Universal pharmacological thromboprophylaxis for total knee arthroplasty may not be necessary in low-risk populations: a nationwide study in Taiwan

Authors

  • C.-H. LEE,

    1. Department of Internal Medicine, National Cheng Kung University Hospital, Tainan
    2. Institute of Clinical Pharmacy and Pharmaceutical Sciences, College of Medicine, National Cheng Kung University, Tainan
    Search for more papers by this author
  • C.-L. CHENG,

    1. Institute of Clinical Pharmacy and Pharmaceutical Sciences, College of Medicine, National Cheng Kung University, Tainan
    Search for more papers by this author
  • C.-H. CHANG,

    1. Institute of Clinical Pharmacy and Pharmaceutical Sciences, College of Medicine, National Cheng Kung University, Tainan
    Search for more papers by this author
  • Y.-H. KAO YANG,

    1. Institute of Clinical Pharmacy and Pharmaceutical Sciences, College of Medicine, National Cheng Kung University, Tainan
    Search for more papers by this author
  • L.-J. LIN,

    1. