Wiater JM

Shoulder arthroplasty can be associated with significant perioperative blood loss, with the overall rate of postoperative allogeneic blood transfusion ranging from 7.4% to 43%.^1-6 Blood transfusions are associated with a range of health risks.⁷ Soft-tissue dissection and cutting and reaming of bone surfaces can be sources of significant blood loss. Directly visualized sources of bleeding can be addressed using standard surgical hemostasis, including electrocautery, suture ligation, compression, and careful avoidance of vascular structures. However, difficult-to-visualize areas and bony sources of bleeding are more difficult to manage.

Numerous products for mitigating perioperative blood loss are commercially available. Topical hemostatic agents have been used in many surgical specialties, including orthopedic surgery, cardiothoracic surgery, neurosurgery, vascular surgery, and general surgery.^8-10 In orthopedic surgery, use of topical thrombin- and fibrin-based products as hemostatic agents has been studied in knee and hip arthroplasty, with varying results.^11-14 Early studies have shown reduced blood loss and postoperative transfusion rates with use of a fibrin sealant or fibrin tissue adhesive,^11,12,15 whereas others have shown no significant benefit of using these hemostatic agents. Massin and colleagues¹⁴ found no difference in blood loss in the setting of total knee arthroplasty (TKA) with use of a fibrin sealant. In a 2012 prospective study, Kim and colleagues¹³ also showed no significant reduction in blood loss in patients treated with a topical thrombin-based hemostatic agent in TKA.

Surgiflo (Ethicon) is a hemostatic matrix that is combined with a topical human thrombin solution before sterile application. The matrix consists of an absorbable porcine gelatin powder that provides a structure for platelet adhesion and aggregation.¹⁶ When used in combination with thrombin, it aids in fibrin clot formation, leading to hemostasis of oozing blood and minor bleeding from small capillaries and venules. According to the manufacturer’s data, it can halt bleeding in less than 2 minutes and retains its efficacy for up to 8 hours.

To our knowledge, there are no reports of studies on use of topical fibrin- or thrombin-based hemostatic agents in shoulder arthroplasty. We conducted a study to investigate perioperative blood loss, transfusion rates, and complications during the hospital stays of patients who underwent shoulder arthroplasty and were treated with or without the Surgiflo topical hemostatic agent. Our hypothesis was that patients intraoperatively treated with this agent would have significantly less perioperative blood loss and lower transfusion rates without increased rates of in-hospital complications.

Patients and Methods

We retrospectively reviewed data from 211 consecutive shoulder arthroplasties performed by Dr. J. Michael Wiater between December 2012 and August 2013. All primary and revision anatomical and reverse total shoulder arthroplasty (TSA) procedures were included. Patients with a preoperative diagnosis of acute fracture, and patients with a diagnosis of any type of blood diathesis, including anemia and platelet disorders that lead to excessive clotting or bleeding, were excluded. Patients treated between May 2013 and August 2013 had the hemostatic matrix applied to the soft tissues before final wound closure. Chart review for any exclusion criteria left 102 patients in the experimental (hemostatic agent) group and 98 patients in the control group.

For all patients, any anticoagulation or anti-inflammatory medication was discontinued 1 week before the elective arthroplasty. An interscalene regional block combined with general anesthesia was used in all cases. All procedures were performed through a standard anterior deltopectoral approach. Patients in the experimental group had 10 mL of the hemostatic agent topically applied to the soft tissues of the wound before closure. Half the mixture (5 mL) was applied to the deep tissues of the axillary recess, subacromial, and joint spaces, and the other half was applied superficially after closure of the deltopectoral interval. A medium Hemovac (Zimmer) drain was used in all cases, with 1 tubing placed in the deep space and another between the deltoid and the skin, both draining to a single drain evacuator.

After surgery, all patients received deep venous thrombosis (DVT) prophylaxis consisting of 5000 units of subcutaneous unfractionated heparin every 8 hours until discharge, and then aspirin 325 mg twice daily for 2 weeks after discharge unless contraindicated. Any long-term anticoagulation therapy discontinued before surgery was resumed on postoperative day 2 (POD 2). All drains were removed on POD 2 unless they had more than 50 mL of output over an 8-hour period. Complete blood cell counts were collected for all patients before surgery and on PODs 1 and 2. Whether to transfuse blood was based on clinical judgment of severe or symptomatic acute blood loss anemia; however, no strict predetermined criteria were followed.

Patient electronic medical records were reviewed for demographic information, including age, sex, height, weight, comorbidities, American Society of Anesthesiologists (ASA) physical status, and preoperative anticoagulation use. Anesthesia records were reviewed for intraoperative estimated blood loss (EBL) and intraoperative autologous blood return (Cell Saver, Haemonetics). Patient laboratory results were reviewed for preoperative and postoperative hemoglobin (Hb) and hematocrit levels. Electronic medical records were also reviewed for incidence of transfusion and any major or minor complications occurring within 90 days of the procedure. All data were collected and reviewed under the approval of the human investigations committee at our institution.

Hemoglobin loss and hidden blood loss (HBL) were calculated as described by Good and colleagues.¹⁷ Total Hb loss was estimated using the total blood volume formula described by Nadler and colleagues.¹⁸ Difference between preoperative Hb level and final Hb level recorded during hospital stay was corrected for units of blood transfused (estimate, 52 g of Hb per unit). Hemoglobin loss was then used to calculate total blood loss, and total drain output was added to total blood loss to determine HBL. These formulas were used:

Hb_loss = Blood Volume (L) × [Hb_initial (g/L) – Hb_final (g/L)] + Hb_transfused

Total Blood Loss (mL) = 1000 × Hb_loss/Hb_initial

HBL (mL) = Total Blood Loss (mL) + Total Drain Output (mL)

All statistical analyses were performed using SPSS Statistics Version 20 (IBM). A Shapiro-Wilk test was used to test for normality. All variables collected were compared between the experimental and control cohorts. For continuous variables, independent t test was used to compare normal data, and the Mann-Whitney rank sum test was used for non-normal data. Categorical variables were compared with the Fisher exact test for 2×2 tables and with the χ² test for larger tables. In all tests, P < .05 was considered statistically significant.

Results

The experimental and control cohorts were demographically similar with respect to age, sex, body mass index (BMI), ASA status, and home anticoagulation treatment (Table 1). Patients who received preoperative anticoagulation therapy were evenly distributed between the 2 patient groups (P = .745). Thirty-five patients in the experimental group and 39 in the control group were taking aspirin. In addition, in the experimental group, 5 patients were taking warfarin, 4 clopidogrel, 1 dabigatran, and 1 prasugrel. In the control group, 6 patients were taking warfarin, 3 clopidogrel, 2 dabigatran, and 1 rivaroxaban. Type of arthroplasty (primary anatomical, primary reverse, revision shoulder arthroplasty) was also evenly distributed (P = .256), and operative time did not vary significantly between cohorts (P = .518).

Markers of operative blood loss were also compared between patient groups (Table 2). There was no significant difference in intraoperative EBL or cell saver volume between cohorts (Ps = .301 and .800). Drain output on PODs 1 and 2 did not differ between cohorts (Ps = .789 and .777); the same was true for total postoperative drain output (P = .906). Hemoglobin levels did vary significantly between groups before surgery (P = .002) and on PODs 1 and 2 (Ps = .027 and .005), with the experimental group having a lower mean Hb level at each time point. Mean Hb loss, however, did not vary significantly (P = .253). There was also no difference in HBL between cohorts (P = .601), the calculation of which accounts for patient height and weight, Hb loss, and transfusions. The incidence of transfusion was 25% in the experimental group and 20% in the control group—not a statistically significant difference (P = .407). Mean (SD) number of transfused units of packed red blood cells was 0.54 (1.05) in the experimental group and 0.40 (0.91) in the control group—again, not a statistically significant difference (P = .377).

Preoperative Hb level under 13 g/dL has been reported as a risk factor for transfusion after surgery.¹⁹ To account for the significantly lower Hb level in the experimental group, we examined the incidence of transfusion in patients with preoperative Hb levels above and below this cutoff. Among patients with preoperative Hb levels under 13 g/dL, transfusion incidence was 45.8% (experimental group) and 42.9% (control group) (P > .99); among those with preoperative Hb levels above 13 g/dL, transfusion incidence was 7.7% (experimental) and 11.1% (control) (P = .760).

To account for reportedly higher blood loss and transfusion rates in revision cases,^1,2,20 we stratified our data by primary and revision cases, comparing them within the entire patient cohort and comparing the experimental and control groups within these subsets. Tables 3 and 4 list the results. Revision cases had more EBL (P < .001), autologous blood return (P < .001), drain output on POD 1 (P = .025), and total drain output (P = .002). There was no significant difference in transfusion rate between primary (22.2%) and revision (27.3%) cases (P = .505) or when the experimental and control groups were compared within primary and revision subsets. Among primary cases, transfusion rates were 23% (experimental) and 21.2% (control) (P = .853); among revision cases, rates were 35% (experimental) and 15% (control) (P = .263). Revisions showed a significant (P = .043) difference in HBL between the experimental and control groups, with more blood loss in the experimental group. EBL and autologous blood return were equivocal. Hb levels and drain outputs were statistically different only for POD 2, but there was no difference between overall Hb loss or total drain outputs. Among primary cases, no parameters of blood loss were statistically significantly different. The significantly lower preoperative and postoperative Hb levels were again seen in the experimental group.

The groups’ complication rates were comparable, and there was no significant risk associated with use of the hemostatic agent (P = .764). In each group, there were no complications that would be of particular concern with use of this agent. These complications included wound complications, deep prosthesis infection, and systemic thromboembolic disease (eg, myocardial infarction, stroke, DVT, pulmonary embolus). Nine patients (5 control, 4 experimental) had minor medical complications, and 2 (1 control, 1 experimental) had major medical complications. The control group’s 5 minor medical complications were acute kidney infection treated with antibiotics (1 patient), persistent urinary retention requiring Foley catheter for short period after discharge (1), minor upper gastrointestinal bleed treated medically (1), recalcitrant tachycardia in setting of chronic atrial fibrillation (1), and vasovagal syncope with no identified cardiovascular cause or periprosthetic complication (1); the control patient with the major medical complication died 2 weeks after surgery, after discharge to the inpatient rehabilitation unit. This death was secondary to pneumonia, sepsis, and eventual multisystem organ failure. The experimental group’s 4 minor medical complications were urinary retention requiring catheterization for short period (1 patient), urinary tract infections diagnosed 2 weeks after surgery and treated with antibiotics (2), and new-onset atrial fibrillation treated medically (1); the experimental patient with the major medical complication developed Takotsubo cardiomyopathy, a nonischemic stress-induced weakening of the myocardium requiring medical management. An experimental patient also had reverse TSA shoulder dislocation 12 days after surgery—thought to be caused by inadequate soft-tissue tension and unrelated to hemostatic agent use. The patient was returned to the operating room for polyethylene liner exchange and metallic spacer implantation.

Discussion

Reported rates of transfusion after shoulder arthroplasty have ranged from 7.4% to 43%, when including revision and reverse TSAs.^2,3 In the present study, the overall transfusion rate was 23% (includes patients who underwent primary or revision shoulder arthroplasties with anatomical or reverse prostheses). Although the risk for complications is low, serious issues may arise with blood transfusions. Allogeneic blood transfusions can cause fluid overload, allergic reactions, fever, acute immune hemolytic reaction, transfusion-related acute lung injury (TRALI), bloodborne infections, and formation of antibodies complicating any future need for transfusions.⁷ According to the National Heart, Lung, and Blood Institute, the chances of becoming infected from transfusion are 1 in 2 million for the hepatitis C and human immunodeficiency viruses and 1 in 205,000 for the hepatitis B virus.⁷ Some studies have also found higher rates of infection after hip or knee arthroplasty in patients who received allogeneic blood transfusions.^21,22 In addition, for hospitals, transfusion costs are significant. One study showed that direct and indirect overhead costs amounted to $522 to $1183 per red blood cell unit.²³ Given the risks and costs associated with blood transfusions, use of an effective intraoperative blood loss management agent could be beneficial in the setting of shoulder arthroplasty.

The use and efficacy of intraoperative blood management agents remain controversial. Numerous agents for managing perioperative blood loss are commercially available. Previous clinical studies have shown variable results with use of topical hemostatic agents, but not in the setting of shoulder arthroplasty.²⁴ In 1999, Levy and colleagues¹¹ showed that use of fibrin tissue adhesive reduced blood loss and postoperative transfusion rates in patients who underwent TKA. In 2001, Wang and colleagues¹⁵ showed that using a fibrin sealant in TKA reduced bloody drainage and maintained higher Hb levels. In 2003, the same group showed that use of fibrin sealant also reduced perioperative blood loss in total hip arthroplasty.¹² More recent studies have had contradicting results,^13,14 similar to ours. A 2012 prospective study failed to show any significant difference in blood loss after TKA in patients treated with a topical thrombin-based hemostatic agent.¹³ The authors did find significantly higher Hb values in the treated group on PODs 1 and 2, though the drain outputs and transfusion rates did not differ.

To our knowledge, the present study is the first to evaluate use of a topical hemostatic agent during shoulder arthroplasty. We did not find a significant difference in perioperative blood loss with application of Surgiflo, a topical thrombin-based hemostatic agent. Interestingly, we found that Hb levels both before surgery and on PODs 1 and 2 were significantly lower in the experimental group. However, the difference was about 0.7 g/dL, which would not be clinically significant. The lower Hb levels on PODs 1 and 2 likely resulted from lower preoperative levels.

Other studies have found higher transfusion rates for revision versus primary shoulder arthroplasty.^1,2,20 In our series, EBL, autologous blood return, and drain output were higher overall for revision versus primary cases. When we stratified by primary and revision cases, we could not detect a difference in transfusion rates between the experimental and control groups. The lack of significant difference in the revision group could be caused by low statistical power, as the control group had only 13 revision cases. Having more patients in the study may have revealed a larger difference in blood loss with use of the hemostatic agent in revision cases.

We also found no significant increase in adverse events related to use of the hemostatic agent. Complications of particular concern would include wound complications, deep prosthesis infection, and systemic thromboembolic disease (eg, myocardial infarction, stroke, DVT, pulmonary embolus). There were no statistical differences in major and minor complications between the groups and no identifiable complications related to the hemostatic agent used.

Our results should be viewed in light of study limitations. First, with this retrospective study, we relied heavily on the accuracy of computer-based patient documentation. In addition, blood loss estimates are imperfect regardless of measurement technique. Intraoperative EBL is often determined by the surgeon and is highly variable, and autologous blood collection does not account for blood lost in operative sponges, instruments, and irrigation. To minimize this issue, we tried to assess perioperative blood loss through multiple data points, including intraoperative EBL, autologous blood returned during surgery, drain output, transfusion rates, and HBL calculations. Also, blood transfusion criteria depend on the physician’s clinical assessment and decision making, as well as patient condition, which could certainly add variability to the transfusion rate between groups. Another limitation is that the procedures studied were not homogeneous, and including primary and revision anatomical and reverse shoulder arthroplasties may have added variability to the results. In this single-surgeon study, however, we were able to ensure that the same standard techniques and hemostasis were applied in all procedures. Last, given the relatively small sample used, more patients may be needed to reveal a significant and clinically relevant difference in blood loss.

Conclusion

Perioperative blood loss poses serious risks to patient health. In light of the varying findings in the literature and the cost of transfusions and blood loss management products, use of these hemostatic agents remains controversial. In the present study, we found no significant difference in perioperative blood loss or transfusion rates with use of a hemostatic agent during shoulder arthroplasty. Therefore, we cannot conclude that this agent is effective for blood loss management in shoulder arthroplasty. Highly powered prospective studies are needed to confirm our findings.

Shoulder arthroplasty can be associated with significant perioperative blood loss, with the overall rate of postoperative allogeneic blood transfusion ranging from 7.4% to 43%.^1-6 Blood transfusions are associated with a range of health risks.⁷ Soft-tissue dissection and cutting and reaming of bone surfaces can be sources of significant blood loss. Directly visualized sources of bleeding can be addressed using standard surgical hemostasis, including electrocautery, suture ligation, compression, and careful avoidance of vascular structures. However, difficult-to-visualize areas and bony sources of bleeding are more difficult to manage.

Numerous products for mitigating perioperative blood loss are commercially available. Topical hemostatic agents have been used in many surgical specialties, including orthopedic surgery, cardiothoracic surgery, neurosurgery, vascular surgery, and general surgery.^8-10 In orthopedic surgery, use of topical thrombin- and fibrin-based products as hemostatic agents has been studied in knee and hip arthroplasty, with varying results.^11-14 Early studies have shown reduced blood loss and postoperative transfusion rates with use of a fibrin sealant or fibrin tissue adhesive,^11,12,15 whereas others have shown no significant benefit of using these hemostatic agents. Massin and colleagues¹⁴ found no difference in blood loss in the setting of total knee arthroplasty (TKA) with use of a fibrin sealant. In a 2012 prospective study, Kim and colleagues¹³ also showed no significant reduction in blood loss in patients treated with a topical thrombin-based hemostatic agent in TKA.

Surgiflo (Ethicon) is a hemostatic matrix that is combined with a topical human thrombin solution before sterile application. The matrix consists of an absorbable porcine gelatin powder that provides a structure for platelet adhesion and aggregation.¹⁶ When used in combination with thrombin, it aids in fibrin clot formation, leading to hemostasis of oozing blood and minor bleeding from small capillaries and venules. According to the manufacturer’s data, it can halt bleeding in less than 2 minutes and retains its efficacy for up to 8 hours.

To our knowledge, there are no reports of studies on use of topical fibrin- or thrombin-based hemostatic agents in shoulder arthroplasty. We conducted a study to investigate perioperative blood loss, transfusion rates, and complications during the hospital stays of patients who underwent shoulder arthroplasty and were treated with or without the Surgiflo topical hemostatic agent. Our hypothesis was that patients intraoperatively treated with this agent would have significantly less perioperative blood loss and lower transfusion rates without increased rates of in-hospital complications.

Patients and Methods

We retrospectively reviewed data from 211 consecutive shoulder arthroplasties performed by Dr. J. Michael Wiater between December 2012 and August 2013. All primary and revision anatomical and reverse total shoulder arthroplasty (TSA) procedures were included. Patients with a preoperative diagnosis of acute fracture, and patients with a diagnosis of any type of blood diathesis, including anemia and platelet disorders that lead to excessive clotting or bleeding, were excluded. Patients treated between May 2013 and August 2013 had the hemostatic matrix applied to the soft tissues before final wound closure. Chart review for any exclusion criteria left 102 patients in the experimental (hemostatic agent) group and 98 patients in the control group.

For all patients, any anticoagulation or anti-inflammatory medication was discontinued 1 week before the elective arthroplasty. An interscalene regional block combined with general anesthesia was used in all cases. All procedures were performed through a standard anterior deltopectoral approach. Patients in the experimental group had 10 mL of the hemostatic agent topically applied to the soft tissues of the wound before closure. Half the mixture (5 mL) was applied to the deep tissues of the axillary recess, subacromial, and joint spaces, and the other half was applied superficially after closure of the deltopectoral interval. A medium Hemovac (Zimmer) drain was used in all cases, with 1 tubing placed in the deep space and another between the deltoid and the skin, both draining to a single drain evacuator.

After surgery, all patients received deep venous thrombosis (DVT) prophylaxis consisting of 5000 units of subcutaneous unfractionated heparin every 8 hours until discharge, and then aspirin 325 mg twice daily for 2 weeks after discharge unless contraindicated. Any long-term anticoagulation therapy discontinued before surgery was resumed on postoperative day 2 (POD 2). All drains were removed on POD 2 unless they had more than 50 mL of output over an 8-hour period. Complete blood cell counts were collected for all patients before surgery and on PODs 1 and 2. Whether to transfuse blood was based on clinical judgment of severe or symptomatic acute blood loss anemia; however, no strict predetermined criteria were followed.

Patient electronic medical records were reviewed for demographic information, including age, sex, height, weight, comorbidities, American Society of Anesthesiologists (ASA) physical status, and preoperative anticoagulation use. Anesthesia records were reviewed for intraoperative estimated blood loss (EBL) and intraoperative autologous blood return (Cell Saver, Haemonetics). Patient laboratory results were reviewed for preoperative and postoperative hemoglobin (Hb) and hematocrit levels. Electronic medical records were also reviewed for incidence of transfusion and any major or minor complications occurring within 90 days of the procedure. All data were collected and reviewed under the approval of the human investigations committee at our institution.

Hemoglobin loss and hidden blood loss (HBL) were calculated as described by Good and colleagues.¹⁷ Total Hb loss was estimated using the total blood volume formula described by Nadler and colleagues.¹⁸ Difference between preoperative Hb level and final Hb level recorded during hospital stay was corrected for units of blood transfused (estimate, 52 g of Hb per unit). Hemoglobin loss was then used to calculate total blood loss, and total drain output was added to total blood loss to determine HBL. These formulas were used:

Hb_loss = Blood Volume (L) × [Hb_initial (g/L) – Hb_final (g/L)] + Hb_transfused

Total Blood Loss (mL) = 1000 × Hb_loss/Hb_initial

HBL (mL) = Total Blood Loss (mL) + Total Drain Output (mL)

All statistical analyses were performed using SPSS Statistics Version 20 (IBM). A Shapiro-Wilk test was used to test for normality. All variables collected were compared between the experimental and control cohorts. For continuous variables, independent t test was used to compare normal data, and the Mann-Whitney rank sum test was used for non-normal data. Categorical variables were compared with the Fisher exact test for 2×2 tables and with the χ² test for larger tables. In all tests, P < .05 was considered statistically significant.

Results

The experimental and control cohorts were demographically similar with respect to age, sex, body mass index (BMI), ASA status, and home anticoagulation treatment (Table 1). Patients who received preoperative anticoagulation therapy were evenly distributed between the 2 patient groups (P = .745). Thirty-five patients in the experimental group and 39 in the control group were taking aspirin. In addition, in the experimental group, 5 patients were taking warfarin, 4 clopidogrel, 1 dabigatran, and 1 prasugrel. In the control group, 6 patients were taking warfarin, 3 clopidogrel, 2 dabigatran, and 1 rivaroxaban. Type of arthroplasty (primary anatomical, primary reverse, revision shoulder arthroplasty) was also evenly distributed (P = .256), and operative time did not vary significantly between cohorts (P = .518).

Markers of operative blood loss were also compared between patient groups (Table 2). There was no significant difference in intraoperative EBL or cell saver volume between cohorts (Ps = .301 and .800). Drain output on PODs 1 and 2 did not differ between cohorts (Ps = .789 and .777); the same was true for total postoperative drain output (P = .906). Hemoglobin levels did vary significantly between groups before surgery (P = .002) and on PODs 1 and 2 (Ps = .027 and .005), with the experimental group having a lower mean Hb level at each time point. Mean Hb loss, however, did not vary significantly (P = .253). There was also no difference in HBL between cohorts (P = .601), the calculation of which accounts for patient height and weight, Hb loss, and transfusions. The incidence of transfusion was 25% in the experimental group and 20% in the control group—not a statistically significant difference (P = .407). Mean (SD) number of transfused units of packed red blood cells was 0.54 (1.05) in the experimental group and 0.40 (0.91) in the control group—again, not a statistically significant difference (P = .377).

Preoperative Hb level under 13 g/dL has been reported as a risk factor for transfusion after surgery.¹⁹ To account for the significantly lower Hb level in the experimental group, we examined the incidence of transfusion in patients with preoperative Hb levels above and below this cutoff. Among patients with preoperative Hb levels under 13 g/dL, transfusion incidence was 45.8% (experimental group) and 42.9% (control group) (P > .99); among those with preoperative Hb levels above 13 g/dL, transfusion incidence was 7.7% (experimental) and 11.1% (control) (P = .760).

To account for reportedly higher blood loss and transfusion rates in revision cases,^1,2,20 we stratified our data by primary and revision cases, comparing them within the entire patient cohort and comparing the experimental and control groups within these subsets. Tables 3 and 4 list the results. Revision cases had more EBL (P < .001), autologous blood return (P < .001), drain output on POD 1 (P = .025), and total drain output (P = .002). There was no significant difference in transfusion rate between primary (22.2%) and revision (27.3%) cases (P = .505) or when the experimental and control groups were compared within primary and revision subsets. Among primary cases, transfusion rates were 23% (experimental) and 21.2% (control) (P = .853); among revision cases, rates were 35% (experimental) and 15% (control) (P = .263). Revisions showed a significant (P = .043) difference in HBL between the experimental and control groups, with more blood loss in the experimental group. EBL and autologous blood return were equivocal. Hb levels and drain outputs were statistically different only for POD 2, but there was no difference between overall Hb loss or total drain outputs. Among primary cases, no parameters of blood loss were statistically significantly different. The significantly lower preoperative and postoperative Hb levels were again seen in the experimental group.

The groups’ complication rates were comparable, and there was no significant risk associated with use of the hemostatic agent (P = .764). In each group, there were no complications that would be of particular concern with use of this agent. These complications included wound complications, deep prosthesis infection, and systemic thromboembolic disease (eg, myocardial infarction, stroke, DVT, pulmonary embolus). Nine patients (5 control, 4 experimental) had minor medical complications, and 2 (1 control, 1 experimental) had major medical complications. The control group’s 5 minor medical complications were acute kidney infection treated with antibiotics (1 patient), persistent urinary retention requiring Foley catheter for short period after discharge (1), minor upper gastrointestinal bleed treated medically (1), recalcitrant tachycardia in setting of chronic atrial fibrillation (1), and vasovagal syncope with no identified cardiovascular cause or periprosthetic complication (1); the control patient with the major medical complication died 2 weeks after surgery, after discharge to the inpatient rehabilitation unit. This death was secondary to pneumonia, sepsis, and eventual multisystem organ failure. The experimental group’s 4 minor medical complications were urinary retention requiring catheterization for short period (1 patient), urinary tract infections diagnosed 2 weeks after surgery and treated with antibiotics (2), and new-onset atrial fibrillation treated medically (1); the experimental patient with the major medical complication developed Takotsubo cardiomyopathy, a nonischemic stress-induced weakening of the myocardium requiring medical management. An experimental patient also had reverse TSA shoulder dislocation 12 days after surgery—thought to be caused by inadequate soft-tissue tension and unrelated to hemostatic agent use. The patient was returned to the operating room for polyethylene liner exchange and metallic spacer implantation.

Discussion

Reported rates of transfusion after shoulder arthroplasty have ranged from 7.4% to 43%, when including revision and reverse TSAs.^2,3 In the present study, the overall transfusion rate was 23% (includes patients who underwent primary or revision shoulder arthroplasties with anatomical or reverse prostheses). Although the risk for complications is low, serious issues may arise with blood transfusions. Allogeneic blood transfusions can cause fluid overload, allergic reactions, fever, acute immune hemolytic reaction, transfusion-related acute lung injury (TRALI), bloodborne infections, and formation of antibodies complicating any future need for transfusions.⁷ According to the National Heart, Lung, and Blood Institute, the chances of becoming infected from transfusion are 1 in 2 million for the hepatitis C and human immunodeficiency viruses and 1 in 205,000 for the hepatitis B virus.⁷ Some studies have also found higher rates of infection after hip or knee arthroplasty in patients who received allogeneic blood transfusions.^21,22 In addition, for hospitals, transfusion costs are significant. One study showed that direct and indirect overhead costs amounted to $522 to $1183 per red blood cell unit.²³ Given the risks and costs associated with blood transfusions, use of an effective intraoperative blood loss management agent could be beneficial in the setting of shoulder arthroplasty.

The use and efficacy of intraoperative blood management agents remain controversial. Numerous agents for managing perioperative blood loss are commercially available. Previous clinical studies have shown variable results with use of topical hemostatic agents, but not in the setting of shoulder arthroplasty.²⁴ In 1999, Levy and colleagues¹¹ showed that use of fibrin tissue adhesive reduced blood loss and postoperative transfusion rates in patients who underwent TKA. In 2001, Wang and colleagues¹⁵ showed that using a fibrin sealant in TKA reduced bloody drainage and maintained higher Hb levels. In 2003, the same group showed that use of fibrin sealant also reduced perioperative blood loss in total hip arthroplasty.¹² More recent studies have had contradicting results,^13,14 similar to ours. A 2012 prospective study failed to show any significant difference in blood loss after TKA in patients treated with a topical thrombin-based hemostatic agent.¹³ The authors did find significantly higher Hb values in the treated group on PODs 1 and 2, though the drain outputs and transfusion rates did not differ.

To our knowledge, the present study is the first to evaluate use of a topical hemostatic agent during shoulder arthroplasty. We did not find a significant difference in perioperative blood loss with application of Surgiflo, a topical thrombin-based hemostatic agent. Interestingly, we found that Hb levels both before surgery and on PODs 1 and 2 were significantly lower in the experimental group. However, the difference was about 0.7 g/dL, which would not be clinically significant. The lower Hb levels on PODs 1 and 2 likely resulted from lower preoperative levels.

Other studies have found higher transfusion rates for revision versus primary shoulder arthroplasty.^1,2,20 In our series, EBL, autologous blood return, and drain output were higher overall for revision versus primary cases. When we stratified by primary and revision cases, we could not detect a difference in transfusion rates between the experimental and control groups. The lack of significant difference in the revision group could be caused by low statistical power, as the control group had only 13 revision cases. Having more patients in the study may have revealed a larger difference in blood loss with use of the hemostatic agent in revision cases.

We also found no significant increase in adverse events related to use of the hemostatic agent. Complications of particular concern would include wound complications, deep prosthesis infection, and systemic thromboembolic disease (eg, myocardial infarction, stroke, DVT, pulmonary embolus). There were no statistical differences in major and minor complications between the groups and no identifiable complications related to the hemostatic agent used.

Our results should be viewed in light of study limitations. First, with this retrospective study, we relied heavily on the accuracy of computer-based patient documentation. In addition, blood loss estimates are imperfect regardless of measurement technique. Intraoperative EBL is often determined by the surgeon and is highly variable, and autologous blood collection does not account for blood lost in operative sponges, instruments, and irrigation. To minimize this issue, we tried to assess perioperative blood loss through multiple data points, including intraoperative EBL, autologous blood returned during surgery, drain output, transfusion rates, and HBL calculations. Also, blood transfusion criteria depend on the physician’s clinical assessment and decision making, as well as patient condition, which could certainly add variability to the transfusion rate between groups. Another limitation is that the procedures studied were not homogeneous, and including primary and revision anatomical and reverse shoulder arthroplasties may have added variability to the results. In this single-surgeon study, however, we were able to ensure that the same standard techniques and hemostasis were applied in all procedures. Last, given the relatively small sample used, more patients may be needed to reveal a significant and clinically relevant difference in blood loss.

Conclusion

Perioperative blood loss poses serious risks to patient health. In light of the varying findings in the literature and the cost of transfusions and blood loss management products, use of these hemostatic agents remains controversial. In the present study, we found no significant difference in perioperative blood loss or transfusion rates with use of a hemostatic agent during shoulder arthroplasty. Therefore, we cannot conclude that this agent is effective for blood loss management in shoulder arthroplasty. Highly powered prospective studies are needed to confirm our findings.

Shoulder arthroplasty can be associated with significant perioperative blood loss, with the overall rate of postoperative allogeneic blood transfusion ranging from 7.4% to 43%.^1-6 Blood transfusions are associated with a range of health risks.⁷ Soft-tissue dissection and cutting and reaming of bone surfaces can be sources of significant blood loss. Directly visualized sources of bleeding can be addressed using standard surgical hemostasis, including electrocautery, suture ligation, compression, and careful avoidance of vascular structures. However, difficult-to-visualize areas and bony sources of bleeding are more difficult to manage.

Numerous products for mitigating perioperative blood loss are commercially available. Topical hemostatic agents have been used in many surgical specialties, including orthopedic surgery, cardiothoracic surgery, neurosurgery, vascular surgery, and general surgery.^8-10 In orthopedic surgery, use of topical thrombin- and fibrin-based products as hemostatic agents has been studied in knee and hip arthroplasty, with varying results.^11-14 Early studies have shown reduced blood loss and postoperative transfusion rates with use of a fibrin sealant or fibrin tissue adhesive,^11,12,15 whereas others have shown no significant benefit of using these hemostatic agents. Massin and colleagues¹⁴ found no difference in blood loss in the setting of total knee arthroplasty (TKA) with use of a fibrin sealant. In a 2012 prospective study, Kim and colleagues¹³ also showed no significant reduction in blood loss in patients treated with a topical thrombin-based hemostatic agent in TKA.

Surgiflo (Ethicon) is a hemostatic matrix that is combined with a topical human thrombin solution before sterile application. The matrix consists of an absorbable porcine gelatin powder that provides a structure for platelet adhesion and aggregation.¹⁶ When used in combination with thrombin, it aids in fibrin clot formation, leading to hemostasis of oozing blood and minor bleeding from small capillaries and venules. According to the manufacturer’s data, it can halt bleeding in less than 2 minutes and retains its efficacy for up to 8 hours.

To our knowledge, there are no reports of studies on use of topical fibrin- or thrombin-based hemostatic agents in shoulder arthroplasty. We conducted a study to investigate perioperative blood loss, transfusion rates, and complications during the hospital stays of patients who underwent shoulder arthroplasty and were treated with or without the Surgiflo topical hemostatic agent. Our hypothesis was that patients intraoperatively treated with this agent would have significantly less perioperative blood loss and lower transfusion rates without increased rates of in-hospital complications.

Patients and Methods

We retrospectively reviewed data from 211 consecutive shoulder arthroplasties performed by Dr. J. Michael Wiater between December 2012 and August 2013. All primary and revision anatomical and reverse total shoulder arthroplasty (TSA) procedures were included. Patients with a preoperative diagnosis of acute fracture, and patients with a diagnosis of any type of blood diathesis, including anemia and platelet disorders that lead to excessive clotting or bleeding, were excluded. Patients treated between May 2013 and August 2013 had the hemostatic matrix applied to the soft tissues before final wound closure. Chart review for any exclusion criteria left 102 patients in the experimental (hemostatic agent) group and 98 patients in the control group.

For all patients, any anticoagulation or anti-inflammatory medication was discontinued 1 week before the elective arthroplasty. An interscalene regional block combined with general anesthesia was used in all cases. All procedures were performed through a standard anterior deltopectoral approach. Patients in the experimental group had 10 mL of the hemostatic agent topically applied to the soft tissues of the wound before closure. Half the mixture (5 mL) was applied to the deep tissues of the axillary recess, subacromial, and joint spaces, and the other half was applied superficially after closure of the deltopectoral interval. A medium Hemovac (Zimmer) drain was used in all cases, with 1 tubing placed in the deep space and another between the deltoid and the skin, both draining to a single drain evacuator.

After surgery, all patients received deep venous thrombosis (DVT) prophylaxis consisting of 5000 units of subcutaneous unfractionated heparin every 8 hours until discharge, and then aspirin 325 mg twice daily for 2 weeks after discharge unless contraindicated. Any long-term anticoagulation therapy discontinued before surgery was resumed on postoperative day 2 (POD 2). All drains were removed on POD 2 unless they had more than 50 mL of output over an 8-hour period. Complete blood cell counts were collected for all patients before surgery and on PODs 1 and 2. Whether to transfuse blood was based on clinical judgment of severe or symptomatic acute blood loss anemia; however, no strict predetermined criteria were followed.

Patient electronic medical records were reviewed for demographic information, including age, sex, height, weight, comorbidities, American Society of Anesthesiologists (ASA) physical status, and preoperative anticoagulation use. Anesthesia records were reviewed for intraoperative estimated blood loss (EBL) and intraoperative autologous blood return (Cell Saver, Haemonetics). Patient laboratory results were reviewed for preoperative and postoperative hemoglobin (Hb) and hematocrit levels. Electronic medical records were also reviewed for incidence of transfusion and any major or minor complications occurring within 90 days of the procedure. All data were collected and reviewed under the approval of the human investigations committee at our institution.

Hemoglobin loss and hidden blood loss (HBL) were calculated as described by Good and colleagues.¹⁷ Total Hb loss was estimated using the total blood volume formula described by Nadler and colleagues.¹⁸ Difference between preoperative Hb level and final Hb level recorded during hospital stay was corrected for units of blood transfused (estimate, 52 g of Hb per unit). Hemoglobin loss was then used to calculate total blood loss, and total drain output was added to total blood loss to determine HBL. These formulas were used:

Hb_loss = Blood Volume (L) × [Hb_initial (g/L) – Hb_final (g/L)] + Hb_transfused

Total Blood Loss (mL) = 1000 × Hb_loss/Hb_initial

HBL (mL) = Total Blood Loss (mL) + Total Drain Output (mL)

All statistical analyses were performed using SPSS Statistics Version 20 (IBM). A Shapiro-Wilk test was used to test for normality. All variables collected were compared between the experimental and control cohorts. For continuous variables, independent t test was used to compare normal data, and the Mann-Whitney rank sum test was used for non-normal data. Categorical variables were compared with the Fisher exact test for 2×2 tables and with the χ² test for larger tables. In all tests, P < .05 was considered statistically significant.

Results

The experimental and control cohorts were demographically similar with respect to age, sex, body mass index (BMI), ASA status, and home anticoagulation treatment (Table 1). Patients who received preoperative anticoagulation therapy were evenly distributed between the 2 patient groups (P = .745). Thirty-five patients in the experimental group and 39 in the control group were taking aspirin. In addition, in the experimental group, 5 patients were taking warfarin, 4 clopidogrel, 1 dabigatran, and 1 prasugrel. In the control group, 6 patients were taking warfarin, 3 clopidogrel, 2 dabigatran, and 1 rivaroxaban. Type of arthroplasty (primary anatomical, primary reverse, revision shoulder arthroplasty) was also evenly distributed (P = .256), and operative time did not vary significantly between cohorts (P = .518).

Markers of operative blood loss were also compared between patient groups (Table 2). There was no significant difference in intraoperative EBL or cell saver volume between cohorts (Ps = .301 and .800). Drain output on PODs 1 and 2 did not differ between cohorts (Ps = .789 and .777); the same was true for total postoperative drain output (P = .906). Hemoglobin levels did vary significantly between groups before surgery (P = .002) and on PODs 1 and 2 (Ps = .027 and .005), with the experimental group having a lower mean Hb level at each time point. Mean Hb loss, however, did not vary significantly (P = .253). There was also no difference in HBL between cohorts (P = .601), the calculation of which accounts for patient height and weight, Hb loss, and transfusions. The incidence of transfusion was 25% in the experimental group and 20% in the control group—not a statistically significant difference (P = .407). Mean (SD) number of transfused units of packed red blood cells was 0.54 (1.05) in the experimental group and 0.40 (0.91) in the control group—again, not a statistically significant difference (P = .377).

Preoperative Hb level under 13 g/dL has been reported as a risk factor for transfusion after surgery.¹⁹ To account for the significantly lower Hb level in the experimental group, we examined the incidence of transfusion in patients with preoperative Hb levels above and below this cutoff. Among patients with preoperative Hb levels under 13 g/dL, transfusion incidence was 45.8% (experimental group) and 42.9% (control group) (P > .99); among those with preoperative Hb levels above 13 g/dL, transfusion incidence was 7.7% (experimental) and 11.1% (control) (P = .760).

To account for reportedly higher blood loss and transfusion rates in revision cases,^1,2,20 we stratified our data by primary and revision cases, comparing them within the entire patient cohort and comparing the experimental and control groups within these subsets. Tables 3 and 4 list the results. Revision cases had more EBL (P < .001), autologous blood return (P < .001), drain output on POD 1 (P = .025), and total drain output (P = .002). There was no significant difference in transfusion rate between primary (22.2%) and revision (27.3%) cases (P = .505) or when the experimental and control groups were compared within primary and revision subsets. Among primary cases, transfusion rates were 23% (experimental) and 21.2% (control) (P = .853); among revision cases, rates were 35% (experimental) and 15% (control) (P = .263). Revisions showed a significant (P = .043) difference in HBL between the experimental and control groups, with more blood loss in the experimental group. EBL and autologous blood return were equivocal. Hb levels and drain outputs were statistically different only for POD 2, but there was no difference between overall Hb loss or total drain outputs. Among primary cases, no parameters of blood loss were statistically significantly different. The significantly lower preoperative and postoperative Hb levels were again seen in the experimental group.

The groups’ complication rates were comparable, and there was no significant risk associated with use of the hemostatic agent (P = .764). In each group, there were no complications that would be of particular concern with use of this agent. These complications included wound complications, deep prosthesis infection, and systemic thromboembolic disease (eg, myocardial infarction, stroke, DVT, pulmonary embolus). Nine patients (5 control, 4 experimental) had minor medical complications, and 2 (1 control, 1 experimental) had major medical complications. The control group’s 5 minor medical complications were acute kidney infection treated with antibiotics (1 patient), persistent urinary retention requiring Foley catheter for short period after discharge (1), minor upper gastrointestinal bleed treated medically (1), recalcitrant tachycardia in setting of chronic atrial fibrillation (1), and vasovagal syncope with no identified cardiovascular cause or periprosthetic complication (1); the control patient with the major medical complication died 2 weeks after surgery, after discharge to the inpatient rehabilitation unit. This death was secondary to pneumonia, sepsis, and eventual multisystem organ failure. The experimental group’s 4 minor medical complications were urinary retention requiring catheterization for short period (1 patient), urinary tract infections diagnosed 2 weeks after surgery and treated with antibiotics (2), and new-onset atrial fibrillation treated medically (1); the experimental patient with the major medical complication developed Takotsubo cardiomyopathy, a nonischemic stress-induced weakening of the myocardium requiring medical management. An experimental patient also had reverse TSA shoulder dislocation 12 days after surgery—thought to be caused by inadequate soft-tissue tension and unrelated to hemostatic agent use. The patient was returned to the operating room for polyethylene liner exchange and metallic spacer implantation.

Discussion

Reported rates of transfusion after shoulder arthroplasty have ranged from 7.4% to 43%, when including revision and reverse TSAs.^2,3 In the present study, the overall transfusion rate was 23% (includes patients who underwent primary or revision shoulder arthroplasties with anatomical or reverse prostheses). Although the risk for complications is low, serious issues may arise with blood transfusions. Allogeneic blood transfusions can cause fluid overload, allergic reactions, fever, acute immune hemolytic reaction, transfusion-related acute lung injury (TRALI), bloodborne infections, and formation of antibodies complicating any future need for transfusions.⁷ According to the National Heart, Lung, and Blood Institute, the chances of becoming infected from transfusion are 1 in 2 million for the hepatitis C and human immunodeficiency viruses and 1 in 205,000 for the hepatitis B virus.⁷ Some studies have also found higher rates of infection after hip or knee arthroplasty in patients who received allogeneic blood transfusions.^21,22 In addition, for hospitals, transfusion costs are significant. One study showed that direct and indirect overhead costs amounted to $522 to $1183 per red blood cell unit.²³ Given the risks and costs associated with blood transfusions, use of an effective intraoperative blood loss management agent could be beneficial in the setting of shoulder arthroplasty.

