Although the optimal surgical treatment of proximal humerus fractures has not been determined, there have been many operative techniques described, including percutaneous fixation, conventional plate fixation, intramedullary fixation with rods or pins, tension band wiring, and blade plate fixation, whose clinical outcomes have varied . The current trend for treating these fractures utilizes locking plate technology as its lower profile may reduce impingement; its multiple divergent locking screw positions allow for improved fixation; and its biomechanical properties provide improved stability and load to failure. A recent biomechanical analysis in which blade-plate fixation was compared with locking plate fixation for the treatment of proximal humeral fractures demonstrated potential advantages with use of the locking plate technology .
In the current study, excellent or satisfactory clinical results were realized in 73.0% of cases, with good radiographical results in 77.8% of cases. These clinical and radiographic findings suggest that locking plate fixation and suture augmentation for proximal humerus fractures provide value as an operative treatment modality. However, room for improvement exists as 14 complications occurred in thirteen of 63 cases, representing a complication rate of 22%. These complications included screw perforation, plate irritation, loss of reduction, osteonecrosis, nonunion, screw loosening, and implant breakage (Table 2).
There have been several reports about the complications encountered with locking plate technology. Egol et al.  reported complications in 12 of 51 patients (24%) following proximal humerus locking plate fixation at 16 months follow-up. The complications occurred in eight patients (16%), including intraarticular screw penetration, osteonecrosis, acute fracture, nonunion, and heterotopic ossification. Similarly, Owsley et al.  reported a 36% complication rate in 53 patients, with intraarticular screw penetration occurring in 23% and a statistically significant higher radiographic complication rate noted in patients older than 60 years of age. In a study by Lee et al. , 20% of 45 patients had postoperative complications that included loss of fixation, adhesive capsulitis, and deep infection, while Sudkamp et al.  reported various complications in 34% of 155 patients including: screw penetration, plate impingement, infection, loss of reduction with or without screw perforation, humeral head osteonecrosis, nonunion, screw loosening, plate pullout, and implant breakage. Brunner et al.  reported an overall complication rate of 35% and Badman et al.  presented 13 complications (16%) in 81 patients and reported varus collapse in 5 patients (6%), intraarticular screw penetration in 3 (3.7%) and osteonecrosis in 5 (6.2%). Königshausen et al.  reported 12 (23.1%) complications in 73 patients. The overall complication rate in the current study was not higher compared to previous reports (22% versus 16- 36%)
Sudkamp et al.  classified the complications related to the locking plate into four categories. They reported 25 (40%) initial incorrect surgical technique-related complications amongst 62 total complications, which included primary intraoperative screw perforations of the humeral head in 21 cases and subacromial impingement due to significant cranial positioning of the plate in 4 cases. They suggested that adherence to proper surgical technique is necessary to avoid iatrogenic errors. Others, including Badman et al.  and Brunner et al. , have reported on primary and secondary intraarticular screw penetration into the glenohumeral joint and recommended that more accurate screw length measurement and shorter screw selection would prevent primary screw perforation.
In the current study, there were three (4.8%) initial incorrect surgical technique-related complications encountered in 63 cases at the end of the operative procedure: primary intraoperative screw perforation in two cases and subacromial impingement in one case. We believe that confirming screw position in more than one plane using an image intensifier will decrease the incidence of these surgical technique-related complications.
Of the four types of complications, we believe that bone- and fracture- related complication are the most important to prevent because they can negatively impact clinical and radiographic outcomes. More importantly though, this type of complication is under the surgeon’s control, and thus can be avoided with meticulous surgical technique. In the current study, all six cases with failed clinical outcomes had bone- and fracture- related complications. Our findings mirror previous work in this field, including Sudkamp et al.  who reported twenty one (13.5%) bone- and fracture- related complications in 155 patients.
The secondary screw perforation with reduction loss in the treatment of proximal humerus fractures with locking plates has already been described. The rigidity of this angled locking device is responsible for screws cutting through osteoporotic bone, leading to humeral head subsidence because of a deficient posteromedial bone buttress or osteonecrosis . According to Brunner et al. , secondary varus angulation was observed in five patients, where screws were placed in three of these patients to support the medial buttress. However, none of the five patients received tension band sutures between the rotator cuff and the plate to neutralize traction forces. Based on their observations, they suggested that traction forces from the rotator cuff should be neutralized using tension band sutures combined with screws supporting the medial calcar, especially when medial support is insufficient. Similarly, Badman et al.  documented that restoration of the medial calcar and supplemental suture fixation may decrease the incidence of hardware-related complications.
In our study, augmentation with a tension band construct using non-absorbable sutures through the rotator cuff to the holes in the plate was applied in all cases. However, despite augmentation with a tension band construct, there were six cases where secondary reduction loss with or without screw perforation occurred in patients without medial support (Tables 3 and 4).
The incidence of humeral head osteonecrosis following locking plate fixation at short-term follow-up has been reported in 3.8% to 25% of cases [6, 7, 11, 12]. In our current study, we had one case of humeral head osteonecrosis (1.6%) in a patient with a two-part proximal humerus fracture-dislocation. It is thought that open reduction and internal fixation may increase the risk of osteonecrosis unless the medial capsular structures and metaphyseal bony attachments are maintained so as to preserve the humeral head blood supply . We believe that we had a low incidence of osteonecrosis because we were able to utilize the locking plate system to indirectly reduce the fracture fragments and avoid additional soft tissue dissection and damage near the fracture site.
