Venous thromboembolism in patients hospitalized for knee joint replacement surgery

Patients undergoing knee joint replacement (KJR) are at high risk of postoperative venous thromboembolism (VTE), but data on the time trends of VTE rate in this population are sparse. In this analysis of the German nationwide inpatient sample, we included all hospitalizations for elective primary KJR in Germany 2005–2016. Overall, 1,804,496 hospitalized patients with elective primary KJR (65.1% women, 70.0 years [IQR 63.0–76.0]) were included in the analysis. During hospitalization, VTE was documented in 23,297 (1.3%) patients. Total numbers of primary KJR increased from 129,832 in 2005 to 167,881 in 2016 (β-(slope)-estimate 1,978 [95% CI 1,951 to 2,004], P < 0.001). In-hospital VTE decreased from 2,429 (1.9% of all hospitalizations for KJR) to 1,548 (0.9%) cases (β-estimate − 0.77 [95% CI − 0.81 to − 0.72], P < 0.001), and in-hospital death rate from 0.14% (184 deaths) to 0.09% (146 deaths) (β-estimate − 0.44 deaths per year [95% CI − 0.59 to − 0.30], P < 0.001). Infections during hospitalization were associated with a higher VTE risk. VTE events were independently associated with in-hospital death (OR 20.86 [95% CI 18.78–23.15], P < 0.001). Annual number of KJR performed in Germany increased by almost 30% between 2005 and 2016. In parallel, in-hospital VTE rates decreased from 1.9 to 0.9%. Perioperative infections were associated with higher risk for VTE. Patients who developed VTE had a 21-fold increased risk of in-hospital death.

Temporal trends of primary surgical knee joint replacements, patients' baseline characteristics and in-hospital events. The total number of KJR performed as well as in-hospital death rates increased substantially with age ( Fig. 1A). While men were more frequently represented than women in the first three decades of life, the proportion of women was higher among patients aged 30 Table S2 in the supplementary material). The median duration of hospitalization for KJR decreased from 15 (13)(14)(15)(16) days in 2005 to 10 (8-12) in 2016 (β-estimate − 0.12 [95% CI − 0.12 to − 0.12], P < 0.001) ( Fig. 2A) and was primarily depending on patients' age ( Fig. S3 in the supplementary material).
The proportion of patients aged 70 years or older among those who underwent KJR, decreased from 54.0% in 2005 to 50.3% in 2016 (β-estimate − 0.14 [95% CI − 0.15 to − 0.13], P < 0.001). In parallel, the prevalence of some important comorbidities such as cancer and coronary artery disease decreased during the same timeframe ( Fig. S2B + C, Fig. S4 and Table S1 in the supplementary material). Although females outweighed male patients with regard to KJR performed throughout the entire study period, the proportion of male patients increased slightly over time ( Fig. S2A and Table S1 in the supplementary material). The rate of myocardial infarction, stroke, intracerebral bleeding complications as well as necessity of transfusions of blood constituents decreased ( Fig. 2B + C, Fig. S5 (Table S3 in the supplementary material). In addition, we analysed the total numbers of in-hospital deaths in patients undergoing primary KJR at each day of hospitalization (day 1-15) and the proportion of deaths, which were related to prior VTE events during hospitalization at each day. The proportion of VTE related death was highest at the first 7 days after admission and decreased over the hospitalization period (Fig. 3). As expected, bleeding complications such as intracerebral and gastrointestinal bleeding as well as transfusions of blood components were more often detected in KJR patients with VTE ( Table 1). The rate of VTE and bleeding complications remained low and decreased over time, as did the need for transfusions of blood constituents (Figs. 2C and 3C as well as Table S2 in the supplementary material). The length of in-hospital stay was in median 12.0 days (IQR 10.0-14.0) and was longer in patients with (14 [12][13][14][15][16][17][18]) vs. without (12 [10][11][12][13][14]

Discussion
VTE, with its clinical manifestations of DVT and PE, is responsible for significant morbidity and mortality in Europe and worldwide 5,12 . Important major provoking risk factors of VTE are represented by orthopaedic-related issues, notably major trauma and surgery, lower-limb fracture, joint replacement, and spinal cord injury 5 .
