Proximal humeral fractures are among the most common osteoporotic injuries in the elderly, accounting for approximately 5% of all fractures. Anatomical reduction and stable fixation are often difficult, making arthroplasty a reasonable option. Historically, hemiarthroplasty (HA) has been widely used in such cases. Recently, the use of reverse total shoulder arthroplasty (RSA) has increased; however, it is generally limited to carefully selected elderly patients with proximal humeral fractures. Several studies have reported that in this specific subgroup, particularly in those with poor bone quality or rotator cuff insufficiency, RSA may provide better clinical outcomes and lower complication rates than HA [1-3].

Although RSA provides reliable pain relief and restoration of elevation in elderly individuals, complications such as instability, scapular notching, glenoid component loosening, and comparatively lower implant survival rates than hip and knee arthroplasty remain concerns. Therefore, in younger and more active patients, HA is still considered an alternative to preserve the bone stock and restore near-anatomic biomechanics [4]. Nevertheless, the clinical outcomes of HA are often unsatisfactory, and tuberosity healing has been identified as a major prognostic factor for postoperative function [5-7].

Malunion or nonunion of the tuberosities leads to decreased strength, a limited range of motion, and poor patient satisfaction. Boileau et al. [5] reported that resorption or malposition of the tuberosity was the primary cause of poor results after HA, and Kralinger et al. [6] also demonstrated significantly lower Constant scores in the nonunion group than in the healing group.

Another concern following HA is the development of secondary glenoid changes, including erosion or arthrosis, which may deteriorate the long-term outcomes. Zhao et al. [8] observed glenoid erosion in 64.9% of patients at an average follow-up of 14.7 years and reported significantly worse clinical outcomes in this group. Parsons et al. [9] also found progressive glenoid wear and joint-space narrowing in all young active patients after HA, which correlated with reduced function. However, the clinical significance and progression pattern of glenoid arthrosis remain unclear.

Therefore, this study aimed to evaluate the clinical and radiographic outcomes of HA performed for proximal humeral fractures with a minimum follow-up of 3 years. This study focused on the influence of tuberosity healing, changes in the acromiohumeral distance (AHD), and the occurrence of glenoid arthrosis to clarify the midterm results and prognostic factors for HA.

This study was conducted at Pusan National University Hospital. The study protocol and data use were approved by the Institutional Review Board (IRB) of Gachon University Gil Medical Center, with appropriate data-sharing agreements in place (IRB No. GFIRB 2018-382).

This was a single-arm retrospective cohort study. It was described according to the STROBE statement available at https://www.strobe-statement.org/.

This study was conducted in Pusan National University Hospital, Korea. Patients were treated between January 2005 and December 2018 and were followed in an outpatient setting according to a standardized postoperative schedule (1, 3, 6, 9, and 12 months and annually thereafter).

Before surgery, templating was performed on both shoulder radiographs and both sides were compared. An appropriate insertion length was determined for implant insertion. All patients underwent a 30° beach-chair position under general anesthesia. A deltopectoral approach was used in all the cases. During the operation, a trial stem was inserted and the tension of the biceps tendon and deltoid muscle was adjusted. Traction was applied to confirm proper reduction and humeral head height. Cement was used to fix humeral stems. The tuberosity was fixed with Ethibond No. 5 (Ethicon) and attached to the stem. The cancellous bone was taken from the surrounding humeral head, and the sutures were bundled. To confirm whether the soft tissues, including the tuberosity, were properly sutured after surgery, we investigated whether a forward elevation of 140°, internal rotation of 50°–70°, and external rotation of 40° were possible, and whether 10–15 mm posterior subluxation was possible. Postoperatively, a shoulder abduction brace was applied for 6 weeks. Passive pendulum and isometric exercises were performed on the first postoperative day. One week postoperatively, passive shoulder motion began, and the range of motion gradually increased. Active range of motion was measured 6 weeks after surgery, and resistance strengthening exercises were allowed at 12 weeks.

Eligible participants were patients who underwent primary shoulder HA for proximal humeral fractures during the study period. During this period, HA was selected when fixation was not feasible because of severe comminution or poor bone quality. Patients were included in the final cohort if they had postoperative follow-up for at least 3 years with available clinical outcomes and radiographic evaluation. Patients were excluded if follow-up was shorter than 3 years (including death before minimum follow-up) or if they were lost to follow-up. A total of 47 patients were initially identified. Among them, two patients died and one was lost to follow-up. Consequently, 44 patients (16 males and 28 females) who were followed up for at least three years were included in the study.

The primary clinical outcome was the Constant score at the 3-year follow-up. Age- and sex-adjusted Constant scores were also evaluated. Secondary clinical outcomes included pain during active motion, measured using a 0–10 visual analog scale; patient satisfaction; and active range of motion. Radiographic outcomes comprised AHD measured on true anteroposterior radiographs; radiolucent lines around the humeral stem at the cement-bone or implant-bone interface; glenoid arthrosis, characterized by joint-space narrowing, subchondral sclerosis or cysts, or osteophyte formation on anteroposterior or axillary views; and tuberosity absorption, defined as nonunion with radiographic absorption or loss of the greater tuberosity.

