Clinical Outcomes of Anatomic Total Shoulder Arthroplasty

Clinical Outcomes of Anatomic Total Shoulder Arthroplasty

Ryan W. Simovitch, MD

Brent Mollon, MD, MSc


Anatomic total shoulder arthroplasty (ATSA) involves the surgical replacement of the entire glenohumeral articulation with components designed to recreate native joint structure and kinematics. ATSA replaces the proximal humerus with a cobalt-chromium or titanium alloy component as in hemiarthroplasty (HA), but resurfaces the glenoid with a polyethylene component. This implant demands the presence of an intact and functional rotator cuff to appropriately balance and power shoulder movements. As a result, rotator cuff compromise (due to irreparable rotator cuff tears [RCTs], proximal humerus fractures, or functional attrition of the rotator cuff musculature) or situations where appropriate shoulder balance cannot be achieved (as in significant glenoid or humeral erosions) are being preferentially managed with reverse total shoulder arthroplasty (RTSA).

The first anatomic total shoulder prosthesis was designed by a dentist and implanted by French surgeon Jules Pean in 1893 for tuberculosis of the shoulder.1 This implant, composed of a paraffin-coated ebonite glenoid and a leather/platinum humeral component, was ultimately removed 2 years later due to chronic infection.2 Modern shoulder arthroplasty design is credited to Neer, who introduced a vitallium HA design to treat displaced proximal humerus fractures in the 1950s.3 In 1976, ATSA became available after Stellbrink designed a polyethylene glenoid component to couple with the humeral heads designed by Neer.2,4 ATSA has since evolved to become the standard of care for advanced glenohumeral arthritis with an intact rotator cuff and has been consistently found to offer superior clinical results to HA.5,6

ATSA is indicated in patients with glenohumeral arthritis who have failed nonoperative treatments or as a revision procedure for ongoing pain following shoulder HA. Contraindications include rotator cuff incompetency, severe bony abnormality, active infection, deltoid palsy, and inability to comply with postoperative instructions.

Fourth-Generation ATSA Design

While initial designs of ATSA prostheses consisted of monobloc humeral components and cemented all-polyethylene glenoid components, as was the case in Neer’s initial design, modern fourth-generation implants have evolved significantly to provide additional options. Variability on the humeral and glenoid side allows multiple combinations to aid surgeons in providing their patients with an “anatomic” reconstruction. The prevalent thought of shoulder surgeons is that a more “anatomic reconstruction” will result in more durable and profound clinical improvements after surgery.

The factors that can be adjusted during the humeral reconstruction aspect of ATSA are head height, head thickness, head diameter, center of rotation (COR), neck-shaft angle, and humeral head offset.7 In the initial monobloc humeral prostheses, the position of the humeral head was dictated by the stem position in the humeral intramedullary canal. Current ATSA prostheses now utilize modular components. The head and stem can be mated on the back table or in situ. Most systems utilize heads of varying diameters and thicknesses with eccentricity built into the head or head and stem, allowing the best anatomic coverage of the resected head independent of the position of the stem in the intramedullary canal. Double eccentricity has been shown to provide an anatomic reconstruction more often than single eccentricity, allowing medial and posterior offset to be corrected simultaneously and independently. The goal of modularity is to allow a humeral implant to restore the native humeral head size and COR in order to avoid overstuffing of the joint, which can result in loss of range of motion (ROM), attrition of the rotator cuff, edge loading of the glenoid, and subacromial impingement. A previous study by Flurin et al8 demonstrated that improved anatomic reconstruction (optimizing humeral head height, humeral head centering, humeral head medial offset, humeral head diameter, and humeral neck angle) had a statistically significant positive effect on clinical outcomes.

Other aspects of the fourth-generation humeral component include methods of fixation as well as the shape and length of the stem. Fixation of humeral stems has largely evolved away from cementation, as methods of porous coating for fixation of the metaphyseal portion of the stem have been shown to provide reliable on-growth and in-growth. Fixation has predominantly
been designed to be metaphyseal and not diaphyseal so as to reduce the rate of stress shielding. Furthermore, the trend has been away from large stems that result in three-point fixation in the diaphyseal bone, also in order to avoid stress shielding and difficult revision scenarios. Most recently, the concept of short stems and stemless humeral components has been introduced in an effort to further reduce stress shielding, minimize instrumentation of the intramedullary canal, and provide for an easier revision.

