Comparison of functional outcomes between medialized and lateralized reverse shoulder prostheses in acute proximal humerus fractures Authors: Luis Delgado-Flores, Miguel Pérez-Abad, Pierluigi Di Felice Ardente, Joan Miquel, Jose Carlos Minarro, Carlos Torrens Published: 2026 ABSTRACT AbstractHypothesis/BackgroundReverse shoulder arthroplasty (RSA) is increasingly used for complex proximal humerus fractures (PHFs). Although lateralized designs have demonstrated advantages in elective indications, their role in acute PHFs remains uncertain. This study evaluated functional outcomes following RSA for acute PHFs, with particular focus on the influence of greater tuberosity (GT) healing and prosthesis design.MethodsIn this multicenter retrospective study, 122 patients with acute PHFs treated with RSA and a minimum follow-up of two years were included. Two centers used lateralized designs, and two centers used medialized designs. Functional outcomes (Constant score) and rates of GT healing, scapular notching, dislocation, infection, and revision surgery were recorded. Welch’s and two-proportion z tests compared groups, and multivariate regression identified independent predictors.ResultsThe mean Constant score at the 2-year follow-up was 57.9 ± 14.1 points, with no significant differences between the medialized (58.4 ± 11.6) and lateralized models (57.4 ± 16.7). No statistically significant differences were observed in GT healing, scapular notching, dislocation, infection or revision surgery between designs. Overall, the GT healed in 65% of patients, and scapular notching was reported in 24.6%. There were significant differences in the Constant score between patients with (61.6 ± 12.3) and without (51.2 ± 14.9) GT healing (p < 0.001) and patients with (52.2 ± 15.5) and without (59.8 ± 13.2) scapular notching (p = 0.010). GT healing was the only independent predictor of higher Constant scores.ConclusionsGreater tuberosity healing was the only independent predictor of functional outcome. Differences between prosthesis designs were not significant and should be interpreted in the context of center-specific treatment strategies.Keywords: Proximal humeral fractures, Reverse shoulder arthroplasty, Prosthesis design, Functional outcome Introduction PHFs account for approximately 5% of all fractures and represent the third most common fracture type in individuals over the age of 65, with a notably higher incidence among women over 70 years of age [1, 2]. Nationwide registry data report an overall incidence of proximal humeral fractures of approximately 138 per 100,000 person-years. Rates are substantially higher in women older than 60 years, reaching nearly 500 per 100,000. Despite the aging population, the age-adjusted incidence has remained relatively stable over time [3]. In younger patients, these injuries are typically the result of high-energy trauma; in contrast, low-energy mechanisms related to osteoporosis are the predominant cause in older adults [2, 4]. Although the majority of PHFs are treated nonoperatively, the overall rate of surgical treatment in the United States Medicare population remained stable between 2010 and 2019 [5]. Surgical options include open reduction and internal fixation, hemiarthroplasty (HA), and RSA [6]. RSA is often selected for complex fractures where fixation is not feasible and poor tendon quality precludes HA, resulting in superior functional outcomes and lower revision rates [7–9]. In the United States Medicare population, the use of RSA for PHFs increased from 8.9% in 2010 to 50.9% in 2019 [5]. Several RSA designs are currently available. The original Grammont prosthesis medialized the center of rotation (COR), thereby increasing the deltoid lever arm; however, this configuration has been associated with a limited range of motion (ROM) and a greater incidence of scapular notching [10–12]. To address these limitations, lateralized COR designs were introduced. Systematic reviews and meta-analyses of both degenerative rotator cuff arthropathy and PHFs have reported comparable overall functional outcomes between medialized and lateralized implants [13–16]. In elective indications, however, lateralized designs have demonstrated superior ROM—particularly in external rotation—and a reduced incidence of scapular notching [13–15, 17]. In contrast, in PHFs, lateralized implants do not consistently have a superior ROM compared with medialized designs [16]. The impact of RSA design on GT healing in PHFs remains controversial [16, 18], and the influence of implant design on scapular notching has been insufficiently studied. Furthermore, the relationships among GT consolidation, scapular notching, and postoperative functional outcomes continue to be a matter of ongoing debate [9, 11, 16, 19–23]. Despite the growing use of RSA for acute PHFs, there is no clear consensus regarding whether medialized or lateralized designs provide superior clinical outcomes in this setting. Available evidence remains limited and sometimes conflicting, particularly concerning GT healing, scapular notching, and functional recovery. This persistent uncertainty underscores the need for comparative clinical studies. The primary objective of this study was to evaluate functional outcomes following RSA for acute PHFs, with particular focus on the role of GT healing and prosthesis