Avascular necrosis (AVN) of the humeral head is a progressive and debilitating condition that predominantly affects young and middle-aged adults, accounting for up to 10% of shoulder arthroplasty indications worldwide. Its etiology is multifactorial, encompassing traumatic injury, corticosteroid exposure, alcohol abuse, systemic autoimmune disease, and idiopathic cases [1]. Regardless of cause, compromised vascular supply leads to subchondral bone collapse, joint incongruity, and secondary degenerative changes, ultimately resulting in chronic pain, restricted motion, and substantial functional impairment. As the disease advances, patients experience progressive disability, diminished quality of life, and increased difficulty performing activities of daily living [2].

Nonoperative management, including physical therapy, analgesics, and bisphosphonates, is often insufficient once structural collapse begins. Consequently, surgical management remains the cornerstone of treatment for advanced-stage humeral head AVN [3]. Two primary arthroplasty options are widely used: hemiarthroplasty (HA) and total shoulder arthroplasty (TSA). HA replaces only the humeral head, preserving the native glenoid and minimizing bone loss, and is historically favored in younger individuals and those without significant glenoid degeneration. However, long-term outcomes after HA may be limited by progressive glenoid erosion, persistent pain, and variable functional recovery depending on glenoid status [4]. In contrast, TSA replaces both the humeral head and glenoid, restoring joint congruity and is associated with improvements in certain range-of-motion domains. Yet TSA carries concerns related to glenoid component loosening, surgical complexity, and implant longevity, particularly in younger or high-demand patients [5].

Although the advantages of TSA versus HA for humeral head osteonecrosis/AVN have been systematically reviewed, contemporary uncertainty persists because published syntheses differ in scope, included eras, and conclusions. For example, a systematic review by Na et al. [6] summarized outcomes and complications after HA and TSA for humeral head osteonecrosis but primarily provided aggregate procedure-specific results (with limited head-to-head comparative pooling across consistent endpoints), leaving the comparative trade-offs that drive procedure selection less clearly quantified. More recently, Daher et al. [7] performed a meta-analysis comparing HA versus anatomic TSA for shoulder osteonecrosis through August 2024, but their included evidence base (8 studies, 504 patients) and their pooled findings demonstrating mixed functional results alongside differences in complication and revision profiles highlight how sensitive conclusions may be to study selection, era, and endpoint availability.

Critically, the unresolved clinical gap is not whether one implant is “universally superior,” but how surgeons should weigh function (patient-reported outcomes and range of motion [ROM]), safety (complications), and durability (revision/survivorship) when glenoid status, patient age, and activity demands vary. In addition, larger comparative cohorts and longer-term survivorship data have emerged in recent years, which may meaningfully shift pooled estimates and help reconcile conflicting conclusions across prior reviews [8,9].

Therefore, we performed an updated systematic review and meta-analysis of comparative studies evaluating TSA versus HA specifically for humeral head AVN, incorporating evidence through November 2025 and focusing on patient-relevant outcomes (American Shoulder and Elbow Surgeons [ASES] score, Simple Shoulder Test [SST] score, ROM, visual analog scale [VAS] pain score), complication rates, and revision risk to better inform contemporary, individualized procedure selection. Rather than assuming the global functional superiority of either procedure, we aimed to quantify comparative benefits and trade-offs that remain clinically consequential in modern decision-making.

Our meta-analysis comparing TSA versus HA in humeral head AVN yielded several important findings in the context of contemporary clinical decision-making. By incorporating comparative studies through November 2025 and including larger and more recent cohorts than prior syntheses, our analysis helps update and contextualize the uncertainty that persists despite previous systematic reviews. HA was associated with significantly better ASES scores (MD, +17.12), whereas TSA demonstrated greater external rotation gains (MD, –10.23°), and HA demonstrated a significantly lower overall complication rate (RR, 0.33). In contrast, VAS pain (MD, –0.95), SST scores (MD, –0.64), internal rotation (MD, –1.60), and revision rates (RR, 0.57) were broadly similar between groups (all P>0.05).

The statistically and clinically significant improvements in ASES scores with HA suggests a relative advantage in patient-reported functional outcomes. HA preserves humeral bone stock and avoids glenoid implantation, and may yield favorable patient-reported outcomes in appropriately selected patients, particularly when glenoid cartilage is preserved. In contrast to the most recent meta-analysis by Daher et al. [7], which did not detect a statistically significant difference in ASES scores, our updated synthesis incorporating more recent comparative cohorts identified a significant difference in ASES scores favoring HA. This finding highlights that patient-reported outcomes after shoulder arthroplasty for AVN do not uniformly favor glenoid replacement and may be influenced by patient selection and expectations.