Our study results demonstrate an increasing demand and rising annual numbers of performed KJR in the aging western populations, in line with what described in other high-income states 2,6,8-11 . These temporal trends regarding total numbers of KJR and perioperative complications are important for adequate management of public health as well as health care service planning. Although KJR can substantially improve patients` mobility and quality of life in patients with advanced gonarthritis, it is accompanied by a high risk of VTE 2,13-16 . While without thromboprophylaxis, the rate of DVT in screening examinations in patients undergoing KJR was reported as approximately 60% (including symptomatic as well as asymptomatic DVT events) 13,14 , the risk of VTE after KJR can be significantly decreased by the use of pharmacological thromboprophylaxis 2 . In accordance with results from studies of KJR conducted in other states reporting a 1.1% in-hospital VTE rate after KJR 17 , we identified an in-hospital VTE rate of 1.3% in patients undergoing elective KJR in Germany. Importantly for adequate health care planning, the VTE rate decreased significantly from 1.9% in 2005 to 0.9% in 2016. This favorable trend may be due to several factors, including the introduction of the direct oral anticoagulants, the decrease in the use of tourniquets and/or surgical time of tourniquet use, a small increase in the number of uncemented KJR procedures 18 , the incline in unicondylar KJR 18 , decreasing age and the introduction of fast-track procedures improving early mobilization and discharge 19,20 .
Current practice guidelines provide partly conflicting recommendations for the medical VTE prophylaxis after KJR 21,22 . Additionally, the appropriate duration of anticoagulation following elective KJR is highly controversial and a dynamically evolving topic. Country-specific guideline recommendations as well as physicians' and patients' preferences, have a major impact on the management strategies. While the German AWMF guidelines recommend low-molecular weight heparins (LMWH), fondaparinux or novel oral anticoagulants (NOAC) at a prophylactic dose for 11-14 days after KJR 22 , the National Institute of Health and Care Excellence guidelines (NICE) in the United Kingdom recommend LMWH, or NOACs, with aspirin recommended for extended prophylaxis 21,23 . The guidelines of the American College of Chest Physicians and the American Association of Orthopaedic Surgeons recommend LMWH, (N)OACs or aspirin 3,6,7,21,24 .
There is still a great controversy about the effect of tourniquets on development of VTE after KJR [25][26][27][28] . The use of a tourniquet improves visualization during KJR surgery and may shorten the operating time 28 . However, there www.nature.com/scientificreports/ is also some evidence that the use of tourniquet during KJR surgery may lead to increased numbers of DVT [25][26][27][28] . Furthermore, previous studies have reported that cemented fixation as compared to cement-less fixation was found as a risk factor for VTE in hip and KJR 29,30 ; nevertheless, the role of cement as a thrombogenic agent and cause of emboli during or after joint replacements is not entirely clear to date [30][31][32] . In Germany more than 90% of the KJR were operated with cement fixation 18 . Finally, the decreasing proportion of VTE in KJR patients may partly be explained by trends regarding patients' age during the observational period. It is known that the incidence of VTE grows exponentially with age [33][34][35] 43,44 , and renal insufficiency 45 . In addition, acute cardiovascular events during hospitalization for elective KJR such as myocardial infarction 40,46 and stroke 37,47 , but also infections 48 occurring during hospitalization were significantly associated with VTE development, although our study design does not permit speculation about cause and effect. In fact, our results demonstrated that systemic infectious diseases (pneumonia and sepsis) were accompanied with a higher VTE risk than traditional VTE risk factors like cancer and thrombophilia. VTE events in patients undergoing primary KJR were associated with a significant 3.1-fold risk of prolonged hospitalization (beyond 14 days).
Our study highlights that perioperative VTE events aggravate the early prognosis of patients operated at the knee joint: perioperative VTE events were associated with a 21-fold increased risk of in-hospital deaths independently from age, sex and comorbidities. In parallel with reduction of VTE events, the KJR patients' in-hospital mortality decreased from 2005 to 2016. In addition, we detected a decreasing rate of periprocedural myocardial infarction, stroke and major bleeding, which might have contributed to this mortality reduction.
Although the median duration of hospitalization for elective KJR decreased from 15 days in the year 2005 to 10 days in the year 2016, the median length of stay in the year 2016 was still substantially longer in Germany than in the United Kingdom (6 days in regular medical care and 3 days for enhanced recovery programs) 49 or in the United States (4 days) 50 . In other European countries, enhanced recovery protocols have been adopted, resulting in short length of hospitalization of only 2 days (in median) 20 . Interestingly, enhanced recovery with halved length of in-hospital stay was accompanied by lower rate of cardiac ischemic events and lower mortality rate at 30 and 90 days follow-up after KJR and hip joint replacement 49 .  www.nature.com/scientificreports/ Recent studies of nationwide cohorts reported that patients treated according these protocols had a low 90-day VTE rate of 0.39% 20 . The results of these mentioned studies contributed to fuel the discussion on whether prolonged pharmacological thromboprophylaxis (after discharge from hospital) is beneficial and needed for all patients treated with an enhanced recovery protocol and hospitalized for a period of 5 days or shorter 19,20 . In these patients, a preventive strategy based on in-hospital thrombo-prophylactic treatment with the Factor Xa inhibitor rivaroxaban followed by low-dose aspirin appeared to be effective and safe for thromboprophylaxis 50 . These results remain to be confirmed in other countries, in which fast-track surgery and enhanced recovery protocols will be introduced or were introduced more recently 21 .