design. Secondary objectives included assessing differences in outcomes and complication rates between medialized and lateralized designs, as well as the influence of scapular notching on functional results. Materials and methods A multicenter retrospective cohort study was conducted across four hospital centers, including patients with acute proximal humerus fractures (PHFs) treated with reverse shoulder arthroplasty (RSA) between January 2017 and December 2021. Inclusion and exclusion criteria The inclusion criteria were patients over 65 years of age, acute complex PHFs (three- or four-part fractures, with or without dislocation), surgeries performed within three weeks postinjury, no prior surgical interventions on the affected shoulder, and the availability of clinical and radiological data with a minimum follow-up of 24 months. Patients with incomplete data or who did not complete follow-up and those who underwent surgery after three weeks posttrauma were excluded. Four experienced shoulder surgeons participated. All procedures were performed by experienced shoulder surgeons with regular clinical practice in shoulder arthroplasty. A lateralized prosthesis (Humelock Reversed, Fx Shoulder Solutions, Viriat, France) was used in two centers, whereas a medialized prosthesis (DeltaXtend, DePuy Synthes, Warsaw, IN, USA) was used in the other two centers. Prosthesis design was determined by institutional preference, resulting in a center-based allocation of implant type rather than randomized assignment. The surgical approach (deltopectoral or anterosuperior) was selected based on surgeon preference. The deltopectoral approach was used in 82 cases and the anterosuperior (deltoid-splitting) approach in 40. Following humeral head resection and glenoid preparation, the metaglene component was intended to be implanted flush with the inferior glenoid rim. In the medialized group, the humeral stem was cemented; in the lateralized group, the humeral component was noncemented, and primary stability was achieved via the use of one or two screws. In both groups, the stem was positioned at 0° of retroversion. Tuberosity reattachment followed consistent biomechanical principles across centers. Fixation consisted of horizontal compression sutures combined with vertical cerclage sutures from the humeral diaphysis to both the greater and lesser tuberosities, using high-strength nonabsorbable braided sutures. The configuration aimed to provide circumferential stabilization and compression of the tuberosities against the metaphysis. Autologous cancellous bone graft harvested from the resected humeral head was systematically used in all cases to fill metaphyseal defects and to enhance biological healing. Although minor variations in the number of sutures and passage technique may have occurred depending on implant system and surgeon preference, the overall fixation construct and grafting strategy were comparable across centers. No metallic cables, wires, or structural allografts were used. Rehabilitation protocol Postoperative rehabilitation followed the routine protocol of each participating center. In general, patients were immobilized in a sling for approximately three weeks, followed by progressive passive mobilization beginning with forward elevation. Internal rotation exercises were introduced once forward elevation reached approximately 120°, and external rotation exercises were subsequently initiated. Supervised physiotherapy was typically continued for three to four months depending on clinical evolution. Due to the retrospective design of the study, adherence to the rehabilitation protocol and potential minor inter-center variations could not be formally monitored. Clinical and radiological evaluation Functional outcomes were evaluated annually using the Constant score [24], and the 24-month postoperative score was predefined and used for all statistical analyses to ensure a uniform time point across patients. Active range of motion was assessed in accordance with the standardized Constant score protocol, using predefined subdomain scores for forward elevation, abduction, external rotation, and internal rotation. Postoperative radiographs were obtained under fluoroscopic guidance to ensure true anteroposterior and outlet views with the arm in a neutral position. Imaging was conducted immediately after surgery and subsequently at 1, 6, and 12 months postoperatively, with annual assessments thereafter. GT healing in the anatomical position was defined based on the following radiographic criteria: (1) the GT was visible on the anteroposterior view with the arm in the neutral position; (2) the superior aspect of the GT aligned with the polyethylene insert of the humeral component; and (3) was at the level of the epiphyseal portion of the prosthesis. GT was considered malpositioned or posteriorly displaced when it was not visualized in the neutral anteroposterior projection. Greater tuberosity healing was initially classified into three categories: (1) anatomical healing, (2) malunion or displaced healing, and (3) nonunion or resorption. For the primary statistical analysis, a dichotomous variable was constructed in which only anatomical healing was considered “healed,” whereas malunion and nonunion/resorption were grouped as “not healed,” given their similarly unfavorable biomechanical implications. Representative radiographic examples of anatomical healing and non-healing are shown in Fig. 1. Fracture patterns were classified according to the Neer classification system [25] into three-part and four-part fractures. Fracture-dislocations were identified and analyzed as an independent subgroup, resulting in three fracture pattern categories included in the statistical analysis. Scapular notching was assessed via the Nérot–Sirveaux classification [26]. The Constant score was analyzed according to GT healing and the scapular notch. Greater tuberosity (GT) healing and scapular notching were independently assessed by two observers: a senior shoulder surgeon and a senior orthopedic resident. Prior to evaluation, both observers received specific instruction on the radiographic criteria described in this study to ensure consistent application of the predefined assessment parameters. Assessments were performed independently and with full blinding to functional outcomes (Constant score) and prosthesis design. Representative anteroposterior (AP) radiographs illustrating the classification of greater tuberosity healing. A Anatomical healing, with the greater tuberosity visible in the neutral AP view and aligned with the humeral component. B Non-healing (absence/resorption), with no visible tuberosity in the neutral AP projection Secondary variables Age, sex, time from injury to surgery, surgeon experience (years of independent shoulder arthroplasty practice), comorbidities (polytrauma, obesity [BMI > 30], hypertension, diabetes, cardiovascular or pulmonary disease), American Society of Anesthesiologists (ASA), approach and fracture type (fracture-dislocations were recorded as a separate variable and analyzed as a predefined subgroup in the statistical analysis), complications (dislocation and infection), and revision surgery were analyzed according to the Constant score, GT healing and scapular notch. Sample calculations and statistical analysis In accordance with Cho et al. [14], a sample size of 50 patients per group was estimated to detect a 10-point difference in the Constant score (power 80%, α = 0.05, SD = 17, with 5% attrition). A 10-point difference in the Constant score was considered clinically meaningful based on previously reported minimal clinically important difference (MCID) values for shoulder arthroplasty, which typically range between 8 and 10 points. This threshold was consistent with the reference study used for the sample size calculation. Although the final groups were slightly unequal in size (65 medialized and 57 lateralized), both exceeded the minimum sample size required per group. The observed standard deviation in the lateralized group (SD = 16.7) was comparable to that used in the a priori calculation, indicating that the study retained adequate power to detect a clinically meaningful difference. The data were collected via a coded Microsoft Excel® database at each center. Statistical analyses were performed via IBM® SPSS® software. Continuous variables are described with means and standard deviations; categorical variables are described with frequencies and percentages. External rotation was analyzed both as a continuous Constant subdomain score and as a categorical threshold (≥ 6 points) to reflect a clinically meaningful functional milestone, corresponding to the ability to place the hand on top of the head. Welch’s t test was used for comparisons of continuous variables with unequal variances and sample sizes. Proportions were compared using chi-square tests or Fisher’s exact test when expected cell counts were < 5. A p value < 0.05 was considered to indicate statistical significance. These tests were selected because of the presence of unequal variances in several continuous outcomes and to ensure accurate inferences in the comparison of proportions. Additionally, multivariate linear regression analysis was performed to identify independent predictors of the final Constant score. Variables included in the multivariable model were selected based on clinical relevance and variables showing an association with the Constant score in univariate analysis (p < 0.10), following established recommendations for purposeful variable selection in regression modeling [27]. The analysis was conducted to adjust for potential confounding factors and to determine the individual effects of GT healing, scapular notching, and other covariates on functional outcomes. Multicollinearity was assessed using variance inflation factors (VIF), and no significant collinearity was detected. Interobserver agreement was assessed using Cohen’s kappa coefficient for binary radiographic outcomes, including greater tuberosity healing and scapular notching. This study was approved by the Research Ethics Committee (Code CEI 22/48), which waived the requirement for informed consent. Results Epidemiological data During the inclusion period, 141 acute PHFs were treated with RSA. Fourteen patients were excluded due to loss to follow-up, defined as the absence of both the predefined 24-month Constant score and the corresponding radiographic assessment, as they did not attend the scheduled evaluation. These exclusions were proportionally distributed between prosthesis groups, and no significant differences in baseline characteristics were observed compared