Similar to the evolution of humeral components in ATSA, glenoid implants have also dramatically evolved. Older generations of ATSA relied on pegged or keeled all-polyethylene components requiring cement fixation. These components fared well but demonstrated high rates of radiolucency over time. Fourth-generation prostheses now include hybrid polyethylene glenoid components capable of achieving initial fixation with cement but durable fixation with biologic on-growth, in-growth, and through growth (FIGURE 21.1). Current glenoids are manufactured from ultrahigh molecular weight polyethylene that is highly cross-linked. Vitamin E9 may be increasingly introduced to reduce oxidation as it has been shown to be effective in the lower extremity literature and early studies pertaining to the shoulder. Current ATSA systems rely on a radial mismatch (difference between radius of curvature of the humeral head and glenoid concave surface), with low or no mismatch having been shown to result in a higher incidence of radiolucent lines and loosening, whereas a higher mismatch can result in edge loading and instability.10 A recent study by Schoch et al11 has challenged the initial thinking that radial mismatch is optimized if kept between 6 and 10 mm as Walch et al12 reported, noting that glenoid loosening as well as glenoid revision rates did not differ with radial mismatches between 3.4 and 7 mm.11 Thus, radial mismatches below 5.5 mm may be acceptable and even desirable.

Trends in ATSA

The advent of fourth-generation prostheses has expanded the indications for ATSA. Modularity and innovation allow the implant to be adapted to a patient’s anatomy, as opposed to adapting the individual patient anatomy to a fixed prosthesis, permitting surgeons to push the envelope in younger patients and patients with severe deformities. However, the frequency of utilization of ATSA has changed with the advent of the increasing popularity and durable outcomes of RTSA. In 2014, the use of RTSA surpassed the use of ATSA and this trend has continued.13 The global shoulder replacement market has been estimated to increase at a compound annual growth rate of 7%, but RTSA growth is dwarfing that of ATSA. Despite the rapid growth of RTSA, ATSA will still account for a significant portion of the market share as HA has largely fallen out of favor for most conditions.


Improvements in surgical technique, patient selection, and implant design have culminated in overall ATSA implant survivorship rates of 95% to 98% at 5 years, 93% to 97% at 10 years, 85% to 88% at 15 years, and 80% to 85% at 20 years.14,15,16,17 Survivorship tends to be impacted by glenoid loosening, humeral stem loosening, infection, and rotator cuff dysfunction.

Overall survivorship has been most clearly linked to glenoid implant design. Historically, glenoid components could be broadly grouped into uncemented metal-backed polyethylene or cemented all-polyethylene components. All-polyethylene components could be further grouped according to how they are secured into the glenoid, either via a large central keel or through multiple pegs (both centrally and peripherally). Although overall glenoid survivorship across all implants is believed to be 96%, 96%, and 95% at 5 years, 10 years, and 15 years,18 respectively, variability does exist across
component types. Independent of component type, it is recognized that an overall higher risk of failure has been identified in male patients and in those for whom the indication for surgery was posttraumatic osteoarthritis or osteonecrosis.18

The lowest rates of glenoid implant survivorship have been reported with uncemented metal-backed glenoid components, and this type of implant is now primarily of historical significance for that reason. Early randomized controlled trials compared these metal-backed implants to a cemented keeled all-polyethylene component and reported a 20% (4/20) loosening rate as well as a 15% revision rate in the metal-backed group compared with none of these complications in the all-polyethylene group. Despite this, the rate of radiolucent lines was much higher in the cemented component group at the time of final follow-up (85% compared with 25%).19 Subsequent studies have reported failure rates of 5%, 11% and 54% at 5 years, 7.5 years, and 12 years, respectively, for metal-backed polyethylene glenoids.15,20 Gauci et al21 studied a cohort of patients younger than 60 years and noted a threefold greater rate of failure at 10.3 years with metal-backed components compared with cemented all-polyethylene components (70% vs 22%). Reported modes of failure in the above studies included loosening, polyethylene dissociation from the metal baseplate, and severe polyethylene wear. Although some studies have reported favorable outcomes15 of newer generation metal-backed glenoids and current use of hybrid implants (distinct from historical metal-backed design) show promise,22 cemented polyethylene components are still considered the current gold standard in ATSA.