The use and efficacy of intraoperative blood management agents remain controversial. Numerous agents for managing perioperative blood loss are commercially available. Previous clinical studies have shown variable results with use of topical hemostatic agents, but not in the setting of shoulder arthroplasty.²⁴ In 1999, Levy and colleagues¹¹ showed that use of fibrin tissue adhesive reduced blood loss and postoperative transfusion rates in patients who underwent TKA. In 2001, Wang and colleagues¹⁵ showed that using a fibrin sealant in TKA reduced bloody drainage and maintained higher Hb levels. In 2003, the same group showed that use of fibrin sealant also reduced perioperative blood loss in total hip arthroplasty.¹² More recent studies have had contradicting results,^13,14 similar to ours. A 2012 prospective study failed to show any significant difference in blood loss after TKA in patients treated with a topical thrombin-based hemostatic agent.¹³ The authors did find significantly higher Hb values in the treated group on PODs 1 and 2, though the drain outputs and transfusion rates did not differ.

To our knowledge, the present study is the first to evaluate use of a topical hemostatic agent during shoulder arthroplasty. We did not find a significant difference in perioperative blood loss with application of Surgiflo, a topical thrombin-based hemostatic agent. Interestingly, we found that Hb levels both before surgery and on PODs 1 and 2 were significantly lower in the experimental group. However, the difference was about 0.7 g/dL, which would not be clinically significant. The lower Hb levels on PODs 1 and 2 likely resulted from lower preoperative levels.

Other studies have found higher transfusion rates for revision versus primary shoulder arthroplasty.^1,2,20 In our series, EBL, autologous blood return, and drain output were higher overall for revision versus primary cases. When we stratified by primary and revision cases, we could not detect a difference in transfusion rates between the experimental and control groups. The lack of significant difference in the revision group could be caused by low statistical power, as the control group had only 13 revision cases. Having more patients in the study may have revealed a larger difference in blood loss with use of the hemostatic agent in revision cases.

We also found no significant increase in adverse events related to use of the hemostatic agent. Complications of particular concern would include wound complications, deep prosthesis infection, and systemic thromboembolic disease (eg, myocardial infarction, stroke, DVT, pulmonary embolus). There were no statistical differences in major and minor complications between the groups and no identifiable complications related to the hemostatic agent used.

Our results should be viewed in light of study limitations. First, with this retrospective study, we relied heavily on the accuracy of computer-based patient documentation. In addition, blood loss estimates are imperfect regardless of measurement technique. Intraoperative EBL is often determined by the surgeon and is highly variable, and autologous blood collection does not account for blood lost in operative sponges, instruments, and irrigation. To minimize this issue, we tried to assess perioperative blood loss through multiple data points, including intraoperative EBL, autologous blood returned during surgery, drain output, transfusion rates, and HBL calculations. Also, blood transfusion criteria depend on the physician’s clinical assessment and decision making, as well as patient condition, which could certainly add variability to the transfusion rate between groups. Another limitation is that the procedures studied were not homogeneous, and including primary and revision anatomical and reverse shoulder arthroplasties may have added variability to the results. In this single-surgeon study, however, we were able to ensure that the same standard techniques and hemostasis were applied in all procedures. Last, given the relatively small sample used, more patients may be needed to reveal a significant and clinically relevant difference in blood loss.

Conclusion

Perioperative blood loss poses serious risks to patient health. In light of the varying findings in the literature and the cost of transfusions and blood loss management products, use of these hemostatic agents remains controversial. In the present study, we found no significant difference in perioperative blood loss or transfusion rates with use of a hemostatic agent during shoulder arthroplasty. Therefore, we cannot conclude that this agent is effective for blood loss management in shoulder arthroplasty. Highly powered prospective studies are needed to confirm our findings.

Reverse shoulder arthroplasty (RSA) is a treatment option for a spectrum of diseases in shoulders with rotator cuff deficiency. There are distinct morphologic changes in the scapular and glenoid anatomy in patients with chronic rotator cuff tears.¹ A muscular imbalance that occurs in the joint as a result of rotator cuff deficiency leads to morphologic changes that eliminate the compressive forces that hold the humeral head against the glenoid.² RSA effectively stabilizes the glenohumeral joint in shoulders with deficient rotator cuffs.^3,4 In early work, Grammont proposed that the glenosphere center of rotation should be medialized (concentric to the central axis of the metaglene or baseplate) and lowered.⁵ Although the medialized center of rotation in Grammont prostheses decreases shear forces and improves the deltoid lever arm, it also tends to result in mechanical impingement between the superomedial aspect of the humeral polyethylene insert and the scapular neck—so-called inferior scapular notching.^6-9

Notching, which has been reported in 50% to 96% of patients who receive a Delta III prosthesis, typically appears within the first few months after surgery but may be seen as late as 14 months after surgery.^5,10-12 Postmortem studies have shown that notching corresponds with erosion of the inferior pole of the glenoid and scapular neck, thought to be caused by the polyethylene cup of the implant.¹³ Although some studies have found that notching stabilizes after 1 year, others have shown notching progressing for up to 4 years after surgery.^11,12,14 The clinical relevance of notching continues to be controversial, but notching has been associated with poorer clinical outcomes, polyethylene wear, and local osteolysis. Component loosening has also been reported with notching of grade 3 or more.^8,10 Ultimately, there is concern that scapular notching could progress, ultimately leading to late glenoid loosening and potentially catastrophic failure.

Scapular anatomy has become an area of increased focus in rotator cuff disorders and in effects on RSA biomechanics.⁹ Recent reports have described important scapular morphology variations that suggest more individualized adjustments are needed during RSA.^9,15 In addition, some investigators have reported that development of notching appears to depend on the height and inclination of the implanted glenoid component, where an inferior position of the glenosphere leads to less impingement and better range of motion.^8,16 Simovitch and colleagues⁸ found the angle between the glenosphere and scapular neck and the craniocaudal position of the glenosphere to be highly correlated with inferior notching. They combined these 2 parameters into a predictive algorithm that provides a guideline (notching index, <35) for prevention of notching.

We conducted a study to evaluate the scapular notching index as a predictive tool and to consider other factors that may be associated with scapular notching occurring with use of Grammont reverse replacement systems. We hypothesized that patients with a notching index of less than 35 would not develop notching and that patients with an index of more than 35 would have increased incidence and severity of notching.

Materials and Methods

Patients treated with RSA for painful cuff tear arthropathy or irreparable rotator cuff tear with pseudoparesis (inability to actively elevate shoulder >90° in presence of free passive anterior elevation) were included in this retrospective review. All patients were treated between 2006 and 2010 by 1 of 2 established senior shoulder subspecialty surgeons. Patients treated with a Delta (DePuy Orthopaedics, Warsaw, Indiana) or an Aequalis (Tornier, Edina, Minnesota) reverse shoulder implant were included in the study. A standard polyethylene liner was used for all patients. These prostheses have the same neck shaft angle, 155º, as they have similar geometric designs, both based on the Grammont design—semiconstrained inverted with a fixed, lowered, medialized center of rotation. Standard instrumentation was used for all procedures. Patients were excluded if any nonstandard techniques or components were used (constrained or high-mobility liner, glenoid bone grafting). Patients who underwent revision for a previous reverse total arthroplasty, a total shoulder arthroplasty, or a hemiarthroplasty, or for treatment of acute fracture, posttraumatic deformity, or posttraumatic arthritis, were also excluded from our analyses. Minimum follow-up for study inclusion was 24 months.

All procedures were performed with the patient in the semi-beach-chair position and with use of a deltopectoral approach. The glenoid was prepared such that minimal reaming was needed to preserve the subchondral plate. The glenoid baseplate was positioned in the recommended inferior position to minimize notching and optimize functional outcomes.¹³ After surgery, all patients were managed with a simple soft immobilizer with or without a pillow with the arm at the patient’s side in internal rotation. Immediate passive mobilization was begun under the direction of physical therapists. Passive and active-assisted exercises were continued with gradual progression to independent activities of daily living at 6 weeks. Clinical evaluations were performed before and after surgery by the operating surgeon or independent research nurse. Active forward flexion, passive external rotation, strength, and visual analog scale (VAS) pain scores were reviewed and recorded. Case-specific complications were also reviewed.

Preoperative and postoperative anteroposterior radiographs were evaluated by 2 independent observers (attending surgeon, junior resident). Per standard technique, each radiograph was positioned horizontal to the scapular plane. Of the 91 patients, 66 had preoperative shoulder radiographs of acceptable quality, with complete visualization of scapular morphology. Radiographs were reviewed to measure the scapular neck angle (SNA), inferior scapular notching, prosthesis–scapular neck angle (PSNA), and peg glenoid rim distance (PGRD) (Figure 1). For the 66 patients with acceptable preoperative radiographs, SNA was determined by subtracting preoperative SNA from postoperative PSNA. Postoperative anteroposterior radiographs were used to classify degree of inferior scapular notching based on the Nerot grading scale (0-4). In addition, glenosphere overhang and glenosphere inclination were measured on postoperative radiographs.

The 91 shoulders were sorted into 2 groups based on degree of scapular notching: group 1, Nerot grade 0 (no inferior notching) and grade 1, and group 2, Nerot grades 2, 3, and 4. Group 1 had 37 patients with a size 36 glenosphere, 3 patients with size 38, and 8 patients with size 42; group 2 had 34 patients with a size 36 glenosphere, 1 patient with size 38, and 8 patients with size 42. All measurements were normalized to account for differences in glenosphere size. Groups 1 and 2 were compared on each radiographic parameter (inferior scapular notching, PSNA, PGRD, SNA).

Notching indexes were calculated ([PSNA × 0.13] + PGRD) and compared with the suggested index of 35.⁸ Simovitch and colleagues⁸ demonstrated that a notching index of more than 35 had 91% sensitivity and 88% specificity in predicting inferior notching, whereas a notching index of 35 or less avoided inferior notching 91% of the time. In this study, notching index was calculated for each patient, and then the mean values of groups 1 and 2 were compared (Table 1).

The effect of scapular notching and other individual radiographic parameters on outcomes was also evaluated with respect to forward flexion, external rotation, VAS pain score, complications, and external rotation lag sign. Mann-Whitney U test was used to test these variables; Spearman rank test was performed to determine correlation between each variable and scapular notching; logistic regression was used to explore the relationship of variables (PGRD, PSNA) as predictors of Nerot degree of inferior scapular notching, and postoperative complications; and independent-samples t test was used to determine group differences for each variable. For each investigation, the level of significance was set at P < .05. A biostatistician performed all statistical analyses using SPSS Version 19 (IBM, Armonk, New York).

