2.1 Study design

The IRB at our institution approved this study (IRB-17-00684-CR002).

One fellowship-trained shoulder and elbow surgeon performed all of the aTSAs in this study. Cases between 2001 and 2015 were ultimately included after applying exclusion criteria. Of the 619 total cases queried, 275 patients were deceased, 32 declined to participate, and 138 were lost to follow-up. This left 173 cases remaining for inclusion in our study. Patients were additionally excluded for a variety of factors: 11 metal-backed glenoids, 12 missing glenoid type, 4 avascular necrosis, 17 missing indications, 44 missing Walch class, 11 missing pre-operative ASES scores.

Of the 138 patients (173 shoulders) identified who had a minimum 2-year follow-up, 74 patients met all inclusion and exclusion criteria for the study, and 36 were eventually included in the 1:2 matched cohort analysis based on the matching algorithm (matchit package version 4.5.2 with cardinal matching, R version 4.1.0). Given the lesser number of keeled glenoids in our cohort, we elected to use a 1:2 keeled:pegged glenoid matching algorithm to maximize the statistical power of our matched sample. A match could not be made for 38 patients (3 keeled and 35 pegged) and they were excluded from the study.

The study population was divided into cohorts based upon glenoid design: pegged or keeled (based on postoperative X-rays). Demographic information was sourced from the electronic medical record. The Walch classification system was utilized to classify preoperative glenoid morphology.^2,29 The cohort groups were then matched for age, sex, pre-op ASES, indication, and Walch classification.

2.2 Study outcomes

Shoulder ROM and PROs were measured throughout the study period. The measurements were collected before and after arthroplasty at each visit. The ROM from the last follow-up visit was used as the postoperative measurement. ROM was evaluated by forward elevation (FE) and external rotation (ER). Vertebral levels were used to evaluate internal rotation (IR) as described by Amroodi et al. PROs collected included the Visual Analog Scale (VAS) pain score, preoperative American Shoulder and Elbow Surgeon (ASES) score, and Simple Shoulder Test (SST) score.¹

The acromiohumeral interval (AHI), Walch glenoid classification, and glenoid loosening was assessed using X-rays. The Walch classification is used to describe glenoid morphology in patients with primary osteoarthritis, and is based on “axial cuts of two-dimensional CT scans.”^2,29 Both pre-and post-operatively imaging was reviewed by multiple fellowship-trained surgeons (PJC and BOP). Sanguanjit et al. describe AHI as “the shortest distance between the inferior cortex of the acromion and the top of the humeral head.”²⁵ The measurements were completed using the first postoperative radiograph that was most recently completed. Glenoid loosening was measured as described by Lazarus et al. and ranged from grade 0 to 5, with grade 5 indicating maximal glenoid loosening. This degree of glenoid loosening requires revision.¹⁴ The glenoid loosening measurements were based on a review of the most recent X-ray only.

2.3 Surgical technique

The deltopectoral approach was utilized for all TSA procedures. All patients received the Bigliani/Flatow Complete Shoulder Solution TSA (Zimmer, Warsaw, IN, USA) implant. In regards to subscapularis management, it was managed using one of two techniques: tenotomy or lesser tuberosity osteotomy. The subscapularis was repaired at the conclusion of the arthroplasty. The surgeon used intraoperative discretion to decide whether to use a pegged or keeled glenoid. All glenoid prostheses were cemented. Patients with smaller native glenoids received a keeled glenoid. For patients receiving a pegged glenoid, a smooth 3 peg implant was used.

2.4 Statistical analysis

Python version 3.8.8 was used to statistically evaluate the data. This study reported data using both continuous and categorical variables. Continuous variables are reported as means ± standard deviations, and categorical variables as counts and percentages. Comparisons between keeled and pegged glenoids were performed using chi-square tests or independent two-sample t-tests, as appropriate. A p-value below 0.05 was used to denote statistical significance.

3.1 Study population

This study analyzed 36 TSAs, which included 12 keeled glenoid components and 24 pegged components. These were matched in a 1:2 ratio by age, sex, pre-op ASES, and Walch classification grade. The mean age at the time of surgery was 65.5 ± 11.3 years, and 65.9 ± 6.1 years in the keeled and pegged glenoid groups, respectively (p = 0.903). The mean follow-up length was 8.9 ± 2.6 years and 9.2 ± 3.7 years in the keeled and pegged glenoid groups, respectively (p = 0.756). Follow-up ranged between 5.0 and 13.4 years in the keeled cohort. Follow up ranged between 3.5 and 17.3 years in the pegged cohort. There were 8 (66.7 %) females in the keeled glenoid group and 16 (66.7 %) females in the pegged glenoid group (p = 1.00). The mean BMI for the keeled glenoid group was 23.4 ± 4.1 and 26.9 ± 6.6 kg/mˆ2 in the pegged glenoid group (p = 0.133) (Table 1). Regarding subscapularis repair technique, the keeled cohort included 5 tenotomies, 6 lesser tuberosity osteotomies, and 1 unknown repair technique while the pegged cohort included 7 tenotomies and 17 lesser tuberosity osteotomies.

