Incidence of and Risk Factors for Symptomatic Venous Thromboembolism After Shoulder Arthroplasty

Of the 533 patients, 288 were female and 245 were male. Mean age at surgery was 65.2 years (range, 16-93 years). Mean (SD) BMI was 29.2 (6.4) kg/m². Mean (SD) preoperative Hb level was 13.7 (1.8) g/dL, and mean preoperative Hct level was 40.1% (4.8%). Mean (SD) length of hospital stay was 2.6 (1.5) days. Mean (SD) time before patients were out of bed was 1.1 (0.7) days. On postoperative day 1, mean Hb level was 11.1 (1.7) g/dL, and mean (SD) Hct level was 33.2% (4.8%). Mean (SD) CCI was 1.1 (0.9).

Anesthesia for the 533 patients consisted of GETA (209 patients, 39.0%), interscalene nerve block (2, 0.4%), or GETA with nerve block (314, 59.0%). After surgery, 125 patients (24.3%) received aspirin as prophylaxis. Diabetes was reported by 83 patients, hypertension by 286, cardiac disease by 74, a history of a clotting disorder by 2, a family history of a clotting disorder by 8, ongoing cancer by 4, a history of cancer by 67, and hormone replacement therapy by 104.

For the entire cohort of 533 patients, the symptomatic VTE rate was 2.6% (14 patients), the DVT rate was 0.9% (5), and the PE rate was 2.3% (12). Although VTE did not cause any deaths, there were 3 cardiac events.

Discussion

VTE after SA is rare. We report an overall VTE incidence of 2.6%, with DVT at 0.9% and PE at 2.3%. These rates are similar to those reported in clinical series and significantly higher than those reported for large institutional or national databases.^2-7 Our results also support a previously reported trend: The ratio of PE to DVT for SA is significantly higher than historically reported ratios for lower extremity arthroplasty.^2,6-8 We have identified many VTE risk factors: raised CCI, preoperative thrombotic event, lower preoperative Hb and Hct levels, lower postoperative Hb level, diabetes, use of GETA without interscalene nerve block, higher BMI, and revision SA. Results of other studies support 3 findings (higher BMI, raised CCI, preoperative thrombotic event); new findings include correlation with Hb and Hct levels, diabetes, type of anesthesia, and revision SA.^6,7 Identification of these other factors may be useful in making treatment decisions in patients symptomatic after SA and in lowering the threshold for performing diagnostic tests in these patients at risk for VTE.

Reported rates of VTE after SA are highly variable, ranging from 0.2% to 13%.¹⁰ Our rationale for investigating VTE rates at a single institution was to estimate the rates that can be expected in a university-based practice and to determine whether these rates are high enough to warrant routine thromboprophylaxis. The rate variability seems to result in part from variability in the data sources. Most studies that have reported very low VTE rates typically used large state or national databases, which likely were subject to underreporting.

Lyman and colleagues⁶ found 0.5% DVT and 0.2% PE rates in a New York state hospital database, but only in-hospital immediate postoperative symptomatic complications were included; slightly delayed complications may have been missed. Farng and colleagues⁵ reported a 0.6% VTE rate, but only inpatient (immediate postoperative or readmission) events were included; all outpatient events were missed. Jameson and colleagues,² using a national database that included only cases involving inpatient treatment, reported 0% DVT and 0.2% PE rates, again missing outpatient events, and relying on appropriate coding to capture events. Using electronic health records from a large healthcare system, Navarro and colleagues⁸ queried for VTE cases and reported 0.5% DVT and 0.5% PE rates. The inclusiveness of their data source for the outcome of interest was potentially improved relative to national or statewide databases—and the resulting data reported in their study should reflect that improvement. However, the authors relied on ICD–9 (International Classification of Diseases, Ninth Revision) coding to screen for VTE events and excluded patients with prior VTE, preoperative prophylaxis (enoxaparin or warfarin), or follow-up of <90 days. As patients with prior VTE are those most at risk (present study OR, 6-7), excluding them significantly reduces the overall incidence of clotting reported.

Only 4 studies specifically used information drawn directly from physicians’ clinic notes, vs data retrieved (using code-based queries) from databases.^1,3,4,7 These studies may provide a better representation of the rate of VTE after SA, as they were not reliant on codes, included both inpatient and outpatient events, and were inclusive of outpatient follow-up of at least 3 months.