Results

Our study cohort consisted of 91 shoulders. Mean follow-up was 41.8 months (range, 24.0-80.8 months). Seventy-five (82%) of the 91 shoulders developed scapular notching. Mean (SD) SNA on preoperative radiographs, used to assess preoperative scapular morphology, was 103.9° (14.5°). For all 91 shoulders, mean (SD) PSNA was 125.6° (16°), and mean (SD) PGRD was 16 (5.4) mm (Table 1). Inclination measurements were available for 86 patients. Mean (SD) inclination from 90° was 2.5° (10.3°) (range, 21°-30°). Mean (SD) SNA (postoperative PSNA minus preoperative SNA) for the 66 patients with acceptable preoperative radiographs was 24.3° (21.3°) (Table 1). Forty-eight of the 91 shoulders were placed in scapular notching group 1 (16 grade-0 shoulders, 32 grade-1 shoulders); the other 43 shoulders were placed in group 2 (33 grade-2 shoulders, 9 grade-3 shoulders, 1 grade-4 shoulder). Mean follow-up was 40 months for group 1 and 43 months for group 2.

There were no significant differences between groups 1 and 2 in SNA (102.8° vs 105.4°; P = .3), PGRD (15.4 vs 16.8 mm; P = . 47), or PSNA (125.8° vs 125.4°; P = .82) (Table 1). In addition, groups 1 and 2 had no significant differences (P > .05) in glenoid overhang and glenosphere inclination (other possible factors influencing notching).

Mean (SD) notching index was 31.8 (4.4) for group 1 and 33.1 (7.2) for group 2. These values were not significantly different (P = .29) (Table 1, Figure 2).⁸ Each was below the recommended threshold of 35 for prevention of notching (Table 1, Figure 2).

To try to understand why mean scapular notching index was low for both groups, we examined the contributing factors individually. Our cohort’s mean PGRD of 16.1 mm (15.4 and 16.8 mm for groups 1 and 2, respectively) was significantly lower than the cohort mean reported by Simovitch and colleagues⁸ (Table 2). Given that PGRD is more strongly weighted in the originally described notching index ([PSNA × 0.13] + PGRD),⁸ it was the primary driver for our notching index results, even though on average our results demonstrated a PSNA higher than that found by Simovitch and colleagues⁸ (Table 2; Figures 3, 4). Analyzing PGRD and PSNA together, we found no relationship between these variables and increased severity of inferior notching (Figure 5).

Regarding the effects of notching severity on outcomes in our study cohort, there were no significant differences between groups 1 and 2 in postoperative function, including forward flexion (123° vs 112.4°; P = .11), external rotation (18.8° vs 16.7°; P = .76), positive lag sign (P = .2), and VAS pain scores (1.2 vs 2.1; P = .15). There were also no significant differences between groups in the rate of complications (P = .92). Regression analysis determined that PSNA, PGRD, glenosphere inclination, glenosphere overhang, and implant manufacturer were not significant predictors of complications.

Discussion

RSA has provided good pain relief and restored function in patients with irreparable rotator cuff disease associated with arthritis.^5,12,17,18 Scapular notching is a complex, multifactorial process. Nevertheless, surgeons remain cautious about the implications of inferior scapular notching, which is being reported by a significant number of patients. Our cohort’s high incidence of scapular notching (82%) in the early postoperative period clearly highlights the importance of predictive models, such as the notching index.⁸ Although concerns about consequences of notching have been expressed, notching severity did not affect outcomes or increase complications in this cohort.^{5,8,11,12,17-19}

We conducted this study to examine use of a predictive tool for scapular notching, the notching index, in a large cohort of patients who underwent primary RSA. This index combines 2 well-established factors that contribute to notching—craniocaudal position and PSNA—into a predictive formula based on statistical analyses performed in a prospective cohort study.^4,5,8,12,18 In their clinical study, Simovitch and colleagues⁸ found that both craniocaudal position and PSNA were tightly coupled with inferior scapular notching, and they developed a notching index that accounts for this relationship. We hypothesized that patients with a notching index of less than the recommended 35 would not develop notching and that patients with a notching index of more than 35 would have increased incidence and severity of notching. With our cohort, the recommended index of 35 was not an appropriate threshold predictive of notching. Furthermore, the 35 threshold applied to our cohort had 89% sensitivity and 21% specificity in predicting notching. Although the sensitivity is high, and correctly predicted true instances of notching, the low specificity compromises the precision of the notching formula ([PSNA × 0.13] + PGRD).

From the formula, it can be inferred that higher PSNA values can be compensated for by decreasing PGRD and inferiorizing the glenosphere. However, this recommendation appears limited based on increasing PSNA values, as in our cohort. The previously described notching formula cannot be universally applied to all patients treated with RSA because of the complexity of this relationship and patient-specific anatomy.

 We assessed other possible anatomical and surgical factors, specific to scapular morphology, that could contribute to scapular notching. In other studies, reaming that produced an inferior tilt of the glenoid increased the likelihood of inferior notching.^8,20,21 Furthermore, we expected less inferior glenoid overhang and smaller glenosphere would predispose patients to more notching.^8,12,19 In our cohort, notching grade was not correlated with inferior tilt, glenoid overhang, or glenosphere size, which may be attributed to minimal variability in glenosphere size and a small range of glenosphere overhang.

There were limitations to this study. We examined only 2 types of RSA systems, and they had very similar Grammont designs. Other RSA designs might not have similar shortcomings with respect to inferior notching. In addition, we examined patient cases at a single time point and did not evaluate the effect of notching over time.

Overall, our results suggest that PGRD and PSNA have little effect on development of higher grade notching, particularly with use of Grammont prostheses. With newer surgical techniques, the recommendation is for inferior craniocaudal placement of the glenosphere, but this may not prevent notching with some types of patient-specific scapular morphology. Clearer surgical guidelines and techniques may help delineate the contribution of each parameter causing inferior scapular notching. Surgeons must weigh the evidence to determine how to correct patient-specific glenoid pathology and orient the glenosphere. Recent studies on bony increased-offset reverse shoulder arthroplasty (bio-RSA) techniques or newer prosthetic designs that considerably alter PSNA and the center of rotation may prevent inferior notching and provide a promising alternative to Grammont designs. Ultimately, longer follow-up is also needed to understand the clinical relevance of increased scapular notching.

Reverse shoulder arthroplasty (RSA) is a treatment option for a spectrum of diseases in shoulders with rotator cuff deficiency. There are distinct morphologic changes in the scapular and glenoid anatomy in patients with chronic rotator cuff tears.¹ A muscular imbalance that occurs in the joint as a result of rotator cuff deficiency leads to morphologic changes that eliminate the compressive forces that hold the humeral head against the glenoid.² RSA effectively stabilizes the glenohumeral joint in shoulders with deficient rotator cuffs.^3,4 In early work, Grammont proposed that the glenosphere center of rotation should be medialized (concentric to the central axis of the metaglene or baseplate) and lowered.⁵ Although the medialized center of rotation in Grammont prostheses decreases shear forces and improves the deltoid lever arm, it also tends to result in mechanical impingement between the superomedial aspect of the humeral polyethylene insert and the scapular neck—so-called inferior scapular notching.^6-9

Notching, which has been reported in 50% to 96% of patients who receive a Delta III prosthesis, typically appears within the first few months after surgery but may be seen as late as 14 months after surgery.^5,10-12 Postmortem studies have shown that notching corresponds with erosion of the inferior pole of the glenoid and scapular neck, thought to be caused by the polyethylene cup of the implant.¹³ Although some studies have found that notching stabilizes after 1 year, others have shown notching progressing for up to 4 years after surgery.^11,12,14 The clinical relevance of notching continues to be controversial, but notching has been associated with poorer clinical outcomes, polyethylene wear, and local osteolysis. Component loosening has also been reported with notching of grade 3 or more.^8,10 Ultimately, there is concern that scapular notching could progress, ultimately leading to late glenoid loosening and potentially catastrophic failure.

Scapular anatomy has become an area of increased focus in rotator cuff disorders and in effects on RSA biomechanics.⁹ Recent reports have described important scapular morphology variations that suggest more individualized adjustments are needed during RSA.^9,15 In addition, some investigators have reported that development of notching appears to depend on the height and inclination of the implanted glenoid component, where an inferior position of the glenosphere leads to less impingement and better range of motion.^8,16 Simovitch and colleagues⁸ found the angle between the glenosphere and scapular neck and the craniocaudal position of the glenosphere to be highly correlated with inferior notching. They combined these 2 parameters into a predictive algorithm that provides a guideline (notching index, <35) for prevention of notching.

We conducted a study to evaluate the scapular notching index as a predictive tool and to consider other factors that may be associated with scapular notching occurring with use of Grammont reverse replacement systems. We hypothesized that patients with a notching index of less than 35 would not develop notching and that patients with an index of more than 35 would have increased incidence and severity of notching.

Materials and Methods

Patients treated with RSA for painful cuff tear arthropathy or irreparable rotator cuff tear with pseudoparesis (inability to actively elevate shoulder >90° in presence of free passive anterior elevation) were included in this retrospective review. All patients were treated between 2006 and 2010 by 1 of 2 established senior shoulder subspecialty surgeons. Patients treated with a Delta (DePuy Orthopaedics, Warsaw, Indiana) or an Aequalis (Tornier, Edina, Minnesota) reverse shoulder implant were included in the study. A standard polyethylene liner was used for all patients. These prostheses have the same neck shaft angle, 155º, as they have similar geometric designs, both based on the Grammont design—semiconstrained inverted with a fixed, lowered, medialized center of rotation. Standard instrumentation was used for all procedures. Patients were excluded if any nonstandard techniques or components were used (constrained or high-mobility liner, glenoid bone grafting). Patients who underwent revision for a previous reverse total arthroplasty, a total shoulder arthroplasty, or a hemiarthroplasty, or for treatment of acute fracture, posttraumatic deformity, or posttraumatic arthritis, were also excluded from our analyses. Minimum follow-up for study inclusion was 24 months.

All procedures were performed with the patient in the semi-beach-chair position and with use of a deltopectoral approach. The glenoid was prepared such that minimal reaming was needed to preserve the subchondral plate. The glenoid baseplate was positioned in the recommended inferior position to minimize notching and optimize functional outcomes.¹³ After surgery, all patients were managed with a simple soft immobilizer with or without a pillow with the arm at the patient’s side in internal rotation. Immediate passive mobilization was begun under the direction of physical therapists. Passive and active-assisted exercises were continued with gradual progression to independent activities of daily living at 6 weeks. Clinical evaluations were performed before and after surgery by the operating surgeon or independent research nurse. Active forward flexion, passive external rotation, strength, and visual analog scale (VAS) pain scores were reviewed and recorded. Case-specific complications were also reviewed.