Preoperatively, there were no significant differences in FE (p = 0.757), ER (p = 0.450), IR (p = 0.268), VAS (p = 0.571), ASES (p = 0.940), and SST (0.263). Additionally, there was no significant difference in the distribution of preoperative glenoid morphologies between the keeled and pegged groups. For the patient population as a whole, the most common preoperative glenoid morphologies were A1 (33 %) and A2 (41.67 %) (Table 2).

3.2 Clinical outcomes

When the ROM and PRO outcomes were compared pre-and post-operatively, both keeled and pegged glenoid groups showed significant improvements in all ROM and PRO measurements. In the keeled glenoid group, FE (p = 0.013), ER (p < 0.001), IR (p = 0.029), ASES (p < 0.001), and SST (p < 0.001) significantly increased, and VAS (p < 0.001) significantly decreased. Similarly, in the pegged glenoid group, FE (p < 0.001), ER (p < 0.001), IR (p < 0.001), ASES (p < 0.001), and SST (p < 0.001) significantly increased, and VAS (p < 0.001) significantly decreased (Table 3).

However, when keeled and pegged glenoid cohorts were compared against each other postoperatively, there was no significant difference in ROM scores: FE (p = 0.333), ER (p = 0.462), and IR (p = 0.411). Similarly, PRO scores did not significantly differ between the keeled and pegged glenoid cohorts: VAS (p = 0.101), ASES (p = 0.192), and SST (p = 0.662) (Table 4).

Pre-to-postoperative changes in ROM and PROs did not differ significantly between the keeled and pegged cohorts. There was no significant difference in FE change (p = 0.743), ER change (p = 0.895), IR change (p = 0.065), VAS change (p = 0.448), ASES change (p = 0.313), and SST change (p = 0.857) (Table 5).

3.3 Radiographic outcomes

Radiographic analysis revealed an overall decrease in the AHI in both the keeled and pegged glenoid cohorts from preoperative to the most recent postoperative follow-up (−3.09 ± 3.06 vs. −2.65 ± 2.56; p = 0.647). However, there were no significant differences in immediate post-operative AHI (11.08 ± 5.01 vs. 10.70 ± 3.48; p = 0.790) or AHI at the most recent follow-up (7.99 ± 4.38 vs. 8.05 ± 2.25; p = 0.955). Additionally, there was no significant difference in glenoid loosening score determination. With regards to the Lazarus classification, most patients received a score of 0/5, however, this result was not statistically significant (pegged: 66.7 % vs. keeled: 83.3 %; p = 0.479) (Table 6).

3.4 Revisions

Among the 36 total TSAs, only one keeled glenoid (8.3 %) was revised after 9.95 years due to glenoid loosening.

4.1 Patient-reported outcomes and clinical efficacy

The findings of this study reveal no significant differences in PROs between pegged versus keeled glenoid implants, with both groups having demonstrated significant improvements in PROs postoperatively. With a larger sample, it is possible that we would have found significant differences based on the type of glenoid component. While there may be biomechanical or procedural preferences for one design over the other, in our small study of 36 shoulders, clinical outcomes did not vary significantly based on the type of glenoid component used. Welsher et al. conducted a meta-analysis of 7 comparative and 25 noncomparative studies examining TSAs that used either pegged or keeled glenoid components and found no significant difference in functional outcomes such as American Shoulder and Elbow Surgeons (ASES) shoulder score and Constant scores. Notably, however, of the seven comparative studies included in their analysis, only two reported these score indices.³⁰

Duey et al. retrospectively reviewed 144 TSAs with either keeled or pegged components and found no difference in ASES scores or simple shoulder test scores (SST) at long-term follow-up (>9.3 years).⁸ Kilian et al. prospectively randomized 59 TSAs to either pegged or keeled glenoid components and found there were no differences in clinical outcome scores or active mobility outcomes between the two cohorts at an average of 7.9-year follow-up.¹³ Throckmorton et al. reported similar outcomes with pegged and keeled designs in terms of pain relief, functional improvement, and risk for loosening at a mean follow-up of 51.3 months for the keeled group and 45.7 months for the pegged group.²⁷ Our results and previous literature reinforce that the restoration of shoulder function and alleviation of pain is not contingent on the type of glenoid component used.