with the analyzed cohort. Three patients underwent surgery three weeks after the fracture. Two patients died during follow-up. This left 122 patients to be included. The study included 81 female patients and 41 male patients, with a mean age of 73 ± 5.6 years. The most frequent comorbidities were hypertension (62%) and obesity (39%). The ASA classification was as follows: 3 patients had ASA I, 74 had ASA II, 42 had ASA III, and 3 had ASA IV. Baseline demographic and clinical characteristics stratified by participating center are presented in Table 1. No relevant differences in age, sex distribution, fracture pattern, or comorbidity profile were observed across centers. The mean time from injury to surgery was 9.9 ± 3.7 days (median 10; range 3–19), with no significant differences between centers (p = 0.94). Time to surgery was not associated with Constant score (p = 0.528), greater tuberosity healing (p = 0.096), or scapular notching (p = 0.80). Surgeon experience was not associated with Constant score in univariate analysis (β = −0.15; 95% CI − 0.92 to 0.63; p = 0.707). Fracture-dislocations were associated with significantly lower Constant scores (49.9 vs. 59.0; p = 0.035) and lower greater tuberosity healing rates (26% vs. 70%; p = 0.003). Baseline characteristics according to prosthesis design are presented in Table 2. The medialized and lateralized groups were comparable with respect to age, sex distribution, ASA classification, fracture pattern, and major comorbidities. Table 1Baseline characteristics by participating centerVariableHospital del Mar (n = 35)Hospital Reina Sofía (n = 30)Fundació Althaia (n = 32)Parc Taulí (n = 25)Mean age (years) ± aSD74.3 ± 5.873.7 ± 6.273.0 ± 5.171.8 ± 5.0 n (%) Female sex23 (66)20 (67)21 (66)17 (68) ComorbiditiesPolytrauma1 (3)1 (3)1 (3)2 (8)Obesity (bBMI>30)12 (34)18 (60)10 (31)7 (28)Hypertension/diabetes mellitus7 (20)10 (33)2 (6.2)3 (12)Cardiopulmonary disease7 (20)3 (10)6 (19)8 (32)cASA III–IV13 (37)14 (47)11 (34)7 (28) Fracture pattern3-part fractures19 (54)16 (53)17 (53)13 (52)4-part fractures12 (34)10 (33)11 (34)9 (36)Fracture-dislocation4 (11)4 (13)4 (13)3 (12)aSD: standard deviationbBMI: body mass indexcASA: American Society of Anesthesiologists score Baseline characteristics by participating center aSD: standard deviation bBMI: body mass index cASA: American Society of Anesthesiologists score Table 2Baseline characteristics according to prosthesis designVariableMedialized (n = 65)Lateralized (n = 57)Mean age (years) ± aSD74 ± 672.5 ± 5 n (%) Female sex43 (66)38 (67) ComorbiditiesPolytrauma2 (3)3 (5)Obesity (bBMI>30)30 (46)17 (30)Hypertension/diabetes mellitus49 (75)42 (74)Cardiopulmonary disease25 (38)24 (42)cASA III–IV27 (42)18 (32) Fracture pattern3-part fractures35 (54)30 (53)4-part fractures22 (34)20 (35)Fracture-dislocation8 (12)7 (12)aSD: standard deviationbBMI: body mass indexcASA: American Society of Anesthesiologists score Baseline characteristics according to prosthesis design aSD: standard deviation bBMI: body mass index cASA: American Society of Anesthesiologists score Functional outcomes At the 2-year follow-up, the mean Constant score was 57.9 ± 14.1. Patients were followed for a mean of 42 months (range 24–62). The subdomain means were as follows: pain 12 ± 3.1, activities of daily living 15.9 ± 3.9, forward elevation 7.1 ± 2.2, abduction 6 ± 1.7, external rotation 5.5 ± 2, internal rotation 5 ± 2.1, and strength 6.2 ± 3.4. In terms of pain, 87% of patients scored 10 or higher on a 0–15 scale, where 0 indicates severe pain and 15 indicates no pain. In terms of forward elevation, 87% of patients scored greater than 90° (6 or more on the Constant score). A complete external rotation (6 points or more on the Constant score) was achieved by 69% of patients, while 31% scored 4 points or less. Internal rotation over the level of the sacroiliac junction was achieved by 51% of the patients. Baseline clinical and demographic characteristics and their corresponding Constant scores are presented descriptively in Table 3. Table 3Constant Score by age, gender, comorbidities, ASA, surgical approach, complications and greater tuberosity healingVariableMean constant scoreaSD Age < 755814.7 > 755713.0 Gender Females5713.8 Males5814.8 Comorbidities Polytrauma4315.4 Obesity (bBMI>30)5914.7 Hypertension/diabetes mellitus6012.8 Cardiopulmonary disease5416.1 c ASA I-II5912.7 III-IV5616.2 Approach Deltopectoral5813.7 Anterosuperior5715.2 Complications Scapular notch Yes (d1: 19, 2: 6, 3: 4, 4: 0)5215.5 No6013.2 Dislocation3014.3 Infection4323.0 Revision surgery3820.7 Greater tuberosity healing Yes6212.3 No5114.9aSD: standard deviationbBMI: body mass indexcASA: American Society of Anesthesiologists scoredScapular notch grades, 1–4 Constant Score by age, gender, comorbidities, ASA, surgical approach, complications and greater tuberosity healing aSD: standard deviation bBMI: body mass index cASA: American Society of Anesthesiologists score dScapular notch grades, 1–4 Comparison between medialized and lateralized models Medialized prostheses were used in 65 patients (53.3%), and lateralized prostheses were used in 57 patients (46.7%). The mean difference in Constant score between medialized and lateralized designs was small (95% CI − 4.22 to 6.24) and did not exceed previously reported MCID (minimal clinically important