An additional and clinically plausible alternative explanation is confounding by glenoid pathology in the HA cohorts. In several included studies, preoperative glenoid cartilage integrity was not uniformly characterized, the pattern of wear (concentric vs. eccentric) was not distinguished, and postoperative glenoid progression was not consistently reported. Consequently, some patients undergoing HA may have had unrecognized, subtle, or eccentric glenoid degeneration at baseline, or experienced early progression, that would predictably compromise pain and function after humeral head replacement alone. If present, this would disproportionately worsen HA outcomes and could inflate the apparent functional advantage of TSA observed in pooled analyses, independent of the true effect of glenoid replacement itself.

Furthermore, the difference in ER favoring TSA is noteworthy. Improved ER is particularly important for activities of daily living (reaching behind, grooming). A large-scale meta-analysis in shoulder arthroplasty populations reported greater ER gains with TSA compared to HA, although results in AVN-specific reviews have been inconsistent [17]. Our findings help clarify this uncertainty by demonstrating a significant ER advantage for TSA in AVN when contemporary comparative data are synthesized. However, this advantage was not accompanied by differences in other motion domains or patient-reported measures such as SST or pain. This finding may reflect restoration of glenohumeral congruity with glenoid resurfacing, which can improve soft-tissue balancing and optimize rotator cuff–deltoid mechanics, thereby facilitating external rotation after anatomic TSA [18]. Internal rotation, in contrast, did not differ significantly, possibly because IR is more constrained by soft-tissue factors rather than bony mechanics. Importantly, this pattern aligns with our SST results, where overall functional capacity appeared similar between procedures, suggesting that while TSA demonstrates advantages in specific metrics such as ER, HA achieves equal or superior patient-reported functional outcomes in selected patients [19].

Pain reduction measured by VAS did not differ significantly between procedures, suggesting that both TSA and HA provide comparable postoperative pain relief in humeral head AVN. This suggests that both procedures achieve effective decompression of humeral pathology and pain relief from the necrotic head, which aligns with older series showing substantial pain relief after humeral head replacement, regardless of glenoid treatment [20]. For patients whose primary goal is pain relief rather than maximal function, HA may therefore still be acceptable. Moreover, HA may be associated with fewer perioperative constraints, such as shorter operative time and reduced hospital stay, which may further influence early postoperative pain levels [21,22]. Additionally, postoperative rehabilitation protocols and effective pain-management strategies play a major role in overall pain reduction, emphasizing that pain outcomes depend not only on prosthesis type but also on perioperative and rehabilitative factors [23].

The significantly lower overall complication rate observed after HA in our meta-analysis reinforces the long-standing concern that glenoid implantation introduces additional procedure-specific risks. The systematic review of Na et al. [6] reported overall complication rates of approximately 10% for HA and 17% for TSA in humeral head AVN, without a statistically significant difference between procedures, highlighting both the evolving nature of TSA outcomes and the heterogeneity of available data. However, our findings are consistent with Daher et al. [7], indicating that HA continues to demonstrate a relative safety advantage, particularly when complication definitions are pooled across diverse study designs. The nature of complications also differs meaningfully between procedures. Complications following HA most commonly involve progressive glenoid erosion, mechanical instability, and secondary pain [24], whereas TSA-related complications are more frequently associated with glenoid component loosening or periprosthetic issues [25]. A registry-based analysis demonstrated that glenoid component failure remains the predominant cause of late revision after TSA, while glenoid wear and secondary osteoarthritis dominate failure patterns following HA [26]. Taken together, these findings suggest that although modern TSA may mitigate some historical risks, HA retains a lower overall complication profile, and complication risk should therefore be weighed carefully against the functional advantages of TSA when selecting the optimal surgical approach. Importantly, patients undergoing TSA were generally older and more likely to have clinically recognized glenoid involvement, raising the possibility that observed differences reflect confounding by indication and differential baseline glenoid status rather than implant effect alone. Furthermore, although procedure-specific complication patterns (e.g., glenoid component loosening after TSA and progressive glenoid erosion after HA) are clinically distinct, inconsistent and incomplete reporting across studies precluded quantitative, type-specific comparative analyses in the present meta-analysis.

Revision rates in our study were not significantly different between TSA and HA, suggesting that long-term implant survival is broadly comparable despite differing complication profiles. This finding is particularly relevant given that some prior meta-analyses [6,7] reported higher revision rates following TSA. Our updated synthesis, incorporating newer survivorship data, suggests that these differences attenuate in contemporary practice. This interpretation is supported by evidence from a comprehensive meta-analysis by Ravi et al. [27], who reported that revision shoulder arthroplasty, regardless of the primary implant, tends to occur at similar frequencies across anatomical designs when followed long term. Their findings showed that failure mechanisms differ between implant types, but overall revision incidence converges over time, with no single arthroplasty configuration demonstrating superior long-term survivorship. Applying this framework to our results, it is likely that HA revisions are more often driven by progressive glenoid degeneration, while TSA revisions arise from glenoid component-related issues. However, the cumulative likelihood of revision ultimately remains similar. Beyond implant design itself, revision risk is also influenced by factors such as rotator cuff integrity, surgical technique, patient age and activity level, bone quality, and postoperative rehabilitation adherence, all of which contribute meaningfully to long-term implant performance. Thus, surgical decision-making should prioritize expected functional outcomes and complication risk rather than anticipated revision probability alone.