Preoperative and postoperative anteroposterior radiographs were evaluated by 2 independent observers (attending surgeon, junior resident). Per standard technique, each radiograph was positioned horizontal to the scapular plane. Of the 91 patients, 66 had preoperative shoulder radiographs of acceptable quality, with complete visualization of scapular morphology. Radiographs were reviewed to measure the scapular neck angle (SNA), inferior scapular notching, prosthesis–scapular neck angle (PSNA), and peg glenoid rim distance (PGRD) (Figure 1). For the 66 patients with acceptable preoperative radiographs, SNA was determined by subtracting preoperative SNA from postoperative PSNA. Postoperative anteroposterior radiographs were used to classify degree of inferior scapular notching based on the Nerot grading scale (0-4). In addition, glenosphere overhang and glenosphere inclination were measured on postoperative radiographs.

The 91 shoulders were sorted into 2 groups based on degree of scapular notching: group 1, Nerot grade 0 (no inferior notching) and grade 1, and group 2, Nerot grades 2, 3, and 4. Group 1 had 37 patients with a size 36 glenosphere, 3 patients with size 38, and 8 patients with size 42; group 2 had 34 patients with a size 36 glenosphere, 1 patient with size 38, and 8 patients with size 42. All measurements were normalized to account for differences in glenosphere size. Groups 1 and 2 were compared on each radiographic parameter (inferior scapular notching, PSNA, PGRD, SNA).

Notching indexes were calculated ([PSNA × 0.13] + PGRD) and compared with the suggested index of 35.⁸ Simovitch and colleagues⁸ demonstrated that a notching index of more than 35 had 91% sensitivity and 88% specificity in predicting inferior notching, whereas a notching index of 35 or less avoided inferior notching 91% of the time. In this study, notching index was calculated for each patient, and then the mean values of groups 1 and 2 were compared (Table 1).

The effect of scapular notching and other individual radiographic parameters on outcomes was also evaluated with respect to forward flexion, external rotation, VAS pain score, complications, and external rotation lag sign. Mann-Whitney U test was used to test these variables; Spearman rank test was performed to determine correlation between each variable and scapular notching; logistic regression was used to explore the relationship of variables (PGRD, PSNA) as predictors of Nerot degree of inferior scapular notching, and postoperative complications; and independent-samples t test was used to determine group differences for each variable. For each investigation, the level of significance was set at P < .05. A biostatistician performed all statistical analyses using SPSS Version 19 (IBM, Armonk, New York).

Results

Our study cohort consisted of 91 shoulders. Mean follow-up was 41.8 months (range, 24.0-80.8 months). Seventy-five (82%) of the 91 shoulders developed scapular notching. Mean (SD) SNA on preoperative radiographs, used to assess preoperative scapular morphology, was 103.9° (14.5°). For all 91 shoulders, mean (SD) PSNA was 125.6° (16°), and mean (SD) PGRD was 16 (5.4) mm (Table 1). Inclination measurements were available for 86 patients. Mean (SD) inclination from 90° was 2.5° (10.3°) (range, 21°-30°). Mean (SD) SNA (postoperative PSNA minus preoperative SNA) for the 66 patients with acceptable preoperative radiographs was 24.3° (21.3°) (Table 1). Forty-eight of the 91 shoulders were placed in scapular notching group 1 (16 grade-0 shoulders, 32 grade-1 shoulders); the other 43 shoulders were placed in group 2 (33 grade-2 shoulders, 9 grade-3 shoulders, 1 grade-4 shoulder). Mean follow-up was 40 months for group 1 and 43 months for group 2.

There were no significant differences between groups 1 and 2 in SNA (102.8° vs 105.4°; P = .3), PGRD (15.4 vs 16.8 mm; P = . 47), or PSNA (125.8° vs 125.4°; P = .82) (Table 1). In addition, groups 1 and 2 had no significant differences (P > .05) in glenoid overhang and glenosphere inclination (other possible factors influencing notching).

Mean (SD) notching index was 31.8 (4.4) for group 1 and 33.1 (7.2) for group 2. These values were not significantly different (P = .29) (Table 1, Figure 2).⁸ Each was below the recommended threshold of 35 for prevention of notching (Table 1, Figure 2).

To try to understand why mean scapular notching index was low for both groups, we examined the contributing factors individually. Our cohort’s mean PGRD of 16.1 mm (15.4 and 16.8 mm for groups 1 and 2, respectively) was significantly lower than the cohort mean reported by Simovitch and colleagues⁸ (Table 2). Given that PGRD is more strongly weighted in the originally described notching index ([PSNA × 0.13] + PGRD),⁸ it was the primary driver for our notching index results, even though on average our results demonstrated a PSNA higher than that found by Simovitch and colleagues⁸ (Table 2; Figures 3, 4). Analyzing PGRD and PSNA together, we found no relationship between these variables and increased severity of inferior notching (Figure 5).

Regarding the effects of notching severity on outcomes in our study cohort, there were no significant differences between groups 1 and 2 in postoperative function, including forward flexion (123° vs 112.4°; P = .11), external rotation (18.8° vs 16.7°; P = .76), positive lag sign (P = .2), and VAS pain scores (1.2 vs 2.1; P = .15). There were also no significant differences between groups in the rate of complications (P = .92). Regression analysis determined that PSNA, PGRD, glenosphere inclination, glenosphere overhang, and implant manufacturer were not significant predictors of complications.

Discussion

RSA has provided good pain relief and restored function in patients with irreparable rotator cuff disease associated with arthritis.^5,12,17,18 Scapular notching is a complex, multifactorial process. Nevertheless, surgeons remain cautious about the implications of inferior scapular notching, which is being reported by a significant number of patients. Our cohort’s high incidence of scapular notching (82%) in the early postoperative period clearly highlights the importance of predictive models, such as the notching index.⁸ Although concerns about consequences of notching have been expressed, notching severity did not affect outcomes or increase complications in this cohort.^{5,8,11,12,17-19}

We conducted this study to examine use of a predictive tool for scapular notching, the notching index, in a large cohort of patients who underwent primary RSA. This index combines 2 well-established factors that contribute to notching—craniocaudal position and PSNA—into a predictive formula based on statistical analyses performed in a prospective cohort study.^4,5,8,12,18 In their clinical study, Simovitch and colleagues⁸ found that both craniocaudal position and PSNA were tightly coupled with inferior scapular notching, and they developed a notching index that accounts for this relationship. We hypothesized that patients with a notching index of less than the recommended 35 would not develop notching and that patients with a notching index of more than 35 would have increased incidence and severity of notching. With our cohort, the recommended index of 35 was not an appropriate threshold predictive of notching. Furthermore, the 35 threshold applied to our cohort had 89% sensitivity and 21% specificity in predicting notching. Although the sensitivity is high, and correctly predicted true instances of notching, the low specificity compromises the precision of the notching formula ([PSNA × 0.13] + PGRD).

From the formula, it can be inferred that higher PSNA values can be compensated for by decreasing PGRD and inferiorizing the glenosphere. However, this recommendation appears limited based on increasing PSNA values, as in our cohort. The previously described notching formula cannot be universally applied to all patients treated with RSA because of the complexity of this relationship and patient-specific anatomy.

 We assessed other possible anatomical and surgical factors, specific to scapular morphology, that could contribute to scapular notching. In other studies, reaming that produced an inferior tilt of the glenoid increased the likelihood of inferior notching.^8,20,21 Furthermore, we expected less inferior glenoid overhang and smaller glenosphere would predispose patients to more notching.^8,12,19 In our cohort, notching grade was not correlated with inferior tilt, glenoid overhang, or glenosphere size, which may be attributed to minimal variability in glenosphere size and a small range of glenosphere overhang.

There were limitations to this study. We examined only 2 types of RSA systems, and they had very similar Grammont designs. Other RSA designs might not have similar shortcomings with respect to inferior notching. In addition, we examined patient cases at a single time point and did not evaluate the effect of notching over time.

Overall, our results suggest that PGRD and PSNA have little effect on development of higher grade notching, particularly with use of Grammont prostheses. With newer surgical techniques, the recommendation is for inferior craniocaudal placement of the glenosphere, but this may not prevent notching with some types of patient-specific scapular morphology. Clearer surgical guidelines and techniques may help delineate the contribution of each parameter causing inferior scapular notching. Surgeons must weigh the evidence to determine how to correct patient-specific glenoid pathology and orient the glenosphere. Recent studies on bony increased-offset reverse shoulder arthroplasty (bio-RSA) techniques or newer prosthetic designs that considerably alter PSNA and the center of rotation may prevent inferior notching and provide a promising alternative to Grammont designs. Ultimately, longer follow-up is also needed to understand the clinical relevance of increased scapular notching.

Reverse shoulder arthroplasty (RSA) is a treatment option for a spectrum of diseases in shoulders with rotator cuff deficiency. There are distinct morphologic changes in the scapular and glenoid anatomy in patients with chronic rotator cuff tears.¹ A muscular imbalance that occurs in the joint as a result of rotator cuff deficiency leads to morphologic changes that eliminate the compressive forces that hold the humeral head against the glenoid.² RSA effectively stabilizes the glenohumeral joint in shoulders with deficient rotator cuffs.^3,4 In early work, Grammont proposed that the glenosphere center of rotation should be medialized (concentric to the central axis of the metaglene or baseplate) and lowered.⁵ Although the medialized center of rotation in Grammont prostheses decreases shear forces and improves the deltoid lever arm, it also tends to result in mechanical impingement between the superomedial aspect of the humeral polyethylene insert and the scapular neck—so-called inferior scapular notching.^6-9

Notching, which has been reported in 50% to 96% of patients who receive a Delta III prosthesis, typically appears within the first few months after surgery but may be seen as late as 14 months after surgery.^5,10-12 Postmortem studies have shown that notching corresponds with erosion of the inferior pole of the glenoid and scapular neck, thought to be caused by the polyethylene cup of the implant.¹³ Although some studies have found that notching stabilizes after 1 year, others have shown notching progressing for up to 4 years after surgery.^11,12,14 The clinical relevance of notching continues to be controversial, but notching has been associated with poorer clinical outcomes, polyethylene wear, and local osteolysis. Component loosening has also been reported with notching of grade 3 or more.^8,10 Ultimately, there is concern that scapular notching could progress, ultimately leading to late glenoid loosening and potentially catastrophic failure.

Scapular anatomy has become an area of increased focus in rotator cuff disorders and in effects on RSA biomechanics.⁹ Recent reports have described important scapular morphology variations that suggest more individualized adjustments are needed during RSA.^9,15 In addition, some investigators have reported that development of notching appears to depend on the height and inclination of the implanted glenoid component, where an inferior position of the glenosphere leads to less impingement and better range of motion.^8,16 Simovitch and colleagues⁸ found the angle between the glenosphere and scapular neck and the craniocaudal position of the glenosphere to be highly correlated with inferior notching. They combined these 2 parameters into a predictive algorithm that provides a guideline (notching index, <35) for prevention of notching.