We experienced plate breakage in one case, when a patient fell and landed on his outstretched, operative limb at two months following his index procedure. The patient subsequently underwent osteosynthesis using a proximal humerus locking plate with iliac crest bone graft, leading to successful bony union. We believe this complication was due to implant fatigue failure and a deficient posteromedial calcar in the setting of a low energy traumatic event. We believe that medial support may be important to resist implant fatigue.
Recently, much attention has been paid to the importance of the medial column for maintaining stable fixation of proximal humerus fractures [9, 13–17, 24–27]. Anatomic reduction and restoration of the medial calcar allow the medial column to both buttress and reduce the stresses of laterally-based plate fixation. Gardner et al.  first emphasized this concept by noting that when mechanical support of the inferomedial region of the proximal humerus was obtained, fracture subsidence was significantly reduced postoperatively. They suggested that mechanical support of the medial column may be achieved either with placement of humeral head screws inferomedially or endosteal fibular allograft strut augmentation when anatomic cortical contact is not possible [14, 26, 27]. They reported that lack of medial support led to a 30% screw perforation rate compared to a 6% screw perforation rate for fractures with an intact medial column.
According to our experience, direct placement of an oblique long locking screw into the inferomedial quadrant of the proximal humeral head is considered as the more important and substantive way to obtain the medial support when there is medial communition. In contrast, anatomical reduction of the medial calcar with good cortical contact, especially in patients without medial cortex comminution, may be additive in their ability to prevent postoperative complications.
Lee at al  reported that absence of comorbidity and the restoration of the medial metaphysis were the most reliable predictors of successful clinical outcomes, while Solberg et al.  recognized that the presence of a metaphyseal segment in the region of the medial calcar greater than 2 mm was associated with better clinical outcomes and independent of Neer fracture type. Others have corroborated the importance of restoring and maintaining the medial calcar to enhance mechanical stability and to avoid reduction loss [13, 15, 16, 25]. In situations where the host bone is osteoporotic and anatomic reduction and restoration of the posteromedial column cannot be achieved, it is recommended that augmentation with endosteal fibular allograft struts or primary arthroplasty be considered . Zhang et al.  reported clinical and radiological outcomes of seventy-two consecutive patients (mean 30.8-month follow-up, MS+ group: 29 patients; MS- group: 39 patients). They showed a statistically significant difference regarding the failure rate (23.1% in the MS- group vs. 3.4% in the MS+ group). They documented that the early loss of fixation was related to higher age and less initial neck-shaft angle of the patients. However, bone mineral density was not significantly associated with loss of fixation. They also observed a significantly lower final neck-shaft angle in the MS- group and greater secondary angle loss in the subgroup of Neer three-part (P = 0.033 and 0.015, respectively) and four-part fractures (P = 0.043 and 0.027). Therefore, they concluded that medial support for proximal humerus fractures seems to have no benefits in Neer two-part fractures, but the additional medial support screws inserted into the inferomedial region of the humeral head may help to enhance mechanical stability in complex fractures and allow for better maintenance of reduction.
In the current study, eight major bone- and fracture-related complications occurred in the non-medial support group, while one major bone- and fracture-related complication occurred in the medial support group. Utilizing regression analysis, we found that only one factor, namely the presence of medial support (and not preoperative fracture type), was responsible for predicting major bone- and fracture-related complications in the treatment of proximal humerus fractures using a locking plate and suture augmentation. There appears to be a trend towards significance for osteoporosis on multivariate regression analysis with p-values approaching 0.05. Further, a subgroup analysis of 30 cases with osteoporosis demonstrated that medial support restoration led to significantly less major bone- and fracture-related complications compared to medial support loss. Thus, even osteoporotic patients can benefit from achieving and maintaining medial support so they decrease their chances of having a major complication such as reduction loss or nonunion.
The limitations of the current study are its retrospective nature and small sample size, potentially introducing bias and β-error. We did not perform a priori power calculation analysis for sample size estimation as this work is a preliminary report of our experience. A post-hoc power analysis demonstrated that the observed power for the addition of the set of independent variables was 0.64458902. (Statistics calculators, version 3.0). Future work will focus on performing a prospective, randomized study that is adequately powered to have sufficient sample size to detect differences between the two groups. There was no control group in the present study; therefore, we cannot determine if another treatment method would have led to different results. In addition, we used two different types of locking plates (PH-LCP and PHILOS plate). These plates differ in profile and plate design, including the number of locking holes and the thickness of the plate, both of which can affect the overall outcome. Although we tried to place the oblique long locking screw into the inferomedial quadrant of the proximal humeral head intentionally after noticing the role of the medial support reported by Gardner et al. , some size mismatching between the humerus and the locking plates precluded the placement of the oblique long locking screw at the intended location. This led to both the MS+ and MS- groups comprising cases treated with both different types of locking plates, which can affect the overall outcome also. With regards to the presence or absence of osteoporosis, regional osteoporosis in the hip or spine was used in this study as a surrogate for the presence of local osteoporosis at the proximal humerus. But, this may be not the ideal method to assess local osteoporosis of the proximal humerus. Additionally, we divided the patients to two groups based on the presence or absence of osteoporosis (i.e., t-score < -2.5). We did not see any linear correlations between the t-scores and the other possible factors. Finally, although we describe a generalized postoperative rehabilitation protocol in the Methods, most cases had an individualized protocol for length of sling immobilization and timing of passive range of motion exercises. This depended on the fracture type, degree of medial comminution, and intraoperative fixation stability. These factors may have affected the clinical and radiographic outcomes.