The key strength of our present analysis of the German nationwide inpatient sample is the very large number of unselected patients undergoing KJR, presenting real-world and practice-based data without selection bias. On the other hand, since our study results are based on ICD discharge codes and OPS diagnostic, procedural and/or surgical coding, it has to be mentioned that misclassification of clinical events, underestimation of event rates driven by underreporting or undercoding by the hospital personnel cannot be excluded. Additionally, we could not analyze any possible repeated hospitalizations of the same patient. A further limitation of our study is the focus on the in-hospital period without assessment of events after discharge, which may also potentially lead to an underestimate of the total number of postoperative VTE events. Due to coding reasons, we were not able to distinguish between unilateral and bilateral KJR. Nevertheless, it has been published that the proportion of bilateral KJR during one hospitalization is < 4% in Germany 51 . Neither prior VTE, nor anticoagulant, antiplatelet treatment or tranexamic acid use were assessed in the German nationwide inpatient sample. Thus, the potential impact of these factors on occurrence of VTE after KJR could not be analyzed, which has to be mentioned as a further limitation of our study 52 . As per routine clinical practice, venous ultrasound to diagnose DVT was primarily performed in symptomatic patients. It must be considered that the results from the German nationwide sample may not be generalizable for other geographic regions, countries, and health systems.
In conclusion, perioperative VTE was recorded in 1. . For this study, we selected and included all surgical patients with elective primary KJR hospitalized in the timeframe between January 2005 and December 2016 in German hospitals. Since we intended to focus only on hospitalized patients with elective primary procedures, we excluded patients with acute fractures. The study flow was described in part previously [53][54][55][56] . The study results are not part of the routine work of the RDC, but is an analysis especially done for our needs and proposed contents (see Fig. S1 in the supplementary material for study flow chart).
Ethical aspects and study oversight. As described above, the investigators had no access to data of individual patients, but only to aggregated results provided by the RDC. Thus, approval by an ethics committee and patients' informed consent were not required, in accordance with German law. The study was prepared according to STROBE recommendations. In our study, we included all hospitalizations, which refer to surgical patients undergoing elective primary KJR in Germany from the year 2005 to the year 2016. All hospitalized patients with primary KJR were identified based on the OPS code 5-822, after exclusion of patients with revision-surgery or replacement of existing prostheses (OPS-code 5-823) as well as acute fractures (excluding patients with distal femur fracture (ICD-code S72.4) and proximal tibia fracture (ICD-code S82.1) 52 .
Study outcomes. The outcomes of this study were defined as VTE (including deep venous thrombosis or thrombophlebitis (DVT, ICD codes I80, I81, I82) and/or PE (ICD code I26)), prolonged hospitalization (of more than 14 days) and death from any causes, respectively recorded during hospital stay.
Statistical methods. Descriptive statistics for patient characteristics of KJR patients with and without perioperative VTE were provided as median and interquartile range (IQR) or absolute numbers and corresponding percentages. The Mann-Whitney-U test was used to test the continuous variables of the groups regarding differences and categorical variables were compared with chi 2 test or Fisher's exact test, as appropriate 52 .
Scientific Reports | (2020) 10:22440 | https://doi.org/10.1038/s41598-020-79490-w www.nature.com/scientificreports/ We analysed total numbers and incidence of VTE events in elective KJR in Germany between the years 2005 and 2016 and tested for temporal trends of VTE incidence and in-hospital death rate using linear regression models. Results are presented as estimated slope beta (β)-estimates and corresponding 95% confidence intervals (CI) 35 .
In addition, we analysed univariate and multivariable logistic regression models in order to detect influences regarding possible predictors of VTE events. Furthermore, we analysed the association between VTE and inhospital death as well as VTE and prolonged hospitalization (of more than 14 days). These results of our study are presented as odds ratio (OR) and 95% CI. Multivariable regression models were adjusted for age, sex, obesity, cancer, coronary artery disease, heart failure, atrial fibrillation/flutter (AF), essential arterial hypertension, chronic obstructive pulmonary disease (COPD), acute and chronic renal failure, and diabetes mellitus. These covariates were selected a priori since they represent known risk factors for VTE and in-hospital death. The software SPSS (version 20.0; SPSS Inc., Chicago, Illinois) was used for computerised analysis. P values of < 0.05 (two-sided) were considered to be statistically significant [53][54][55][56] .