Does switching to anatomically pre-contoured clavicle plates lead to fewer failures and removals than reconstruction plates? - a propensity score matched retrospective cohort study of 114 patients


 Background: Plate fixation is frequently used to treat displaced midshaft clavicular fractures, however the ideal plate choice remains subject to discussion; reconstruction locking compression plates (RLCPs) are cheaper and can be easily contoured, whereas anatomically pre-contoured locking compression plates (ALCPs) are thought to provide better stability and therefore lower rates of mechanical failure. Objective: To compare the incidence of mechanical failures, functional and radiological outcomes in patients with midshaft clavicular fractures treated with ALCPs versus RLCPs. Methods: A propensity score matched retrospective cohort study was conducted across two trauma centers. 114 consecutively recruited patients with displaced midshaft clavicular fractures, who were treated with plate fixation and had a minimum follow-up of 6 months, were matched on gender, age, fracture grading, energy of injury, and fracture location. The resulting groups included 53 ALCP-treated fractures and 53 matched controls treated with RLCPs. Results: During a mean follow-up of 20.5 months, there were no implant deformities in the ALCP group whereas the RLCP group had 6 patients (11.3%, p=0.012) with implant deformities (5 occurrences of plate bending with fracture union, and 1 plate breakage with nonunion). Despite the higher rate of plate deformities in the RLCP group, there were no statistically significant differences in number of patients recovering full shoulder range of motion (ALCP 90.6%, RLCP 88.7%, p=0.751), incidence of rest pain (ALCP 13.2%, RLCP 9.4%, p=0.542), or implant removals (ALCP 49.1%, RLCP 56.6%, p=0.439). Conclusion: ALCPs may be superior to RLCPs in terms of implant stability but appear to produce similar clinical results. Keywords: anatomical plate; locking plate; clavicle fracture; plate fixation; midshaft clavicle


Background
Fractures of the clavicle represent 2.6-4% of all fractures in adults, with the majority (69-82%) occurring in the midshaft of the clavicle [1]. Midshaft clavicular fractures are commonly treated nonoperatively with good results [2][3][4][5][6], however plate fixation is indicated when the fracture is severely displaced or causing neurovascular injury. While the precise indications for surgery remain controversial, operative treatment generally results in better early pain control and lower rates of malunion and non-union, albeit at a slightly increased risk of surgical complications [7][8][9][10]. Different plate options are available for the fixation of midshaft clavicular fractures, however the ideal plate choice remains subject to discussion.
Traditionally, 3.5 mm reconstruction plates have been used to repair clavicular fractures since they can be easily contoured in two planes to match the S-shaped profile of the clavicle. Satisfactory outcomes have been reported following the use of these implants [11][12][13][14]. In addition, angle-stable locking screws have demonstrated improved resistance to pull-out in biomechanical studies [15][16][17][18][19].
Recently, anatomically pre-contoured plates have become more popular and have demonstrated satisfactory outcomes [20][21][22][23][24]. The proposed advantages of precontoured plates include improved stiffness, lower profile, and minimal need for additional contouring.
Presently, clinical studies directly comparing anatomical and reconstruction plates are lacking. One recent study of 55 cases reported faster union and better function compared to reconstruction plates, with no mechanical failures in either group [25].
We conducted a retrospective study in a larger cohort of patients and compared patients receiving the two implants using propensity score matching. The objectives of our study were to: (1) compare the radiological outcomes including fracture union 5 and implant stability and (2) compare the incidence of clinically adverse complications and implant removals between anatomically pre-contoured plates and reconstruction plates in midshaft clavicular fractures.

