We retrospectively analyzed all the patients diagnosed with PHFs between January 2017 and June 2019 in the Third hospital of Hebei medical university and the Third Hospital of Shijiazhuang city. The inclusion criteria were as follows: (1) patients over 18 years of age; (2) two-part, three-part, or four-part of PHFs on Neer classification; (3) fresh fracture; (4) no previous ipsilateral humeral surgery. Meanwhile, the exclusion criteria were as follows: (1) pathologic fractures or open fractures; (2) concomitant neurovascular injury; (3) fractures associated with shoulder dislocation; (4) mental illness. The committee of our institution waived the requirement for written informed consent because this present study was retrospective.

The patients were operated under brachial plexus block or general anesthesia. All surgeries were conducted by the same team, which was proficient in both techniques. All patients were lying on a radiolucent operating table in a beach chair position.

For the intramedullary nail group, the deltoid-splitting approach was applied. After exposing the deltoid muscle, the anterior and middle deltoid bundles were split bluntly along the direction of the muscle fibers to reach the fracture site. The entry point was revealed after temporary fixation of the fracture fragments with 2.0 mm Kirschner wires, which was 1 cm medial to the greater tuberosity. Then the Kirschner wire was used as a “joystick” to manipulate the humerus head and reduce the displacement. After the tuberosities were reduced, the three- or four-part fracture type would turn into a two-part type. Then the main nail (Targon nail) was inserted, ensuring that the end of the intramedullary nail was 2–3 mm below the cartilage of the humeral head. At last, the proximal and distal screws were locked.

For the locking plate group, the classical deltopectoral approach was applied. Both the indirect and direct reductions were performed to reduce the fracture fragments with the assistance of the C-arm. The bone defect of the humeral head was evaluated, and autologous or allogeneic iliac bone grafting was used for patients with poor bone quality and larger defects to increase the stability after plate fixation. At least five locking screws were inserted proximally, with a minimum of 3 distal screws. The nonabsorbable sutures passed through the holes on the plate (PHILOS, DePuy Synthes) and were knotted to repair the rotator cuff and increase its stability.

Clinical data and radiographic materials of all the included patients were collected, including gender, age, injury type, comorbidities, operation time, time to surgery, blood loss, and Neer classification [10]. The visual analog scale (VAS) for pain and the Constant-Murley Score was determined for shoulder function assessment.

Shoulders were immobilized for four weeks with an abduction pillow sling. Passive movement of the shoulder began on the second day after surgery. Active movements started 4–6 weeks postoperatively.

Clinical follow-up was conducted by two orthopedics attending physicians at 1 and 3 months after surgery. Postoperative plain radiographs were taken at each follow-up. Furthermore, all surgery-related complications were recorded, such as screw breakage, superficial infection, fat liquefaction, varus deformity, delayed union, and acromion impingement.

All statistical analyses were performed using IBM SPSS Statistics for Windows, version 20.0 (SPSS, Inc., Chicago, IL, USA). The distributions of all variables were evaluated for normality using the Shapiro–Wilk test. Data that satisfies normality were presented as the mean ± standard deviation. Those data that did not meet normality were presented as medians and quartiles. Chi-squared test were used to analyze the difference in gender distribution, Neer classification of humeral head fractures and injury mechanism between the two groups. The nonparametric test and Student t test were applied to analyze continuous variables. A value of P less than 0.05 was considered statistically significant.