Demographic and perioperative variables were obtained from electronic medical records and operative reports.

Paired comparisons of the AHD between immediate postoperative radiographs and radiographs obtained at the 3-year follow-up were performed using the Wilcoxon signed-rank test. The Mann-Whitney U test and Fisher exact test were used for between-group comparisons (tuberosity absorption vs no absorption; ≥60 vs. <60), as appropriate. All statistical analyses were performed using IBM SPSS ver. 20.0 (IBM Corp.). A P-value <0.05 was considered statistically significant.

At the 3-year follow-up, the mean VAS score was 2.1 (range, 0–5). Regarding patient satisfaction, 12 patients (27%) reported being very satisfied and 16 patients (36%) were satisfied with the operated shoulder. The mean Constant score was 46.6 (range, 13–71), the age- and sex-adjusted Constant score was 53.5 (range, 19–92). The mean active shoulder range of motion was as follows: forward elevation, 120.6 (range, 20–145); abduction, 114.5 (range, 20–140) and external rotation, 52.4 (range, 10–65) (Table 2).

At the 3-year follow-up, all stems were radiographically stable without evidence of loosening; however, radiolucent lines were observed in 10 patients (23%). The radiolucent lines were first detected at a mean of 25.2 months (range, 12–36). Glenoid arthrosis was observed in 18 patients (41%) at a mean of 34.4 months (range, 12–36) after surgery. At the 3-year follow-up, absorption of the tuberosity was observed in 17 cases (39%), with a mean of 11.2 months (range, 6–24) after surgery. The AHD was 8.4 mm immediately after surgery and 4.4 mm at 3-year follow-up (P<0.001) (Table 3).

Among the 44 patients, 17 showed tuberosity absorption and 27 did not. The median age was 72 years (interquartile range [IQR], 65–76 years) in the absorption group and 59 years (IQR, 47–72 years) in the nonabsorption group. At the 3-year follow-up, the median VAS score was 3 (IQR, 1–4) in the absorption group and 2 (IQR, 1–3) in the nonabsorption group. The median Constant score was significantly lower in the absorption group than in the nonabsorption group (35 [IQR, 34–35] vs. 50 [IQR, 49–71]; P=0.004), as was the median age- and sex-adjusted Constant score (40 [IQR, 30–50] vs. 60 [IQR, 44–78]; P=0.002). The median forward elevation, abduction, and external rotation were 80° (IQR, 50°–105°), 95° (IQR, 85°–100°), and 30° (IQR, 30°–35°), respectively, in the absorption group, compared with 145° (IQR, 145°–160°), 130° (IQR, 120°–140°), and 60° (IQR, 55°–70°) in the nonabsorption group. Differences in all range of motion parameters were statistically significant between the two groups (all P<0.001) (Table 4).

Among the 44 patients, 24 were aged ≥60 years and 20 were aged <60 years. The median age was 73.5 years (IQR, 68.5–78 years) in patients aged ≥60 years and 46.5 years (IQR, 38.5–56.5 years) in those aged <60 years. At the 3-year follow-up, the median VAS score was 2 (IQR, 1–3) in patients aged ≥60 years and 2 (IQR, 0.5–4) in patients aged <60 years. The median Constant score was significantly lower in patients aged ≥60 years than in those aged <60 years (35 [IQR, 34–49.5] vs. 50 [IQR, 50.0–70.5]; P=0.001), as was the median age- and sex-adjusted Constant score (46 [IQR, 34.5–58.5] vs. 60 [IQR, 40–80]; P=0.014). Glenoid arthrosis was observed in 10 patients (41.7%) aged ≥60 years and eight patients (40.0%) aged <60 years, with no statistically significant difference between the groups (P=1.0). Tuberosity absorption occurred more frequently in patients aged ≥60 years than in those aged <60 years (13 [54.2%] vs. 4 [20.0%]; P=0.030). Patients aged <60 years demonstrated greater range of motion. The median forward elevation and abduction were significantly higher in patients aged <60 years than in those aged ≥60 years (150° [IQR, 145°–160°] vs. 110° [IQR, 72.5°–145°], P<0.001; and 135° [IQR, 119°–140°] vs. 100° [IQR, 90°–120°], P<0.001, respectively). External rotation was also significantly greater in patients aged <60 years (65° [IQR, 61°–80°] vs. 33° [IQR, 30°–57.5°]; P<0.001) (Table 5).

One patient underwent plate fixation for a periprosthetic fracture. One patient underwent revision surgery because of a postoperative infection. Two patients experienced axillary nerve injury after trauma, and conservative treatment was administered.

This study investigated the midterm clinical and radiographic outcomes of shoulder HA for complex proximal humeral fractures with a minimum follow-up of 3 years. Overall, the patients demonstrated moderate Constant scores and satisfactory pain control at the 3-year follow-up, with approximately 64% reporting satisfaction with the procedure. Radiographically, 39% of patients exhibited tuberosity absorption at 3 years postoperatively, which was associated with inferior clinical outcomes. These findings emphasize the importance of anatomical restoration for optimal functional recovery. Additionally, radiolucent lines were identified in 23% of the cases and glenoid arthrosis developed in 41% of the cases, suggesting that mid- to long-term monitoring is essential for these patients.