Methods
All consecutive cases coded under the International Classification of Diseases 9th revision (ICD-9-CM) procedure codes 79. 19 formerly Synthes, Oberdorf, Switzerland). The exclusion criteria were: (1) pathological fractures (excluding osteoporotic) and (2) fracture location lateral to the coracoid. Patients with a follow-up of at least 6 months and those with a known complication prior to 6 months were included for matching and analysis.
Operative procedure and rehabilitation All patients were operated in the supine position under general anaesthesia. A direct approach was used along the clavicle axis or the Langer's lines. Fixation was performed by anatomical reduction, lag screw and neutralization plate fixation, or 6 by bridging fixation when comminuted. All patients received implants from the same manufacturer. RLCPs were bent in two planes to match the clavicle profile, while ALCPs were only contoured in one plane or by twisting. For both implants, at least two locking or cortical screws were inserted in both ends. Gentle mobilization limited to shoulder level was allowed immediately after operation for the first six weeks, followed by full range of motion and strengthening exercises.

Radiological and clinical outcomes
During follow-up, standard radiographs in AP view were retrieved and examined by two surgeons, who each had at least 10 years of experience, to assess (1) fracture union and (2)  Propensity score matching Propensity score matching (PSM) [28] was performed to select cases and match the baseline characteristics of the two groups and minimize confounding from patient selection. Five baseline variables were selected for matching: age, sex, fracture configuration (AO/OTA type A: Simple vs wedge/comminution: types B/C), fracture location (lateral two fifths of the shaft vs middle of shaft or medial shaft) and highenergy injury (such as sports injuries, traffic accidents, and falls from above 2 meters). The PSM procedure was carried out using standard nearest neighbour 7 matching, and a caliper value of 0.2 [29] using SPSS software (version 23, IBM, Armonk, USA), R (version 3.10, The R Foundation, Vienna, Austria) and the Thoemmes plugin (version 3.04) [30]. After matching, baseline variables were reported and compared between the groups.

Statistical analysis
Statistical analysis was performed using SPSS software. Differences between the ALCP and RLCP groups were compared using the Mann-Whitney test. A p-value of < 0.05 was considered statistically significant throughout the study. The relative risk, absolute risk reduction and number needed to treat was calculated for outcomes with statistically significant differences.

Results
Between 2005 and 2015, we obtained 119 patients with unilateral clavicle fractures that fulfilled the inclusion criteria and exclusion criteria. Five patients were lost to follow-up and the remaining 114 had a minimum follow-up of at least 6 months (mean 20.5 months). This included 59 patients with ALCPs and 55 patients with RLCPs. After propensity score matching, there were 53 patients in each group. No baseline factors had statistically significant differences between groups before (Mann-Whitney test, p ≥ 0.300) or after matching (Mann-Whitney test, p ≥ 0.317).
The maximum standardized mean difference was reduced from 0.162 before matching to 0.119 after matching. Table 1 shows the baseline patient characteristics for both groups before and after matching. Six patients (11.3%) patients in the RLCP group were observed to have postoperative implant deformation during the 0-3 months follow-up, versus none (0%) in the ALCP group (Mann-Whitney test, p = 0.012). Out of these six patients, five had fracture union despite progressive angulation of the reconstruction plate. One patient had implant breakage and non-union after RLCP and was re-operated after three months with bone grafting and revision plating using an ALCP (Fig. 2).
Out of the six patients with plate deformations, three had comminuted fractures ( Fig. 3), one fracture was located at the lateral third, and two suffered from a highenergy injury mechanism. These numbers were not sufficient for a statistically meaningful subgroup analysis of predisposing risk factors other than implant type.
Screw pull-out was not observed as a mechanism of failure in any patients.
One patient treated with ALCP had iatrogenic injury to the brachial plexus with median nerve paraesthesia that persisted beyond two years. There were no patients with re-fractures or wound infections. Up to the latest follow-up, all patients had their clavicle fracture completely united, including the patient with non-union who received reoperation.
With RLCP being the risk-exposed group, the relative risk (RR) for implant deformation was 7.69%. The absolute risk reduction (ARR) was 11.3% and the number needed to treat (NNT) was 9.0. The combined incidence for any suboptimal clinical events including re-operations for removals was 54.7% (29 events) for the ALCP group and 69.8% (37 events) for the RLCP group (Mann-Whitney test p = 9 0.111). Table 2 displays a list of all outcomes before and after PSM.

Discussion
Clavicle fractures are one of the most common orthopedic injuries, with a bimodal peak incidence occurring in youth and the elderly [31]. Although most patients can expect good outcomes with conservative treatment [2][3][4][5][6], recent studies suggest that internal fixation results in earlier rehabilitation with better rates of successful bone union [7][8][9][10]. However, it does bear the increased risk of surgery-related complications [14]. Surgical treatment is generally indicated in cases with shortening > 2 cm, skin impingement, or painful non-union, however the precise set of indications remains disputed [8].
In terms of structural features, the two plate types used in our study can be compared as follows: the RLCP is a straight plate made of low-stiffness, annealed metal which facilitates manual contouring (reconstruction), whereas the ALCP is made of high-stiffness, cold-worked metal which is pre-shaped to match the Sshaped profile of the clavicle. Both use angle-stable locking screws for better pullout resistance [15][16][17][18], with the ALCP also accommodating optional smaller screws for fixation at the distal end.
Biomechanical studies have demonstrated a considerable advantage in plate stiffness between for ALCPs compared to RLCPs. The average cantilever failure load observed (40-42N) for reconstruction plates in one study [16] was only a quarter of that observed (170-184N) for cold-worked plates in a separate study by the same authors [15]. When tested in the tension band mode under optimal positioning, the RLCP constructs started to fail at significantly lower forces than the conventional plates (251N-355N vs 300N-345N) [15,16]. ALCPs have also been shown to withstand over three times the force in axial loading (1790N/mm vs 5740N/mm) and over twice as much torsion (130 N m/mm vs 300 N m/mm) [32]. Failure by screw hole fracture typically occurs at a similar load (about 450N) for both implants. [33].
RLCPs may be mechanically inferior to ALCPs when used to treat comminuted clavicle fractures. Taylor used a 3D mathematical model to demonstrate that the clavicle can withstand a combination of bending and torsional forces in the X, Y, and Z axes [34]. This supports the disputed claim that the clavicle does not have a true "tension-side", and therefore the tension band effect may not work for simple fractures, requiring the use of a sturdier plate. Finite element analysis has shown that anatomical plates may significantly reduce local stress under complex loads when compared to reconstruction plates [35]. These results suggest that RLCPs may not be able to withstand physiological stresses as well as ALCPs when used as a bridge in comminuted fractures.
To our knowledge, ours is one of the largest comparative clinical studies between ALCPs and RLCPs for the treatment of unstable clavicle fractures. While our study was not a randomized control trial (RCT), we obtained two reasonably balanced treatment groups using PSM [29]. This technique attempts to approximate an RCT by matching multiple confounding variables between groups, thereby minimizing bias and increasing the validity of the results. The drawbacks of PSM include the need for larger samples, a reduction in statistical power, and the risk of overlooking important confounders at the planning stage.
Our observed implant removal rates (49% of ALCPs and 57% of RLCPs) are comparable to the results of studies by Schemitsch (n = 153, 38% removed of mixed implant types) [37] and VanBeek (n = 42, 64% non-anatomical and 11% anatomical plates removed, but with shorter overall follow-up) [24]. It is worth noting that Schemitsch also found shorter body height (< 175 cm) to be a risk factor for implant removal.
RCLPs are designed with toothed edges whose indents reduce the cross-sectional moment for sideways contouring. In contrast, ACLPs have a lower profile and may produce less soft tissue irritation. Unfortunately, cadaver studies have demonstrated that "anatomically fitting" plates do not actually fit the bone in 5-32% of the population [38,39]. This is especially true in women, whose clavicles are shorter and display more exaggerated curvature.
The limitations of our study include the retrospective design, possibility of selection bias despite matching, lack of functional outcome scores, non-standardized removal indications, not comparing the speed-of-union as some studies did using CT scans, and having only a moderate sample size of 114 cases.

1.
Reconstruction plates in clavicles have a higher rate of implant deformation than anatomically pre-contoured plates due to weaker mechanical properties.

2.
Patients treated with reconstruction plates should therefore adopt a less aggressive rehabilitation programme.   (10):S1-S6.    A patient with a comminuted midshaft fracture, treated by lag screws and an RLCP. At three