**4. Discussion**

The purpose of our study was to evaluate clinically and radiographically the outcomes of shoulder arthroplasty in young patients with PANHH. Both HA and TSA improved the clinical function of affected patients significantly after 8.2 years (mean FU). Revision surgeries for glenoid wear were low (2%). Complications after TSA were excessively high with 32% exhibiting glenoid loosening and 26% receiving revision surgery.

Mansat et al. [2] reported outcomes on 19 HA with a mean FU of 7 years. Mean Constant score (58 points) was significantly improved at last FU. Post-irradiation PANHH yielded the worst results. At long term, with a mean FU set at 12 years, Smith et al. [6] confirmed in 31 HA that mean motion range had still improved significantly (*p* < 0.001).

In our study, two hemi-metal implants (2%) had revision surgery for glenoid wear at 2.4 and 4.1 years postoperatively, with pre-operative Ficat classification at stage 3. Mansat et al. [2] related painful glenoid wear developed in 2 of the 14 HA (14%) at 6.2 and 9.6 years of FU. Only one patient with low Constant score (33 points) had revision surgery. At long term, Smith et al. [6] noted 14 glenoid erosions out of 23 shoulders (61%), but only 2 patients had revision surgery for TSA (7%). The estimated survival rate for HA in their study was 100% after 5 years and 92% after 10 and 15 years.

In our study, survival rate without revision surgery was 97% (89–99%) and 95% (68–99%) at 5 and 10 years. No significant differences between different types of HA, regarding postoperative complication, revision surgery or clinical outcomes were observed. Nevertheless, anatomic cemented stems with metal head (Aequalis, Tornier SAS-Wright Medical) were the device which had the longest follow up and excellent treatment outcomes.

In Herschel et al. [20] study, valgus position of the prosthetic humeral and glenoid cysts were identified as risk factors for glenoid erosion after HA. The size of the humeral head component was not correlated with glenoid erosion in the study of Al-Hadithy et al. [21].

TSA gave excellent results at short and middle term but it exposed patients to glenoid implant loosening.

In our study, glenoid loosening occurred in 6 of 19 TSA (31.6%) between 7 and 10 years FU but this cohort was small. Four TSA had been reoperated. Two glenoids components considered as loosened did not undergo a new surgery (CS = 49 and 67).

Schoch et al. [4] followed 71 TSA after PANHH with a mean follow up of 7.7 years. Pain and range of motion were significantly improved. Among them, 11 (15%) underwent reoperation at a mean time of 4.4 years (range, 0.6–11 years) after index arthroplasty. Four patients (5%) needed to be reoperated for aseptic glenoid loosening.

In a prospective study, Parch et al. [3] prospectively evaluated 13 TSA at a mean followup of 30.2 months (range, 14–49 months). Shoulder function assessed by the Constant score improved from 18 (adjusted score, 24%) to 51 (adjusted score, 69%; *p* < 0.001). They observed that patients younger than 65 years obtained lower adjusted Constant scores (mean, 58%; *n* = 7) than patients older (mean, 82%; *n* = 6; *r*s = −0.73, *p* = 0.02). During follow up, the patient with the lowest adjusted Constant scores was the one with progressive glenoid erosion preoperatively.

Relatively few studies compared the outcomes between HA and TSA for PANHH in the literature. Recently, a study by Ristow et al. [5] assessed 10 TSA and 19 HA and showed no significant differences in clinical outcomes with a mean follow-up of 3.9 years (range, 1–8.5 years). Mean age at surgery was 49.2 years (range, 16–77 years). It demonstrated a trend of better outcome scores with TSA but without statistical significance. Traumatic cases concerned 20% of their patients which impacted the results.

Feeley et al. [1] compared 26 HA vs. 17 TSA with 4.8 years of FU. TSA was associated with lower ASES score and decreased forward flexion compared to hemiarthroplasty (*p* < 0.05). There were significantly more reoperations in the TSA group (22%) among which 4 exhibited glenoid loosening. Schoch et al. [4] compared 67 HA vs. 71 TSA with a mean FU of 9.3 years. At the time of final follow-up, active elevation was significantly higher in the HA group (*p* = 0.04).

In our study, despite a shorter follow up with HA, 2 HA had revision surgery for glenoid wear with a mean follow up of 11.9 years. Twenty years later, the percentage of reoperation-free patients was calculated to be 87%. Fifteen percent of TSA had revision surgery with a mean time of 4.4 years at index surgery. Four of eleven patients were reoperated for aseptic glenoid loosening. Reoperation-free survival rate was calculated to be 79% (CI, 67–92).

Our study has inherent limitations due to its retrospective and multicentric design. It mixed different kinds of hemiarthroplasties with a heterogeneous follow-up. Moreover, the cohort of TSA was smaller with a smaller FU than HA. Nevertheless, it analyses one of the longest FU in the literature about shoulder arthroplasty for PANHH. There were no statistical differences between clinical outcomes and post-operative complications with the stage of the osteonecrosis. Glenoid wear occurred rarely after HA. TSA seemed to be complicated by glenoid loosening more. Humeral metal-head implants gave excellent results and are still a good option for HA.