We conducted a study to evaluate the scapular notching index as a predictive tool and to consider other factors that may be associated with scapular notching occurring with use of Grammont reverse replacement systems. We hypothesized that patients with a notching index of less than 35 would not develop notching and that patients with an index of more than 35 would have increased incidence and severity of notching.

Materials and Methods

Patients treated with RSA for painful cuff tear arthropathy or irreparable rotator cuff tear with pseudoparesis (inability to actively elevate shoulder >90° in presence of free passive anterior elevation) were included in this retrospective review. All patients were treated between 2006 and 2010 by 1 of 2 established senior shoulder subspecialty surgeons. Patients treated with a Delta (DePuy Orthopaedics, Warsaw, Indiana) or an Aequalis (Tornier, Edina, Minnesota) reverse shoulder implant were included in the study. A standard polyethylene liner was used for all patients. These prostheses have the same neck shaft angle, 155º, as they have similar geometric designs, both based on the Grammont design—semiconstrained inverted with a fixed, lowered, medialized center of rotation. Standard instrumentation was used for all procedures. Patients were excluded if any nonstandard techniques or components were used (constrained or high-mobility liner, glenoid bone grafting). Patients who underwent revision for a previous reverse total arthroplasty, a total shoulder arthroplasty, or a hemiarthroplasty, or for treatment of acute fracture, posttraumatic deformity, or posttraumatic arthritis, were also excluded from our analyses. Minimum follow-up for study inclusion was 24 months.

All procedures were performed with the patient in the semi-beach-chair position and with use of a deltopectoral approach. The glenoid was prepared such that minimal reaming was needed to preserve the subchondral plate. The glenoid baseplate was positioned in the recommended inferior position to minimize notching and optimize functional outcomes.¹³ After surgery, all patients were managed with a simple soft immobilizer with or without a pillow with the arm at the patient’s side in internal rotation. Immediate passive mobilization was begun under the direction of physical therapists. Passive and active-assisted exercises were continued with gradual progression to independent activities of daily living at 6 weeks. Clinical evaluations were performed before and after surgery by the operating surgeon or independent research nurse. Active forward flexion, passive external rotation, strength, and visual analog scale (VAS) pain scores were reviewed and recorded. Case-specific complications were also reviewed.

Preoperative and postoperative anteroposterior radiographs were evaluated by 2 independent observers (attending surgeon, junior resident). Per standard technique, each radiograph was positioned horizontal to the scapular plane. Of the 91 patients, 66 had preoperative shoulder radiographs of acceptable quality, with complete visualization of scapular morphology. Radiographs were reviewed to measure the scapular neck angle (SNA), inferior scapular notching, prosthesis–scapular neck angle (PSNA), and peg glenoid rim distance (PGRD) (Figure 1). For the 66 patients with acceptable preoperative radiographs, SNA was determined by subtracting preoperative SNA from postoperative PSNA. Postoperative anteroposterior radiographs were used to classify degree of inferior scapular notching based on the Nerot grading scale (0-4). In addition, glenosphere overhang and glenosphere inclination were measured on postoperative radiographs.

The 91 shoulders were sorted into 2 groups based on degree of scapular notching: group 1, Nerot grade 0 (no inferior notching) and grade 1, and group 2, Nerot grades 2, 3, and 4. Group 1 had 37 patients with a size 36 glenosphere, 3 patients with size 38, and 8 patients with size 42; group 2 had 34 patients with a size 36 glenosphere, 1 patient with size 38, and 8 patients with size 42. All measurements were normalized to account for differences in glenosphere size. Groups 1 and 2 were compared on each radiographic parameter (inferior scapular notching, PSNA, PGRD, SNA).

Notching indexes were calculated ([PSNA × 0.13] + PGRD) and compared with the suggested index of 35.⁸ Simovitch and colleagues⁸ demonstrated that a notching index of more than 35 had 91% sensitivity and 88% specificity in predicting inferior notching, whereas a notching index of 35 or less avoided inferior notching 91% of the time. In this study, notching index was calculated for each patient, and then the mean values of groups 1 and 2 were compared (Table 1).

The effect of scapular notching and other individual radiographic parameters on outcomes was also evaluated with respect to forward flexion, external rotation, VAS pain score, complications, and external rotation lag sign. Mann-Whitney U test was used to test these variables; Spearman rank test was performed to determine correlation between each variable and scapular notching; logistic regression was used to explore the relationship of variables (PGRD, PSNA) as predictors of Nerot degree of inferior scapular notching, and postoperative complications; and independent-samples t test was used to determine group differences for each variable. For each investigation, the level of significance was set at P < .05. A biostatistician performed all statistical analyses using SPSS Version 19 (IBM, Armonk, New York).

Results

Our study cohort consisted of 91 shoulders. Mean follow-up was 41.8 months (range, 24.0-80.8 months). Seventy-five (82%) of the 91 shoulders developed scapular notching. Mean (SD) SNA on preoperative radiographs, used to assess preoperative scapular morphology, was 103.9° (14.5°). For all 91 shoulders, mean (SD) PSNA was 125.6° (16°), and mean (SD) PGRD was 16 (5.4) mm (Table 1). Inclination measurements were available for 86 patients. Mean (SD) inclination from 90° was 2.5° (10.3°) (range, 21°-30°). Mean (SD) SNA (postoperative PSNA minus preoperative SNA) for the 66 patients with acceptable preoperative radiographs was 24.3° (21.3°) (Table 1). Forty-eight of the 91 shoulders were placed in scapular notching group 1 (16 grade-0 shoulders, 32 grade-1 shoulders); the other 43 shoulders were placed in group 2 (33 grade-2 shoulders, 9 grade-3 shoulders, 1 grade-4 shoulder). Mean follow-up was 40 months for group 1 and 43 months for group 2.

There were no significant differences between groups 1 and 2 in SNA (102.8° vs 105.4°; P = .3), PGRD (15.4 vs 16.8 mm; P = . 47), or PSNA (125.8° vs 125.4°; P = .82) (Table 1). In addition, groups 1 and 2 had no significant differences (P > .05) in glenoid overhang and glenosphere inclination (other possible factors influencing notching).

Mean (SD) notching index was 31.8 (4.4) for group 1 and 33.1 (7.2) for group 2. These values were not significantly different (P = .29) (Table 1, Figure 2).⁸ Each was below the recommended threshold of 35 for prevention of notching (Table 1, Figure 2).

To try to understand why mean scapular notching index was low for both groups, we examined the contributing factors individually. Our cohort’s mean PGRD of 16.1 mm (15.4 and 16.8 mm for groups 1 and 2, respectively) was significantly lower than the cohort mean reported by Simovitch and colleagues⁸ (Table 2). Given that PGRD is more strongly weighted in the originally described notching index ([PSNA × 0.13] + PGRD),⁸ it was the primary driver for our notching index results, even though on average our results demonstrated a PSNA higher than that found by Simovitch and colleagues⁸ (Table 2; Figures 3, 4). Analyzing PGRD and PSNA together, we found no relationship between these variables and increased severity of inferior notching (Figure 5).

Regarding the effects of notching severity on outcomes in our study cohort, there were no significant differences between groups 1 and 2 in postoperative function, including forward flexion (123° vs 112.4°; P = .11), external rotation (18.8° vs 16.7°; P = .76), positive lag sign (P = .2), and VAS pain scores (1.2 vs 2.1; P = .15). There were also no significant differences between groups in the rate of complications (P = .92). Regression analysis determined that PSNA, PGRD, glenosphere inclination, glenosphere overhang, and implant manufacturer were not significant predictors of complications.

Discussion

RSA has provided good pain relief and restored function in patients with irreparable rotator cuff disease associated with arthritis.^5,12,17,18 Scapular notching is a complex, multifactorial process. Nevertheless, surgeons remain cautious about the implications of inferior scapular notching, which is being reported by a significant number of patients. Our cohort’s high incidence of scapular notching (82%) in the early postoperative period clearly highlights the importance of predictive models, such as the notching index.⁸ Although concerns about consequences of notching have been expressed, notching severity did not affect outcomes or increase complications in this cohort.^{5,8,11,12,17-19}

We conducted this study to examine use of a predictive tool for scapular notching, the notching index, in a large cohort of patients who underwent primary RSA. This index combines 2 well-established factors that contribute to notching—craniocaudal position and PSNA—into a predictive formula based on statistical analyses performed in a prospective cohort study.^4,5,8,12,18 In their clinical study, Simovitch and colleagues⁸ found that both craniocaudal position and PSNA were tightly coupled with inferior scapular notching, and they developed a notching index that accounts for this relationship. We hypothesized that patients with a notching index of less than the recommended 35 would not develop notching and that patients with a notching index of more than 35 would have increased incidence and severity of notching. With our cohort, the recommended index of 35 was not an appropriate threshold predictive of notching. Furthermore, the 35 threshold applied to our cohort had 89% sensitivity and 21% specificity in predicting notching. Although the sensitivity is high, and correctly predicted true instances of notching, the low specificity compromises the precision of the notching formula ([PSNA × 0.13] + PGRD).

From the formula, it can be inferred that higher PSNA values can be compensated for by decreasing PGRD and inferiorizing the glenosphere. However, this recommendation appears limited based on increasing PSNA values, as in our cohort. The previously described notching formula cannot be universally applied to all patients treated with RSA because of the complexity of this relationship and patient-specific anatomy.

 We assessed other possible anatomical and surgical factors, specific to scapular morphology, that could contribute to scapular notching. In other studies, reaming that produced an inferior tilt of the glenoid increased the likelihood of inferior notching.^8,20,21 Furthermore, we expected less inferior glenoid overhang and smaller glenosphere would predispose patients to more notching.^8,12,19 In our cohort, notching grade was not correlated with inferior tilt, glenoid overhang, or glenosphere size, which may be attributed to minimal variability in glenosphere size and a small range of glenosphere overhang.

There were limitations to this study. We examined only 2 types of RSA systems, and they had very similar Grammont designs. Other RSA designs might not have similar shortcomings with respect to inferior notching. In addition, we examined patient cases at a single time point and did not evaluate the effect of notching over time.

Overall, our results suggest that PGRD and PSNA have little effect on development of higher grade notching, particularly with use of Grammont prostheses. With newer surgical techniques, the recommendation is for inferior craniocaudal placement of the glenosphere, but this may not prevent notching with some types of patient-specific scapular morphology. Clearer surgical guidelines and techniques may help delineate the contribution of each parameter causing inferior scapular notching. Surgeons must weigh the evidence to determine how to correct patient-specific glenoid pathology and orient the glenosphere. Recent studies on bony increased-offset reverse shoulder arthroplasty (bio-RSA) techniques or newer prosthetic designs that considerably alter PSNA and the center of rotation may prevent inferior notching and provide a promising alternative to Grammont designs. Ultimately, longer follow-up is also needed to understand the clinical relevance of increased scapular notching.