4.2 Range of motion analysis

This study showed that both keeled and pegged glenoid groups demonstrated significant improvements in FE, ER, and IR postoperatively without significant differences between the two cohorts. This aligns with findings from previous studies from Throckmorton and Kilian et al. which also reported that ROM was similar when comparing glenoid configuration designs.^13,27 Duey et al. also reported that FE and ER were comparable between glenoid configurations, however, they uniquely highlight a significant difference in IR between groups, with patients that received a pegged component demonstrating greater IR.⁸

4.3 Radiographic outcomes

Radiographic outcomes, particularly the incidence of lucencies, are thought to play a critical role in evaluating the long-term success of TSA implants. Glenoid lucencies are often viewed as a harbinger of implant loosening and the potential development of pain, decreased function, and the need for revision surgery. Radiographically, the results in the current study showed no significant differences in the incidence and severity of periprosthetic radiolucencies between pegged and keeled designs. These findings may imply that both component designs have the ability to provide comparable long-term stability. Existing literature on radiographic outcomes of pegged and keeled components has mixed results. Duey et al. showed that glenoid loosening differed between keeled and pegged implants at long-term follow-up.⁸ Trail and Nuttall compared patients with pegged and keeled components in their study with a mean follow-up of 5.7 years and found that 90 % of keeled components had a radiolucent line of more than 1 mm in at least one zone compared to 36 % in the pegged component group.²⁸ The incidence of radiological evidence of loosening in the keeled group was 53 % although none required revision and was significantly greater than the incidence of radiological loosening in that study's pegged designs. Conversely Welsher et al. found no significant differences in the rate of lucencies discovered postoperatively between pegged and keeled glenoid designs.³⁰ Similarly, Moulton et al. evaluated patients after a minimum of 5 years and found no difference in radiographic loosening between glenoid types. However, they were also limited by a small sample size of 47 total patients.¹⁷ Likewise, Kilian et al.’s prospective randomized trial echoed these findings, showing no substantial difference in glenoid lucencies between the two designs.¹³ Papadonikolakis et al. conducted a systematic review of 3853 TSA and found that the annual rate of radiolucent lines was 7.3 %.²⁰ However, the rate of symptomatic glenoid loosening per year was determined to be 1.2 %, and the rate of surgical revision due to glenoid issues was just 0.8 % per year. Keeled glenoid components were associated with a higher incidence of asymptomatic radiolucent lines compared to pegged components. Furthermore, the study observed no significant evidence of a decrease in the rate of symptomatic loosening over time, suggesting that this remains a persistent challenge in shoulder arthroplasty despite advances in surgical techniques and implant design. Based on these results, Papadonikolakis et al. suggest that the clinical impact of findings such as radiolucency may be limited. However, radiolucencies could be an early sign of eventual symptomatic component loosening that is poorly studied due to the follow-up required to capture such events.

4.4 Revision

The mean follow-up duration was 8.9 years for the keeled glenoid group and 9.2 years for the pegged glenoid group. Only one keeled glenoid (8.3 %) underwent revision which was due to glenoid loosening. This low revision rate is consistent with findings from Fox et al., who conducted a retrospective review of 302 cemented-keeled TSAs.¹⁰ They reported that despite the occurrence of radiographic changes—particularly notable changes emerging after five years—the rate of revision remained low. Specifically, they observed glenoid component survival rates of 99 % at five years and 93 % at ten years.

4.5 Limitations

This study acknowledges several limitations inherent in its design and methodology. Firstly, the retrospective nature of this single-surgeon, single-institution study may limit the generalizability of the findings. The span of surgical dates from 2001 to 2015 means that advancements in surgical techniques, cementing processes, and implant designs over time may have influenced the earlier outcomes within our dataset. Additionally, the choice of implant type was guided by surgeon preference, glenoid size, intraoperative assessment, and availability. These factors could have introduced bias into the results. The relatively small sample size further constrains the statistical power of the study, possibly obscuring subtle distinctions between the groups. Moreover, the follow-up duration may not have been sufficient to detect significant radiographic changes or differences in revision rates, which studies have shown may take at least five to ten years to manifest significantly.¹⁰

Despite these constraints, the study is strengthened by the use of matched cohort groups, which provide valuable insights into the performance of keeled and pegged glenoid designs. To build on the findings presented here and to strengthen the validity of the results, future research should aim to conduct prospective studies with larger sample sizes across multiple centers.