In contrast to glenoid prostheses, humeral stem loosening is uncommon. According to a summary of 22 arthroplasty series published since 1980, humeral loosening occurred in only 4 of 1183 shoulders (0.3%) studied.23 While the factors associated with humeral component failure include patient factors (young age, posttraumatic osteoarthritis, male gender) and the use of metal-backed glenoids,22 no differences in failure rates between cemented and uncemented modern humeral components24 have been noted.

Although a prerequisite for ATSA is an intact and functional rotator cuff, the integrity of the rotator cuff can change over time. The onset of rotator cuff dysfunction over time following ATSA has been termed secondary rotator cuff dysfunction. Young et al,25 in a multicenter study examining 518 ATSAs performed for osteoarthritis with over 5-year follow-up, determined that the rate of survivorship free of secondary rotator cuff dysfunction was 100% at 5 years, 84% at 10 years, and 45% at 15 years. This is a common reason for revision of ATSA to RTSA.

It is possible that humeral component design may indirectly lead to glenoid failure due to nonanatomic prosthesis placement and improper soft tissue balance, resulting in edge loading as well as increased joint reaction forces. As a result, short and stemless humeral components are being evaluated for their ability to recreate humeral offset and restore COR while facilitating ease of revision surgery if required in the future.24 Nonetheless, short stem and stemless humeral components represent the newest trends in ATSA design, and it remains to be seen if they have similar survivorship rates to standard length stems at longer-term follow-up.26


The remaining portion of this chapter will explore the clinical and radiographic results of ATSA prostheses and the factors that influence outcome. Some of the most common clinical metrics will be listed and briefly described, but a more in-depth analysis of these can be found in Chapter 27.

Patient-reported outcome measures (PROMs) that are often cited in the literature include the Simple Shoulder Test (SST), the American Shoulder and Elbow Surgeons (ASES) questionnaire, the Shoulder and Pain Disability Index (SPADI), the Oxford Shoulder Score (OSS), a pain Visual Analog Score (VAS), Single Assessment Numeric Evaluation (SANE), and Subjective Shoulder Value (SSV). These metrics utilize questions that are administered to the patient but do not have an examiner objective component. They are designed to allow a patient’s clinical course to be longitudinally evaluated over time.

Other clinical metrics combine patient-reported measures and function with an examiner’s objective findings such as ROM and strength. These include but are not limited to the University of California at Los Angeles (UCLA) score and the Constant score.

ROM assessment is an important metric that allows an objective comparison of outcomes. However, there is no standardized method for collecting ROM, so the variability of technique can influence direct comparison of outcome studies within the literature. In the future, digitized ROM assessment utilizing smart devices will likely impact the accuracy of these measurements in the clinical setting.

Other data collection tools that help to compare a cohort’s improvement over time and also compare different cohorts within a single study as well as between studies are Health-Related Quality of Life (HRQOL) surveys. These include but are not limited to PROMIS-10, SF-36, SF-12, and VR-12. A more detailed explanation can also be found in Chapter 27.


Historically, the clinical results of outcome studies have been reported by showing differences in means and recording statistical significance, generally shown as a P-value. Statistical significance denotes that a difference is not due to random chance. However, statistical significance does not equate to what a patient sees
as a meaningful clinical change. One method to report whether or not a difference or improvement is clinically relevant to a patient is comparing it to the calculated minimal clinically important difference (MCID) or substantial clinical benefit (SCB) values. The concepts of MCID and SCB are explained in more detail in Chapter 27.

Values for MCID and SCB of common clinical metrics have previously been reported for ATSA (TABLE 21.1).27,28,29 Of note, studies by Simovitch et al27,28 and Werner et al29 both evaluated the MCID and SCB values for ASES after ATSA in distinct cohorts of patients and determined values that were nearly identical. MCID and SCB values after ATSA have been demonstrated to be higher than the values after RTSA. Furthermore, these values appear to be impacted by gender, length of follow-up, age at the time of surgery, and preoperative function.27,28,29 MCID and SCB values can be used as additional means by which to evaluate clinical improvement following ATSA as opposed to solely statistical significance that can be influenced by a myriad of factors including sample size.


The majority of available clinical and radiographic data on ATSA pertains to a “standard” length modular humeral stem. As a result, it remains the most robustly studied in terms of clinical and radiographic results. For the purposes of the following discussion, we will consider the glenoid component in ATSA to be an all-polyethylene design unless otherwise stated.

