The average minimum follow-up was 12 months, and the mean follow-up was 58, 6 months.

Two studies encompassed in this review did not specify the functional evaluation scale [19, 29], and among the remaining studies only 12 authors presented the functional results after conversion/revision to HA.

ASES was the most frequently reported outcome measure, followed by the Constant score (Tables 4 and 5). SANE and SST were also commonly used as secondary measures, whereas SSV appeared less often (Table 5). Other instruments—including UCLA, QuickDASH, SPADI, University of Pennsylvania, and Modified Neer—were reported sporadically; their results are summarized in the following subsections (Tables 4 and 6).

One study (Streubel) reported a single HHR to HA conversion assessed with the Modified Neer score; preoperative values were not provided and the postoperative outcome was unsatisfactory [35].

In standard HA revisions without implant exchange, Sajadi et al. reported similar UCLA improvements for both instability (N = 2: 8.5 → 12.5; Δ + 4.0) and soft-tissue failure (N = 4: 8.5 → 12.5; Δ + 4.0) (Table 6) [2]. Grubhofer et al. presented two-stage infection revisions to HA from non-specified index implants (NS, N = 8) with low postoperative scores (Constant 22; SSV 28) [26]. In the largest HA → HA series (Gauci et al.), functional scores were not reported, precluding comparison [19].

Across the seven studies reporting conversions from aTSA to HA, outcomes varied by indication (Table 4). Khoo observed the highest postoperative scores after humeral loosening (ASES 66; SANE 70; N = 8) and similarly favorable results in glenoid loosening (ASES 65; SANE 66; SST 7.4; N = 18), whereas soft-tissue failure yielded lower values (SANE 57; SST 5.6) [27]. Namdari reported inferior means in glenoid loosening (ASES 58; SANE 54; N = 17), plausibly reflecting larger glenoid vault defects requiring bone grafting [24]; Aibinder provided Modified Neer scores for glenoid loosening at 8.3-year follow-up (N = 11) but without preoperative values, limiting interpretability [36]. Infection-related revisions showed the poorest function: Grubhofer reported Constant 22 and SSV 28 in a staged setting (not specified index implants) [26], whereas Seitz (aTSA → spacer → HA, N = 5) reported a University of Pennsylvania score of 63 after cancellous grafting with biologic resurfacing, suggesting that glenoid and soft-tissue management can modulate recovery [3]. Preoperative data were inconsistently available across studies, so Δ could not be calculated in several cohorts, and subgroup denominators—particularly for humeral loosening—were small; these trends should therefore be interpreted with caution.

Across rTSA to HA conversions, outcomes differed by indication (Table 5). Glenoid loosening generally yielded higher postoperative function than instability, with infection showing the poorest results. Song reported ASES means of 59.7 for glenoid loosening (N = 3) and 53.3 for instability (N = 3) [25]. Gamradt observed the lowest scores in instability/infection cohorts, whereas glenoid loosening reached ASES 49.7; SST was 2.7 for glenoid loosening and 1 for infection/instability [37]. In a severe bone-loss series, Nezwek reported pooled ASES 52 across indications (largest share glenoid loosening), with SANE 70 and SST 3.3 [38]. On Constant scoring, Lädermann (glenoid loosening) reported 37 (N = 28) [28], Kriechling reported 33 across mixed rTSA indications (12 patients) (SSV 35) [30], and Glanzmann reported 25.2 with additional disability metrics (QuickDASH 63, SPADI 36.7) [39]. Low postoperative values in infection-driven two-stage revisions were also described by Grubhofer (Constant 22, SSV 28), although those HA conversions derive from non-specified index implants (NS) rather than exclusively rTSA [26]. Preoperative values were inconsistently reported across studies, preventing calculation of Δ in several cohorts; small subgroup denominators further limit generalizability. Taken together, glenoid loosening outperforms instability, and infection yields the poorest results, with the largest subgroup (Lädermann, N = 28) driving the most stable Constant estimates.

The authors reported a total of 77 complications accounting for 29% of total cases (Table 3): 62 were postoperative complications, 12 intraoperative, and 3 were not specified by the author.

These complications led to 42 reoperations (54.77% of all complications; 15.7% of the total implants): Sajadi et al. reported two glenoid reimplantations due to persistent painful glenoid [2], among Glanzmann’s cohort of patients, one patient, who experienced shoulder pain due to excessive glenoid medialization, was converted to resection arthroplasty [39].

Aibinder reported two intraoperative humeral fractures, treated with cerclage wiring; postoperatively, two patients experienced medial glenoid erosion. One underwent a simple reimplantation of the glenoid component, while the other underwent reimplantation of the glenoid component coupled with posterior capsule plication [36].

Grubhofer et al. reported five reinterventions without specifying which kind of complication or reintervention did the patient’s experience [26].

In Namdari’ study, five patients underwent further reintervention for glenoid component reimplant due to postoperative glenoid sided arthrosis [24].

In Gauci’s retrospective analysis 20 reinterventions were performed after revision to HA, the main causes were stiffness or pain in 12 cases, tuberosity lysis or migration in 3 cases, cuff tear in 3 cases, periprosthetic instability, and hematoma accounted for 1 case each; reinterventions were carried out without exchange of the implant [19].

Kriechling reported one intraoperative complication in the form of Iliac crest fracture, treated with plate fixation [30].

Overall, seven patients belonging to Khoo’s cohort required additional surgery: four underwent revision to total shoulder arthroplasty (two anatomic and two reverse), while three had revision HA to address pain and stiffness [27].