Three of the 4 studies used the Mayo Clinic Total Joint Registry.^1,3,4 Hoxie and colleagues¹ reported an 11% rate of PE after HA performed for fracture (we excluded SA for fracture). As several other investigators have reported an association between trauma and increased risk for VTE, postoperative anticoagulation should be considered in this patient population (though it was not the focus of the present study).^6-8 Sperling and Cofield³ and Singh and colleagues⁷ reported on the risk for PE among SA patients at the Mayo Clinic. Sperling and Cofield³ included only those events that occurred within the first 7 days after surgery; Singh and colleagues⁷ included events out to 90 days after surgery. Sperling and Cofield³ reported a 0.17% PE rate; Singh and colleagues⁷ reported 0.6% PE and 0.1% DVT rates. Sperling and Cofield³ reported on 2885 SAs; Singh and colleagues⁷ reported on 4019 SAs from the same database. As it is unclear whether these 2 studies had complete information on all patients, underreporting may be an issue. Information was obtained through “clinic visits, medical records and/or standardized mailed and telephone-administered questionnaires.”⁷The fourth study, a prospective study of 100 patients by Willis and colleagues,⁴ had the best data on development of symptomatic PE after SA. The authors reported a 2% PE rate and a high (13%) DVT rate. Because US was not performed before the surgical procedures, the number of patients with new and existing DVT cases could not be determined. However, all PEs were new, and the 2% rate found there is similar to the 2.3% in our study. Therefore, we think these rates capture the data most accurately and avoid the underreporting that marks large databases.⁴Studies have identified various factors that increase the risk for VTE after SA. Singh and colleagues⁷ identified the risk factors of age over 70 years, female sex, higher BMI (25-29.9 kg/m²), CCI above 1, traumatic etiology, prior history of VTE, and HA. However, their use of univariate regression analysis may have confounded the effects—one factor may have become a surrogate for another (ie, trauma and HA, as most fractures treated with SA during the study period were treated with HA). Lyman and colleagues⁶ also found advanced age and trauma were associated with higher VTE risk, and reported prior history of cancer as a risk factor as well. Navarro and colleagues⁸ identified trauma as a risk factor, as in the other 2 studies.^6,7 Our data support prior history of VTE, higher BMI, and raised CCI as increasing the risk for VTE.

Other factors identified in the present study are use of GETA without interscalene nerve block, lower preoperative and postoperative Hb levels, diabetes, and revision SA. Because of the limited number of events, only ORs with and without limited control of confounders were performed. Just as in the study by Singh and colleagues,⁷ uncontrolled confounding could have occurred. A nerve block may be protective, as less postoperative pain may allow patients quicker mobilization and therapy. Diabetes may be a surrogate for other medical comorbidities, as reflected by the higher overall risk with raised CCI. Lower preoperative and postoperative Hb levels were associated with clotting and may be representative of patients with poorer overall health and more complicated surgical procedures (eg, revision SA). In an earlier study, we found increased risk for transfusions in revision SA relative to primary SA.¹¹ Lower preoperative Hb level correlated with development of VTE after lower extremity arthroplasty.¹² Postoperative use of aspirin was not found to significantly reduce the incidence of clotting, though this finding may have resulted from lack of power. Therefore, from the present data, there is nothing to conclude about the efficacy of aspirin in preventing thrombosis.

Our findings can be placed in the context of the Virchow triad. Specifically, 3 categories of factors are thought to contribute to thrombosis: hypercoagulability, hemodynamic stasis, and endothelial injury. In grouping factors, we identified prior thrombotic event and obesity as increasing hypercoagulability; revision SA, more comorbidities, lower Hb and Hct levels, diabetes, and GETA as increasing hemodynamic stasis; and revision SA (longer operating room times) as leading to stasis. More comorbidities can be associated with delayed postoperative ambulation, and diabetes and lower Hb and Hct levels can be surrogates for more comorbidities. Surgery performed with the patient under GETA without interscalene nerve block can lead to higher levels of pain and less early mobility.

The present findings have made us more aware of patients at risk for VTE, and we have lowered our threshold for evaluating them for potential clots. Before this study, we used warfarin or enoxaparin for anticoagulation in patients with a history of VTE or active cancer. We are continuing this protocol, but not with other patients. Patients with many comorbidities, lower preoperative Hb level, revision SA, high BMI, or diabetes are carefully monitored for clots early in the postoperative course. Our new threshold for these high-risk patients is to order diagnostic testing, including duplex US or CT angiography. Now, even mild oxygen requirements or mild tachycardia within postoperative week 1 typically prompt a study in these patients. We hope this increased awareness will limit the potential negative consequences associated with development of VTE. Given the present data, we do not think the simple presence of increased comorbidities, lower preoperative Hb, revision SA, high BMI or diabetes should rule out performing SA; rather, it should increase surgeons’ postoperative vigilance in evaluating for potential clots.

Limitations of our study include its retrospective nature and reliance on clinic chart review. Patients were not directly questioned about venous thrombus at follow-up, so all events may not have been captured. Although retrospective review has its drawbacks, it allows for accurate identification of events, even uncoded events. Therefore, more events are likely to be captured with this technique than with large database analyses using only coding information. We tried to identify as many cases as possible by reviewing all outpatient records (orthopedic, nonorthopedic), inpatient records, radiologic studies, and scanned outside records. Another limitation is that having a small number of VTE events limited our ability to perform a multivariate analysis, and uncontrolled confounding likely resulted. Only a very large multi-institutional study can capture enough events to allow a multivariate analysis. A third limitation is that the small number of events may have underpowered the study. Having more patients would have allowed other potential factors to be identified as being significantly associated with VTE. Last, as the study captured only symptomatic VTE events, it may have underreported VTE events. Given our complete review of the medical records, however, most clinically significant events likely were captured.