Clinical Results of ATSA With Standard Stem

ATSA has been demonstrated to result in good to excellent clinical results in the context of implant survival with rates of 80% to 85% at 20 years as noted above.17 Multiple series have documented statistically significant improvement in ROM, PROMs, and improved VAS pain scores in approximately 90% of patients or greater.30,31,32,33,34 The authors reviewed a prospectively collected multicenter international database that utilizes a single-platform ATSA system with dual eccentricity (Equinoxe, Exactech, Gainesville, FL). After examining the results of 752 ATSAs with a minimum of 5-year follow-up (mean 87 ± 29 [SD] months), we determined a 94% satisfaction rate at the most recent follow-up. There was a statistically significant increase in ROM and PROMs (ASES, SST, Constant, UCLA, SPADI) scores as well as reduction in pain VAS scores from preoperative to postoperative time points. All-cause revision rate in this analysis was 5.6%.

Utilizing the same database percentage but applying the MCID and SCB value for metric analysis, Simovitch et al28,29 noted that 92.7% of patients achieved MCID for ASES versus 79.5% achieving for SCB; 94.7% achieved MCID for the Constant score versus 84.9% achieving SCB; 90.8% achieved MCID for UCLA score versus 81.3% achieving SCB; 92% achieved MCID for SST score versus 81.7% achieving SCB; 91.5% achieved MCID for SPADI versus 73.4% achieving SCB; and 88.9% achieved MCID for pain VAS versus 71.6% achieving SCB. ROM was similar with 82.4% achieving MCID for active abduction versus 65.3% achieving SCB; 79.5% achieving MCID for active forward flexion
versus 62% achieving SCB; and 81.5% achieving MCID for active external rotation compared with 69.2% achieving SCB.

Radiographic Results of ATSA With Standard Stem

Radiographic evaluation of standard-stem humeral components has identified long-term humeral loosening rates to be as low as 0% for cemented prosthesis and as low as 1.5% for press-fit stems.18,35,36,37 Despite this, the development of radiolucent lines and medial calcar osteolysis over time has been well documented. In uncemented press-fit standard-length stems, stress shielding of the medial calcar occurs as force is transferred distally to the metaphyseal/diaphyseal region of the humerus, leading to resorption of the bone in the unloaded proximal metaphyseal region. In contrast, osteolysis is also felt to occur through the generation of polyethylene wear particles from the glenoid, leading to bone resorption through a phagocytic response to these particles over time. A study by Verborgt et al38 of 37 uncemented press-fit standard-length humeral components followed up over a mean of 9.2 years demonstrated radiolucency in 59%, with 14% demonstrating tilting not visualized on initial postoperative radiographs. Another study by Cole et al39 evaluated 47 press-fit standard-length humeral stems utilized for ATSA with a minimum of 5-year follow-up and found a 43% rate of medial calcar osteolysis. The largest such study includes ATSA and HA, but similarly identified a 63% rate of stress shielding in uncemented stems and a 43% rate of proximal humeral osteolysis over a mean follow-up of 8.2 years.40 Despite radiographic changes, an association between clinical metrics or ROM and the above radiographic findings has not been consistent and may be more related to glenoid component failure.38,39,40 However, these radiographic changes reflect bone loss occurring and this deficiency can impact ease of revision surgery, need for bone grafting or bulk prostheses, and function after revision.

The rate of stress shielding likely differs with varying implant characteristics and technique of insertion. Stress shielding may be impacted by stem length, proximal coating, implant shape, absence of a collar, and the canal fill ratio. Nagels et al41 examined a cohort of patients with and without stress shielding and noted that a higher canal filling ratio (ratio of stem width to the humeral intramedullary canal width) was associated with a higher rate of stress shielding. While the impact of humeral stress shielding on clinical results is not entirely understood, it is an undesirable result, potentially resulting in implant instability and complicated revisions because of bone loss.