In addition, interpretations of these findings must account for the role of AVN disease stage, which is a key determinant of outcomes following shoulder arthroplasty. AVN staging and glenoid cartilage status were inconsistently reported across the included studies, precluding stratified or sensitivity analyses based on disease severity. As a result, the pooled estimates likely reflect a heterogeneous population that includes a predominance of moderate-to-advanced disease, in which TSA is more commonly selected. This limitation is particularly relevant for HA, as prior studies have demonstrated that outcomes after HA are more favorable in earlier-stage AVN with preserved glenoid cartilage, whereas TSA is generally favored in advanced stages with glenoid involvement [10,28]. Accordingly, in the present analysis we cannot define AVN stage-specific superiority and should be interpreted as reflecting comparative outcomes across mixed, and likely more advanced, AVN stages rather than guiding procedure selection for individual disease stages.

Taken together, our results carry implications for surgical planning in humeral head AVN. For patients with moderate to advanced glenoid involvement, or those whose expectations include high functional demand, particularly activities requiring substantial external rotation, TSA may offer specific functional advantages, provided that the higher complication risk and potential confounding by patient selection are carefully considered. Conversely, HA retains a clear role in younger patients, or those with preserved glenoid cartilage, lower functional requirements, or situations where bone preservation is a high priority. Carefully selected patients may continue to experience durable outcomes after HA, particularly when glenoid integrity is maintained [29]. In the context of improving implant technology and evolving patient expectations, however, our findings suggest that TSA may be considered more broadly in AVN when glenoid wear is present or anticipated.

This meta-analysis has several limitations. First, the evidence base consists primarily of non-randomized retrospective cohort studies, with only one randomized trial included. This limits the strength of causal inference and increases susceptibility to selection bias and unmeasured confounding. However, this reflects the current state of the literature in humeral head AVN, where randomized comparative trials are scarce, and observational comparative cohorts constitute the primary source of available evidence. Accordingly, the findings of this study should be interpreted as comparative outcome patterns rather than causal effects.

Second, heterogeneity in surgical techniques, including implant designs, fixation methods, glenoid management strategies, and rehabilitation protocols, may contribute to variability in functional outcomes and complication rates across studies. Such heterogeneity likely mirrors real-world variation in surgical decision-making and patient selection in AVN rather than methodological inconsistency alone.

Third, follow-up durations varied widely, with some studies reporting short-term results only, potentially underestimating late complications such as component loosening, progressive glenoid erosion (especially after HA), or long-term revision risk. Stratified or sensitivity analyses based on AVN disease stage would be particularly important, given the known stage-dependent performance of HA and TSA. However, such analyses were not feasible due to inconsistent reporting of AVN staging systems (e.g., Ficat or Cruess classification), glenoid cartilage status, and baseline disease severity across studies, as well as the limited number of studies contributing to each outcome. Consequently, the pooled results should be interpreted with caution, as they may disproportionately reflect outcomes in more advanced AVN. Similarly, inconsistent characterization of glenoid wear morphology (concentric vs. eccentric) and incomplete reporting of postoperative glenoid erosion/progression after HA limit our ability to determine how much baseline or progressive glenoid pathology influenced pooled functional differences. In addition, most studies reported complications only as aggregate event rates, which precluded pooled analyses of specific complication types or revision mechanisms despite their clear clinical relevance.

Several studies lacked detailed reporting on baseline AVN severity, rotator cuff integrity, and comorbidities, which may influence postoperative function and survivorship. Radiographic assessment methods and outcome definitions were not fully standardized across studies, further contributing to heterogeneity. As a result, subgroup analyses based on disease stage or anatomical factors could not be performed without risking underpowered or unstable estimates. Lastly, cost-effective comparisons between TSA and HA were rarely addressed, despite the clear financial implications of prosthesis type and revision burden. Future research should prioritize well-designed prospective comparative studies, standardized radiographic and functional assessments, and long-term survivorship analyses to better guide decision-making for patients with humeral head AVN.

In patients with humeral head AVN, pooled evidence from nine comparative studies demonstrates that HA is associated with higher ASES scores and lower overall complication rates, whereas TSA is associated with greater external rotation. Pain (VAS), SST, internal rotation, and revision rates were similar between procedures. These findings highlight a functional-motion-safety trade-off and support individualized procedure selection based on glenoid status, patient factors, and treatment goals. However, stage-specific superiority could not be demonstrated due to inconsistent reporting of AVN severity. Future studies should prioritize standardized staging/reporting, long-term survivorship, predictors of optimal selection, and cost-effectiveness comparisons.