difference) thresholds (8–10 points), suggesting that a clinically meaningful difference between designs is unlikely within the limits of this study. The pain scores were also similar: 12 ± 2.5 for medialized patients and 12 ± 3.6 for lateralized patients. When analyzed using Constant ROM subscores, no statistically significant differences were observed between medialized and lateralized designs in forward elevation, abduction, internal rotation, or total ROM score. External rotation showed a non-significant trend favoring the lateralized group (p = 0.083). For external rotation, both models achieved a similar percentage of 6 points or more: 66% for the medialized model and 71% for the lateralized model. GT healing occurred in 62% of the medialized patients and 68% of the lateralized patients, and the difference was not statistically significant. Regarding complications, scapular notching was observed in 31% of medialized implants compared with 17% of lateralized implants. Although this difference did not reach statistical significance (p = 0.091), the confidence interval remained compatible with a clinically relevant increase in notching in the medialized group. The lateralized group showed two additional dislocations and one additional infection compared with the medialized group. However, these differences did not reach statistical significance and should be interpreted cautiously given the small number of events (Table 4). Table 4Comparison between medialized and lateralized RSA designsFunctional outcomeVariableMedialized (65)Lateralized (57)95% ICP-valueMeanSDaMeanSDa Constant score 5811.65716.7−4.22–6.240.703Pain122.5123.6−0.77–1.520.515Activities of daily living163.2164.6−1.07–1.820.608Strength6.52.96.03.9−0.68–1.830.358 b Range of motion 236.1247.7−2.84–2.200.804Forward elevation > 90º7.42.16.82.4-0.26–1.350.182Abduction > 90º6.21.76.11.8-0.54–0.710.792cER: hand on top of the head5.21.75.82.3-1.4–0.090.083dIR: on top sacroiliac joint4.81.95.12.4-1.1–0.50.465Radiographic outcomes and complicationsMedialized (65)Lateralized (57)95% ICP-valueGreater tuberosity healing40 (62%)39 (68%)-23.8–10.00.427Scapular notch20 (31%)10 (17%)−1.7–28.20.091Dislocation2 (3%)4 (7%)−11.8–3.90.315*Infection2 (3%)3 (5%)−9.3–5.00.543*Revision surgery2 (3%)5 (9%)−14.2–2.80.177*Acromial fracture1.5%0%—1.000*Values are presented as n (%)*Fisher’s exact test was used when expected cell counts were <5aSD: standard deviationbRange of motion corresponds to the Constant score subdomain (0–40)cER: external rotationdIR: internal rotation Comparison between medialized and lateralized RSA designs Values are presented as n (%) *Fisher’s exact test was used when expected cell counts were <5aSD: standard deviationbRange of motion corresponds to the Constant score subdomain (0–40) cER: external rotation dIR: internal rotation Greater tuberosity healing GT healing in an anatomic position was observed in 79 patients (65%). No secondary surgical procedures were performed specifically for isolated tuberosity non-healing during follow-up. Sensitivity analysis using the original three-category classification yielded consistent findings. Prostheses with GT healing achieved significantly higher Constant scores (62 vs. 51; 95% CI = 5.11–15.72; p < 0.001) and significantly better scores in subdomains such as pain (13 vs. 11; p = 0.015), activities of daily living (17 vs. 14; p = 0.002), strength (7 vs. 5; p < 0.001), and ROM (25 vs. 21; p = 0.002). Patients with GT healing achieved greater forward elevation (95% vs. 72%; p < 0.001), abduction (91% vs. 67%; p < 0.001), and internal rotation (60% vs. 35%; p = 0.009). Although external rotation was greater in this group (75% vs. 58%), the difference did not reach statistical significance (p = 0.059). Patients classified as ASA I–II had a higher rate of GT healing than did those classified as ASA III–IV (73% vs. 51%; p = 0.016). No significant associations were found between GT healing and age, sex, comorbidities (polytrauma, obesity, hypertension, diabetes, or cardiac or pulmonary disease), or surgical approach (Table 5). Interobserver agreement was almost perfect for GT healing (κ = 0.86) and substantial for scapular notching (κ = 0.67), according to the Landis and Koch classification. Table 5Greater tuberosity healing and scapular notch by age, gender, ASA, and Constant scoreVariableGreater tuberosity healingScapular notchYes (79)No (43)95% ICP-valueYes (30)No (92)95% ICP-valueProportionProportionProportionProportion Age < 7564%36%-0.16–0.180.85626%74%-0.08–0.230.347> 7566%34%22%78% Gender Females70%30%-0.35–0.020.06827%73%-0.23–0.080.354Males54%46%20%80% a ASA I-II73%27%-0.39 - -0.040.01623%77%-0.13–0.190.684III-IV51%49%27%73% Constant score MeanSDbMeanSDb95% ICp-valueMeanSDbMeanSDb95% ICp-valuePain132.9113.20.24–2.590.015113.5132.9-2.68–0.160.053Activities of daily living173.5144.20.80–3.780.002154.5163.7-3.32–0.310.067Strength73.353.21.07–3.51< 0.00153.373.4-3.08 - -0.260.020Range of motion255.8217.71.72–7.120.002217.6246.4-6.31- -0.020.028Forward elevation > 90ºProportionProportion95% ICp-valueProportionProportion95% ICp-value95%72%-0.37 - -0.09< 0.00187%87%85%-0.14–0.140.967Abduction > 90º91%67%-0.39 - -0.08< 0.00177%74%-0.09–0.250.307cER: hand on top of the head75%58%-0.34–0.010.05953%74%0.01–0.410.035dIR: on top sacroiliac joint60%35%-0.43 - -0.070.00937%55%-0.01–0.390.074Values are presented as n (%)aASA: American Society of Anesthesiologists scorebSD: standard deviationcER: external rotationdIR: internal rotation Greater tuberosity healing and scapular notch by age, gender, ASA, and Constant score 87% 85% Values are presented as n (%) aASA: American Society of Anesthesiologists score bSD: standard deviation cER: external rotation dIR: internal rotation Complications Scapular notching was identified in 30 patients (24.6%), most commonly grade 1 (19 patients, 15.6%). The presence of notching was associated with significantly lower Constant scores (52 vs. 60; 95% CI = -13.96 – -1.24; p = 0.020). In terms of subdomains, patients with notching tended to report worse outcomes for pain (11 vs. 13; p = 0.053) and activities of daily living (15 vs. 16; p = 0.067), although these differences did not reach statistical significance. In contrast, significant differences were observed for strength (5 vs. 7; p = 0.020) and ROM (21 vs. 24; p = 0.028). There were no significant differences in forward elevation (87% in either group; p = 0.967) or abduction (77% vs. 85%; p = 0.307). However, external rotation was significantly lower in patients with scapular notching (53% vs. 74%; p = 0.035), and internal rotation was also significantly lower (37% vs. 55%), although this difference was not statistically significant (p = 0.074). No statistically significant associations were observed between scapular notching and age, sex or ASA score (Table 5). A total of 10 complications were observed (8.2%): 2 cases of combined infection and dislocation (1.6%), 3 cases of isolated infection (2.5%), 4 cases of isolated dislocation (3.3%), and 1 acromial fracture (0.8%). Overall, dislocation occurred in 6 patients (4.9%) and infection in 5 patients (4.1%), with 2 overlapping cases. Revision surgery was required in 7 patients (5.7%). Infection was managed with debridement, antibiotics, and implant retention (DAIR) in 4 cases (3.3%), while 1 patient (0.8%) required a two-stage revision. Two additional patients (1.6%) with isolated recurrent instability underwent polyethylene exchange to increase stability. Both patients with combined infection and dislocation underwent revision surgery and are included among the five infection-related revisions. Multivariate analysis According to the multivariate linear regression analysis, GT healing emerged as the only independent predictor of higher Constant scores (β = 9.2; 95% CI: 3.9–14.4; p < 0.001). Although scapular notching was negatively associated with Constant scores (β = − 7.2; 95% CI: − 12.7–1.7), this relationship did not reach statistical significance (p = 0.110). Other variables, including fracture-dislocation ((β = − 1.98; p = 0.635), polytrauma (β = -10.1; p = 0.134), age > 75 years, ASA classification III–IV, and type of prosthetic design (medialized vs. lateralized), were not independently associated with final functional outcomes (Table 6). The final multivariate model explained 17% of the variance in Constant score (R² = 0.17). Although polytrauma was not an independent predictor in the multivariate model, the observed difference in Constant score suggests a potential clinically relevant effect that may warrant further investigation in larger cohorts. Table 6Multivariate linear regression analysis of predictors for final Constant scoreVariableβ Coefficient95% CI for βp-valueTuberosity Healing9.23.9–14.4< 0.001Scapular Notching-7.2-12.7–1.70.11Fracture-dislocation-1.98-10.2–6.30.635Polytrauma-10.1-23.2–3.20.134Age > 75 years-2.2-7.3–2.90.391Modela-2.5-7.3–2.30.306ASAb-1.1-6.1–3.90.662a Lateralized or medialized reverse shoulder arthroplastybASA: American Society of Anesthesiologists score Multivariate linear regression analysis of predictors for final Constant score a Lateralized or medialized reverse shoulder arthroplasty bASA: American Society of Anesthesiologists score Discussion The management of complex PHFs in elderly patients remains controversial. In recent years, however, the use of RSA has increased as an alternative to HA and osteosynthesis [3, 5, 8]. Therefore, understanding the factors that influence long-term functional outcomes after RSA for acute PHFs is essential. The most relevant finding of this study is that greater tuberosity healing emerged as the only independent predictor of functional outcome following RSA for acute proximal humerus fractures. In contrast, no significant differences in functional outcomes were observed between medialized and lateralized designs. Therefore, the comparison between groups should be interpreted as reflecting differences in treatment strategies rather than the isolated effect of implant lateralization. In addition, differences in stem fixation (cemented vs. uncemented) may have influenced tuberosity healing, biomechanics, and complication profiles, further supporting that our findings reflect the overall implant construct rather than prosthesis design alone. These findings are consistent with the recent systematic review and meta-analysis by Holt et al. [16] on PHFs, as well as with previous reviews by Cho et al. [14], Nunes et al. [15], and Berton et al. [13] on elective indications. Additionally, the Constant subdomains—pain, activities of daily living, strength, and ROM—were similar between the two designs. Within the individual studies