head reconstruction,” “radial head dislocation,” “radial head resection,” “open reduction and internal fixation/radial head fractures,” “radial head replacement,” “radial head arthroplasty” and radial head prosthesis.” Unpublished findings were not considered in this study. Two reviewers independently assessed the publications for inclusion in this meta-analysis. Any discrepancies were resolved through discussion between the reviewers.

Data abstraction

Two independent reviewers extracted trial details and collated onto a predefined database. From each study, the data extracted included sample size, patient age, gender, sides of operation, fixation, follow-up period, clinical outcomes and complications.

Quality assessed

In order to adopt the same method to evaluate the RCTs and nRCTs, two reviewers independently applied the “assessing risk of bias” table to assess the risk of bias of the studies included. The evaluation also included allocation concealment, participant, personnel, outcome assessor blinding, attrition bias, incomplete outcome data, selective outcome reporting and other sources of bias. Discrepancies were resolved through discussion between the reviewers.

Data analysis

Continuous data were assessed using mean difference (MD) with corresponding 95 % confidence interval (95 % CI). Dichotomous data were assessed using relative risk (RR) and 95 % CI. p < 0.05 was regarded as statistically significant. If a standard error was seen in the final outcome value, we calculated the z-score with the formula: z = mean difference/standard error [11].

The heterogeneity of the studies was assessed by the Cochrane Q test. For statistical heterogeneity with a χ 2 heterogeneity of p > 0.10 and the I 2 statistic <50 %, a fixed-effects model was adopted; otherwise, a random-effects model was chosen. The review manager 5.3 was used to perform this analysis.

Additional analyses were performed according to different age (the younger or the elder), study design (RCT or nRCT) and different modified Mason classification (III or IV).

Results

Identification and selection of studies

A total number of 2184 records and abstracts were extracted from five electronic databases, of which 1439 duplicates records were excluded. Ten studies comparing ORIF with RHA were identified, while two of them were excluded because they were biomechanical or cadaver studies (Fig. 1). Finally, two RCTs [13, 19] and six nRCTs [12, 14–18] were included in this review (Table 1).

Fig. 1
figure 1

Flow diagram of included and excluded studies

Full size image
Table 1 Description of the studies included in the meta-analysis
Full size table

Study characteristics and quality

There were 138 ORIF cases and 181 RHA cases in eight studies. Five studies [12, 14, 17–19] reported the gender of the patients. A total number of 48 males and 42 females underwent ORIF as compared to 59 males and 62 females who underwent RHA. Two studies [14, 17] reported the side of operation, which included 18 left elbows and 14 right elbows for ORIF as compared to 17 left elbows and 11 right elbows for RHA. The internal fixations of ORIF were plate and screws but one [12], and the RHA adopts the prosthesis of radial head. The maximum follow-up period was >3 years [14], and the minimum period was approximately 1 year [12, 18], Table 1. All the studies used the Mason classification system or modified Mason classification system to evaluate the severity of the radial head fractures. Four studies [12–14, 18] used the Mason classification, two studies [15, 17] used the modified Mason classification while the other two [16, 19] used both. According to the modified Mason classification, patients were classified again. Four studies [12, 13, 17, 18] were classified as Mason type III totally, and the other four studies [14–16, 19] were classified as Mason type IV totally.

Quality appraisal

The “assessing risk of bias” was applied to assess the risk of bias of the eight studies. All the studies had one or more limitations in study design (Fig. 2). The methodological quality of the eligible trials was moderate to low. Two RCTs had higher quality than six nRCTs (Fig. 2).

Fig. 2
figure 2

Risk of bias assessment of included studies

Full size image

Primary outcomes

Four kinds of overall evaluation were adopted by eight studies. The elbow functional evaluation criteria (Broberg and Morrey) were assessed in four studies [12, 13, 18, 19]. One hundred and seventy-eight elbows (85 ORIF and 93 RHA, respectively) were available. The outcome of satisfaction rate was 94.6 % in RHA group and 72.9 % in ORIF group, which showed no significantly difference (RR = 0.72; 95 % CI 0.44, 1.18, p = 0.20). Heterogeneity was high (I 2 = 89 %, χ 2 = 28.01, p < 0.001), as shown in Fig. 3.