As a standard treatment for proximal humeral fractures, shoulder HA is generally performed for complex comminuted fractures, dislocations, or head-split fractures. Mighell et al. [13] reported that 93% of patients who underwent HA for proximal humeral fractures were pain-free and satisfied with their clinical outcomes at follow-up, including American Shoulder and Elbow Surgeons (ASES) score and Simple Shoulder Test (SST) scores. Several reports have suggested HA for proximal humeral fractures because many patients have good functional results [14-16]. Implants have recently been developed for RSA. Therefore, RSA is becoming more common than HA for the treatment of proximal humeral fractures [17]. However, most studies have compared elderly patients, and RSA in young patients remains controversial. If plate fixation is difficult for proximal humeral fractures in relatively young patients, HA is still used as a primary treatment.

Gronhagen et al. [18] asserted that the classification of fractures was not correlated with the Constant score in the midterm follow-up and noted that the status of soft tissue, not just the fracture classification, was important as a postoperative result. We also found no correlation between the fracture type and postoperative results and could not objectively determine the preoperative state of the soft tissues. Gronhagen et al. [18] concluded that early postoperative surgery affects the outcome, which is an important factor already mentioned by several authors, and we could not analyze the correlation statistically because of the delayed operation in this study [13,19].

Many factors affect the functional outcomes of HA. The age of the patients, time to surgery after injury, the position of humeral stem and rotator cuff integrity were reported as important factors [20,21], and it is known that the anatomical reduction and union of greater tuberosity after surgery is the most important factor on postoperative function [5,19,22]. Boileau et al. [5] reported that the outcomes of 66 patients who underwent HA were related to displacement and malposition of the tuberosity. Initial malposition was observed in 27% of patients. Finally, it was reported that 50% of the cases showed an abnormal position of the tuberosity. In this case, the superior displacement of the implant, shoulder stiffness or weakness, and continuous pain were unsatisfactory. In our cases, nonunion or absorption of the tuberosity affected the postoperative prognosis. In addition, Boileau et al. [5] described factors related to malunion of the tuberosity, such as initial replacement of the implant, malposition of the tuberosity, and age >75 years. Anatomical union of the tuberosity is the most important factor in the HA. Moreover, rehabilitation should be delayed in such patients.

Antuna et al. [23] also described age as an important factor. Elderly individuals have relatively poor bone quality, which affects the healing of the tuberosity, and rehabilitation also shows poor functional outcomes. We also found that preservation of tuberosity healing was a key determinant of postoperative function.

The reported incidence of glenoid arthrosis varies depending on the radiographic definition and follow-up duration. However, there are few cases in which reoperation is required, even in cases of glenoid arthrosis [19,24]. Gronhagen et al. [18] found 35% of these cases and no such correlation, but interpreted it as having poor results in postoperative Constant scores. Mighell et al. [13] observed radiologic changes in degenerative joints in six of eight patients (72%). Two of them involved dissociation of the implant, and one case involved nonunion of the tuberosity. To solve this problem, reoperation was performed. At the time of reoperation, the patient had severe arthritis of the glenoid, concomitant glenoid replacement was performed at the time of reoperation. In our study, glenoid arthrosis was observed in 41% of patients, but the effect on postoperative results was not statistically significant. However, in our cases, it is thought that midterm follow-up was not sufficient to determine the progression of the degenerative lesion to the joint, and long-term follow-up is needed to determine the effect of the degenerative change in the joint on the prognosis.

First, it had a retrospective design with a moderate sample size, which may limit the strength of the statistical inferences. Second, although all patients were followed up for at least 3 years, long-term outcomes beyond this period (e.g., at 5 or 10 years) were not assessed. Third, our radiographic assessments (tuberosity healing, glenoid changes, and AHD) were based on plain radiographs, which may not fully capture the three-dimensional anatomic details. Despite these limitations, our study offers valuable insights into the midterm behavior of HA in proximal humeral fractures, highlighting the importance of tuberosity healing and early detection of glenoid changes. Future prospective multicenter studies with long-term follow-up and advanced imaging modalities (e.g., CT and 3D reconstruction) are needed to better delineate the factors that influence sustained outcomes after HA.

These findings are most generalizable to patients in Korea with proximal humeral fractures treated with HA because internal fixation is not feasible due to severe comminution or poor bone quality, using similar surgical technique and rehabilitation. Generalizability to other populations, surgeons, implant systems, and healthcare settings may be limited. In addition, because this was a single-cohort study without a non-HA comparison group, the results should not be interpreted as comparative effectiveness versus reverse shoulder arthroplasty or fixation strategies.

Shoulder HA for complex proximal humeral fractures provided satisfactory pain relief and functional recovery at the 3-year follow-up, when adequate tuberosity healing was achieved. However, tuberosity absorption was associated with poorer long-term outcomes. Continuous radiographic monitoring and precise surgical reconstruction of the tuberosities are essential to obtain optimal results.