included in Holt et al.’s review, some cohorts reported modest improvements in external rotation or reduced notching with lateralized designs. However, these findings were not consistently reproduced across studies. In our cohort, no functional advantage of lateralization was observed, which may reflect differences in fracture pattern distribution, tuberosity healing rates, or the elderly population characteristics inherent to acute PHF treatment. The original Grammont design has been associated with limited external rotation due to an increased deltoid lever arm without humeral lateralization, which reduces tension on the posterior rotator cuff and deltoid [10]. In contrast, lateralized virtual models improve external rotation by increasing tension on the external rotators [28]. Although several studies have shown improved external rotation with lateralized designs in elective settings [14, 15, 17], this improvement was not observed in our study or in that of Holt et al. [16]. This discrepancy could be attributed to the advanced age of the patients included in this study, in whom the infraspinatus and teres minor tendons are often degenerated. Additionally, the early rehabilitation pathways used in both groups may have minimized functional differences, which is consistent with observations by Tuphe et al. [29]. Although the biomechanical study by Costantini et al. [30] suggested that lateralized designs could increase strength by reducing deltoid loading and increasing joint stability, no significant differences in strength were observed between designs in our cohort. Nevertheless, these biomechanical explanations remain theoretical and should be interpreted with caution, as no dedicated kinematic or muscle activation analysis was performed in the present study. Another aim of this study was to evaluate the influence of prosthetic design on GT healing and to determine whether GT healing correlates with better functional outcomes. Paras et al. [31] reported lower GT healing rates when RSA was used for PHFs than in those with elective indications. In our study, GT healing was observed in 65% of the patients, which is consistent with previously reported rates [7–9]. No significant differences were found in the healing rates between the lateralized and medialized designs (68% vs. 62%; p = 0.427). However, GT healing was the only independent predictor of higher Constant scores in the multivariate linear regression analysis (p < 0.001; 95% CI = 4.5–14.5). The relationship between prosthetic design and GT healing remains controversial. Verdano et al. [18] reported higher healing rates with medialized designs, possibly due to the reduced tension on the tuberosity sutures associated with a medialized COR. Conversely, Holt et al. [16] reported higher healing rates with lateralized designs attributed to advances in reinsertion techniques and prosthetic designs that optimize rotator cuff biomechanics and reduce loading on the tuberosities [19, 28]. Similarly, the impact of GT healing on functional outcomes remains debated. The meta-analysis by He et al. [19] on the use of RSA for PHFs demonstrated that GT healing is associated with improved ROM and functionality, a finding corroborated by Jain et al. [20] and Papadopoulos et al. [21]. However, prospective and retrospective studies by Forcada et al. [9] and Torrens et al. [23] revealed no differences in Constant scores based on GT healing status. The discrepancy between these findings and ours may be related to differences in cohort composition, fracture severity, and statistical methodology. In our study, the proportion of tuberosity non-healing was substantial (35%), which may have increased the statistical power to detect functional differences. Moreover, the use of multivariable adjustment allowed the independent effect of tuberosity healing to emerge after controlling for confounding variables, whereas previous studies may have relied primarily on unadjusted comparisons. These methodological differences could partly explain the divergent conclusions. Finally, we evaluated differences in complication rates and the impact of scapular notching between prosthetic designs. No statistically significant differences in overall complication rates were observed, which is consistent with the findings of Cho et al. [14], although other reviews, such as those by Nunes et al. [15] and Berton et al. [13], reported lower complication rates with lateralized designs. Scapular notching was the most frequent complication (24.6%), with a higher incidence in the medialized designs, although this difference did not reach statistical significance (p = 0.091). In an adjusted logistic regression analysis including prosthesis design and follow-up duration, medialized design was associated with higher odds of scapular notching (OR 1.86; 95% CI 0.76–4.59), although this did not reach statistical significance (p = 0.176). This finding is consistent with previous studies on elective indications, where lateralized designs have been associated with a reduced incidence of scapular notching [13–15]. However, studies that compare both designs in PHFs are limited. It has been reported that scapular notching is less common in PHFs than in cuff tear arthropathy [26]; for example, Torrens et al. [23] reported an incidence of 14.6% in RSAs for PHFs. The higher