Fig. 3
figure 3

Forest plot. Outcome: elbow functional evaluation criteria (Broberg and Morrey)

Full size image

For QuickDASH score, 99 elbows were evaluated in three studies [14, 15, 17] (41 ORIF and 58 RHA, respectively). No significant difference was observed between the two treatments (MD = 0.73; 95 % CI −1.50, 2.95, p = 0.52). Meta-analysis showed a high heterogeneity (I 2 = 77 %, χ 2 = 8.68, p = 0.52).

MEPS was evaluated in two studies [16, 19], with 78 elbows (31 in ORIF and 47 in RHA). A significantly higher score in RHA group was shown in the meta-analysis (MD = −7.08; 95 % CI −12.93, −1.24, p = 0.02). The heterogeneity was low (I 2 = 0 %, χ 2 = 0.16, p = 0.69).

The elbow score (Broberg and Morrey) was assessed in two studies [13, 18]. One hundred and seventeen elbows (58 ORIF and 59 RHA, respectively) were available for meta-analysis. There were statistically significant differences between two groups (MD = −15.53; 95 % CI −23.16, −7.91, p < 0.001). The heterogeneity was high (I 2 = 93 %, χ 2 = 13.86, p < 0.001).

Secondary outcomes

There was a statistically significant difference in operation time and VAS. The ORIF group showed significantly longer operation time (MD = 33.78; 95 % CI 26.59, 40.97, p < 0.001) and higher VAS than RHA group (MD = −0.58; 95 % CI −0.66, −0.49, p < 0.001) (Table 2).

Table 2 Results of the secondary outcomes
Full size table

For range of motion, the degree of flexion (MD = −1.94; 95 % CI 1.31, 2.58, p < 0.001), pronation (MD = −2.63; 95 % CI −3.17, −2.08, p < 0.001) and supination (MD = 1.43; 95 % CI 0.78, 2.08, p < 0.001) showed statistically significant differences between two groups. There was no statistically significant difference in the degree of extension (MD = 1.52; 95 % CI −2.53, 5.58, p = 0.46) (Table 2).

Complications

There was significantly higher incidence of bone nonunion or absorption (RR = 6.70; 95 % CI 1.59, 28.16, p < 0.001) and internal fixation failure (RR = 7.03; 95 % CI 2.51, 25.01, p < 0.001) in the ORIF groups as compared to the RHA groups (Fig. 4). Other complications were reported in different studies, with each study reporting different complications. There were no significant differences in infection, overstuffed, heterotopic ossification, stiffness, nerve palsy, arthrosis and revision, as shown in Table 2 and Fig. 5.

Fig. 4
figure 4

Forest plot. Outcome: bone nonunion or absorption

Full size image
Fig. 5
figure 5

Forest plot. Outcome: revision

Full size image

Risk of bias and publication bias assessment

Two RCTs and six nRCTs were included in this study. Outcomes were analyzed both separately and together when being reported in RCTs and nRCTs. There were statistical differences in the elbow functional evaluation criteria (Broberg and Morrey) in the RCTs, but not in the nRCTs. In contrast, there were statistical differences in the incidence of internal fixation failure in the nRCTs, but not in the RCTs. Other outcomes were coincidental in the RCTs and nRCTs (Table 3).

Table 3 Additional analyses
Full size table

The average age of one study [18] was more than 65 years old, and the others were less than 51 years of age (Table 1). So we divided the studies into two groups according to age. The average age of the elder group was more than 65 years old, and the younger group was less than 51 years old. All outcomes were coincidental between two groups (Table 3).

According to the modified Mason classification, four studies [12, 13, 17, 18] were classified as Mason type III totally, and other four studies [14–16, 19] were classified as Mason type IV totally. So we divided the studies into two groups according to modified Mason classification (Mason III group and Mason IV group). There were statistical differences in the Elbow extension and flexion in the RCTs, but not in the nRCTs. In contrast, there were statistical differences in Mason III group, but not in Mason IV group. Other outcomes were coincidental between two groups (Table 3).

Discussion

This is a quantitative comparative meta-analysis comparing ORIF and RHA for the treatment of radial head fractures. Ultimately, two RCTs and six nRCTs were included in this meta-analysis.

In overall evaluation, the elbow score (Broberg and Morrey) [13, 18] and MEPS [16, 19] of RHA groups was significantly higher than ORIF group. The elbow functional evaluation criteria (Broberg and Morrey) were considered to be satisfactory if the result was good or excellent and unsatisfactory if it was fair or poor. The satisfaction rate of RHA groups was higher than ORIF groups, though there was no statistically significant difference between them. The elbow functional evaluation criteria (Broberg and Morrey) are a widespread used overall evaluation for elbow, which consists of the motion, strength, stability and pain. Recent studies [12, 13, 18] suggested that the better outcomes of medium-short-term in RHA groups could be explained that radial head prosthesis might provide early stability, allowing early rehabilitation and avoiding the failure of internal fixation.