Modularity in Current ATSA Designs

The humeral stem and head modular components work in tandem to reconstruct the humeral side of an ATSA. It is intuitive that a more anatomic reconstruction will result in improved clinical outcomes. Concerns that a humeral head component which fails to accurately reconstruct the COR and native offset will translate to poorer clinical outcomes have been evaluated in two case-control studies to date. Utilizing five radiographic parameters to obtain an Anatomic Reconstruction Index (ARI), Flurin et al8 evaluated 49 ATSA patients with an average follow-up of 9.1 years. While the authors identified that more anatomic reconstructions trended toward better clinical results, the relatively low number of patients that were deemed to have a poor ARI ultimately limited the strength of their findings. In contrast, Chalmers et al42 evaluated radiographic parameters in 95 patients with ATSA and a mean follow-up of 4.3 ± 1.7 years and found no statistical relationship with clinical outcomes in their series. While the bulk of literature pertains to the ability to achieve anatomic reconstruction in ATSA, there remains a paucity of literature regarding its effect on clinical outcomes.


Due to concerns about stress shielding43,44 as well as an effort to reduce blood loss,45 decrease operative time,45 limit instrumentation of the humeral intramedullary canal, and preserve bone if revision is needed, short stem and stemless humeral implants have been developed and are gaining in popularity. Early results are promising. Nonetheless, standard-length stems have performed well clinically with excellent long-term survivorship, so adoption of short stem and stemless components should proceed cautiously while long-term data are collected.

The clinical results of short stem and stemless ATSA have been favorable and similar to studies with standard-length stems. ATSA with these components has resulted in dramatic improvements of PROMs, pain, and ROM when utilized in patients with osteoarthritis.46,47

Finite Element Analysis of Short Stem and Stemless ATSA

Improvements in radiographic results are anticipated based on an expected reduction in stress shielding. Razfar et al43 conducted a finite element analysis study examining proximal humeral bone stresses between stemless, short stem, and standard-length stem models. They noted that a reduction in stem length resulted in a progressive increase in proximal cortical stresses with the stemless model approximating the intact state of the proximal humerus. In addition, they found that stresses in the proximal trabecular bone increased with the use of a short stem or stemless device and that the stemless model demonstrated trabecular stresses that well exceeded the intact state. These findings, based on our understanding of bone remodeling, suggest that the frequency of stress shielding should decrease with short stem and stemless implants compared with their standard-length counterparts.

Radiographic Outcomes of Short Stem ATSA

The incidence of stress shielding has been examined in short stem humeral implants utilized for ATSA. Casagrande et al48 reported a high rate of humeral bone adaptations with a first-generation short stem implant (Ascend Monolithic; Wright Medical, Memphis, TN, USA) in a cohort of 73 patients including calcar osteolysis (17%), radiolucency (71%), and condensation lines (19%). They determined that 10% of the stems were loose and 9% were “at risk”; ultimately, 12% were revised due to aseptic loosening. Another study utilizing the same prosthesis reported a high rate of humeral bone adaptations (52%). However, a subsequent study by Schnetzke et al46 examined a cohort of patients undergoing ATSA with a second-generation short stem (Ascend Flex, Wright Medical, Memphis, TN, USA) modified to have a proximal plasma spray and different geometry, and this yielded a lower rate of bone adaptations (29%) and no stem loosening. Another study by Morwood et al49 compared the first-generation Ascend Monolithic (Wright Medical, Memphis, TN, USA) with the second-generation Ascend Flex (Wright Medical, Memphis, TN, USA) and noted a dramatic reduction in humeral stem radiolucencies and stems at risk for loosening. A separate study that examined a cohort of patients with the Ascend Monolithic (Wright Medical, Memphis, TN, USA) and Ascend Flex (Wright Medical, Memphis, TN, USA) compared to a cohort with an Apex short stem (Arthrex, Naples, FL, USA) demonstrated that despite a higher filling ratio, the Apex (Arthrex, Naples, FL, USA) had a significantly lower rate of bone adaptations. These studies indicate that no one design or patient factor controls the propensity for stress shielding and bone adaptations. Rather it is a complex interplay of design parameters, patient anatomy, and implantation techniques. Multiple short stems have entered the market, many of which have been designed to maximize the ability to avoid stress shielding.