incidence observed in our cohort could be attributed to the longer follow-up, as scapular notching tends to progress over time [15]. In the present study, radiographic progression of notching over time was not formally analyzed, as the 24-month assessment was predefined for functional comparison. However, the independent association between scapular notching and lower Constant scores suggests a potential functional impact. Biomechanically, inferior scapular erosion may alter joint mechanics, reduce effective deltoid efficiency, and impair external rotation strength, which is consistent with the subdomain differences observed in our cohort. In our study, scapular notching was associated with lower Constant scores (52 vs. 60; p = 0.020; 95% CI = -13.96 - -1.24), particularly in the subdomains of strength and ROM. While some studies have reported no significant associations between scapular notching and functional outcomes [10, 23], our results are consistent with those of Mollon et al. [11] and Simovitch et al. [22] in elective RSA settings. With respect to other complications, dislocation occurred in 5% and infection in 4% of cases, which is consistent with previous reports on the use of RSA for PHFs [12]. Given the contradictory findings in the literature, prospective randomized studies are necessary to evaluate the influence of prosthetic design on functional outcomes, GT healing, and complication rates in patients with PHFs treated with RSA. In the present study, GT healing was identified as an independent predictor of better functional outcome, whereas scapular notching showed a negative association that did not reach statistical significance after adjustment. However, these findings should be validated by studies with larger sample sizes and longer follow-up periods. Several limitations must be acknowledged. First, the retrospective and non-randomized design introduces inherent risks of selection bias and unmeasured confounding. Prosthesis allocation was center-based rather than randomized, and therefore implant-related effects cannot be fully separated from center-specific factors, including surgeon experience, technical nuances, perioperative management, and rehabilitation practices. Second, stem fixation differed systematically between groups (cemented in medialized implants and uncemented in lateralized implants), making fixation method colinear with prosthetic design and preventing independent analysis of cementation as a covariate. Consequently, the results reflect the overall implant construct rather than design alone. Third, surgical approaches (deltopectoral or anterosuperior) were selected according to surgeon preference, and multiple surgeons with varying experience levels participated. Although all were experienced shoulder surgeons, subtle technique-related differences cannot be excluded. Fourth, greater tuberosity healing was assessed using predefined radiographic criteria without advanced imaging modalities, and although interobserver agreement was high, the classification system used has not been formally validated. Finally, the study was not powered to detect differences in relatively infrequent complications such as dislocation, infection, or acromial fracture. Despite these limitations, the multicenter design, inclusion of a relatively large and well-characterized cohort, predefined radiographic assessment criteria with blinded interobserver evaluation, and a uniform 24-month functional endpoint strengthen the methodological rigor and clinical relevance of the study. Conclusions In this multicenter retrospective cohort, greater tuberosity healing was the only independent predictor of functional outcome following RSA for acute PHFs. Although no significant differences were observed between medialized and lateralized designs, these findings should be interpreted cautiously given the center-based allocation and differences in implant constructs. Author contributions All authors contributed substantially to the conception and design of the study, data collection, analysis, and manuscript preparation. All authors read and approved the final version of the manuscript. Funding The authors received no financial support for the research, authorship, or publication of this article. • Conflicts of interest: The authors declare no conflicts of interest related to this work. • Ethics approval: The study was approved by the local institutional ethics committee (Code CEI 22/48). • Consent to participate: The requirement for informed consent was waived by the ethics committee. • Consent for publication: Not applicable. • Availability of data and materials: The anonymized dataset is available from the corresponding author upon reasonable request. • Authors’ contributions: All authors contributed substantially to the conception and design of the study, data collection, analysis, and manuscript preparation. All authors read and approved the final version of the manuscript. Data availability The anonymized dataset is available from the corresponding author upon reasonable request. Declarations Competing interest The authors declare no competing interest. Footnotes Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. References Associated Data Data Availability Statement The anonymized dataset is available from the corresponding author upon reasonable request.