In complications, five studies reported the incidence of revision. The outcome of revision rate was 16.7 % in RHA group and 20.1 % in ORIF group, which showed no significant difference. Internal fixation failure (9/15) and bone nonunion (4/15) were the main reasons for revision in ORIF. Six studies [12–16, 19] reported the incidence of internal fixation failure. The ORIF groups (15.5 %) had significantly higher incidence than RHA groups (0 %). Five studies [12–16] reported the incidence of bone nonunion or absorption. The ORIF groups (18.5 %) had significantly higher incidence than RHA groups (0 %). In our meta-analysis, all the studies are fixed with plate and screws but one [12]. However, there were still some fragments which were unable to be stably fixed and lack of blood supply due to the primary injury and iatrogenic factors. It gave rise to an increase in the incidence of internal fixation failure. Stable fixation and the protection of the blood supply of the fractures were needed, but it was difficult for severe comminuted fractures [15, 17, 18].

Overstuffed of prosthesis (8/18) and subluxation (4/18) were the main reasons for revision in RHA. Overstuffed of prosthesis was defined by prosthesis over lengthening ≥2 mm [20]. This highlights the need for proper diameter and length of the prosthesis. While the excised radial head can be used for diameter determination, length determination should be made using prostheses models repeatedly during the operation to determine the position for the radial head amputation [21]. A retrospective study [22] about RHA showed that there were 11 cases of overstuffed in 47 cases with revision, but the most common indication for revision was painful loosening (31/47). The subluxation was the secondary important reason of revision in the studies included in our meta-analysis. Watters et al. [15] suggested that lateral ulnar collateral ligament (LUCL) needed to be repaired in both ORIF and RHA, and then, the hanging arm test was performed to assess stability of elbow joint. If instability persisted, repairing of the medial collateral ligament (MCL) or external fixator was necessary. A retrospective study [23] of 258 patients with RHA for more than 10 years of follow-up showed that 62 patients needed revision, and the primary reason for revision was heterotopic ossification (53.2 %). Duckworth et al. [1] reported lower age was increased risk of revision in RHA. Four studies [12, 13, 16, 19] reported the types of prosthesis, with one bipolar and three monopolar. A cadaveric test showed a similar biomechanics outcomes between bipolar prosthesis and monopolar prosthesis [24].

Two studies [17, 19] reported the operation time, and meta-analysis showed ORIF group had longer operation time. The reduction and fixation in ORIF were more difficult than RHA with radial head resection and prosthesis installation, so loner operation time was needed. Five studies [14–16, 18, 19] reported the range of motion. Although there were significant differences, the clinical significance was not obvious. The maximum value of mean difference was less than three degrees. Three studies [13, 17, 19] reported the incidence of stiffness. The incidence of stiffness in ORIF group (8.2 %) was higher than that in RHA group (3.4 %) but no statistically significant differences. Finally, only one study [16] reported the iatrogenic nerve injury. The incidence of RHA groups (20.0 %) was higher than ORIF groups (8.3 %) but no statistically significant differences.

However, there were three limitations in this meta-analysis. First, two RCTs and six nRCTs were included. Therefore, the nRCTs had a greater caseload and were more likely to suffer from various biases. The additional analysis showed different results between the RCTs and nRCTs. Secondly, all the studies used the Mason classification system or modified Mason classification system to evaluate the severity of the radial head fractures. However, a marginal fracture with displacement >2 mm and dislocation was classified as modified Mason type IV but Mason Type II. Although all the cases can be classified as modified Mason type III and IV, the bias persisted. Thirdly, in terms of follow-up period, none was more than 5 years in this study. Although the longest follow-up period was up to three and a half years, there were a few complications which required longer period of follow-up to obtain accurate results. Therefore, the outcomes in this study might exist inaccurate.

In conclusions, radial head arthroplasty afforded higher satisfaction rate, better elbow score (Broberg and Morrey) and MEPS, shorter operation time, lower incidence of bone nonunion or absorption and internal fixation failure when compared to open reduction and internal fixation. As the follow-up period of all the studies in this meta-analysis were medium-short-term, longer-term studies will be required to ascertain whether the apparent benefits of RHA will be offset by late complications.