Radiographic Results of Stemless ATSA

Experience with stemless ATSA devices has more closely followed the predictions of bone adaptation according to the finite element analysis by Razfar et al.43 Multiple studies with midterm and long-term follow-up have demonstrated a humeral lucency rate between 0% and 2.3% with the absence of signs of stress shielding or osteolysis.47,50,51 These components hold promise to dramatically reduce bone adaptations and have demonstrated robust clinical improvement similar to ATSA with standard humeral stems. One concern voiced about stemless humeral components in ATSA is potentially their inability to recreate the patient’s normal anatomy because of a lack of eccentricity. However, a radiographic study by Gallacher et al52 demonstrated that a stemless device was able to recreate the humeral head size within 2 mm of normal anatomy and the COR within 3 mm of normal anatomy in 76% of the studied patients. Furthermore, in this series, the incidence of radiolucent lines was only 2.6%.


While the design of cemented all-polyethylene peg and keel glenoid components has evolved to include convex-backed components and partially cemented designs to achieve central peg or cage bony ingrowth, the clinical superiority of these designs continue to be debated. Radiographically, though, they show favorable results.

Polyethylene Keel Versus Peg Design

Keeled and pegged implants have primarily been compared by rates of radiolucent line formation at the bone/cement interface. Early concerns resulted from studies demonstrating high rates of radiolucent lines on the immediate postoperative radiographs of keeled designs (compared with pegged designs)53 and were further supported by randomized controlled trials with follow-up radiographs at 2 years.54,55 It remains unclear if the presence of immediate or the later development of radiolucencies translates into negative outcomes or clinical failure. In fact, overall survival rates for cemented keeled glenoid components have been favorable, with implant survival of 95% at 10 years and 92% at 15 years averaged across all keeled designs.18 The long-term data for pegged implants are less robust, with the longest series reported by McLendon et al56 demonstrating comparable survivorship (time to revision) of pegged compared with keeled implants at a mean 5-year follow-up.18 However, this cohort of pegged polyethylene components demonstrated survival free of radiographic failure of 92% at 5 years and only 43% at 10 years. It is important to note that this study evaluated the Cofield 2 pegged prosthesis (Smith and Nephew), whose three inline peg configurations differ from more modern designs utilizing peripheral pegs placed in a nonlinear configuration. Aside from concerns for radiolucency, keeled components require more contiguous bone removal for implantation and hence may complicate revision compared with the standard nonlinear peripheral pegs in modern all-polyethylene glenoid implants.

Polyethylene Hybrid Glenoid

Pegged glenoid components have seen a more recent evolution toward a “biologic hybrid” design. In this design, a central peg or cage has an interface for bony ingrowth, with three smaller peripheral cemented pegs used to achieve initial stability. The proposed method of bone ingrowth for some implants that utilize a polyethylene three-dimensional surface for ingrowth is unclear, though others utilize porous three-dimensional tantalum
or titanium surfaces with grit blast or plasma spray that encourages bone fixation. Regardless of specific hybrid design, short- to medium-term radiographic studies have been favorable. For example, radiographic follow-up across multiple designs has demonstrated bone ingrowth around the central peg to be between 68% and 91%, with osteolysis around the central peg to range from 7% to 25% in short- to medium-term follow-up.53,57,58 A recent study by Friedman et al59 compared a hybrid polyethylene glenoid with a central titanium grit blast cage and an all-polyethylene pegged glenoid component in age, gender, and length of follow-up matched study (FIGURE 21.1). They found that the caged hybrid glenoid had significantly less (P < 0.05) radiolucent glenoid lines (9% vs 38%), aseptic glenoid loosening (1.3% vs 3.8%), glenoid revision (2.5% vs 6.9%), and associated radiolucent humeral lines (3% vs 9.1%).59 Biologic fixation with hybrid components may result in improved glenoid survivorship.

In contrast to the robustness of radiographic studies, there is a paucity of clinical studies comparing keeled, pegged, and hybrid cemented components. The published survivorship data that exist do not demonstrate clinical superiority of any one design across comparable timeframes.23 With the outcomes of the majority of modern generation implants being limited to medium-term follow-up, longer-term follow-up of modern cemented polyethylene and hybrid biologic designs is required to truly compare the clinical outcomes of the two designs. There is evidence that the overall risk of glenoid component failure is higher in male patients as well as in ATSA performed for posttraumatic osteoarthritis or osteonecrosis.18

Only gold members can continue reading. Log In or Register to continue

Jun 23, 2022 | Posted by in ORTHOPEDIC | Comments Off on Clinical Outcomes of Anatomic Total Shoulder Arthroplasty
Premium Wordpress Themes by UFO Themes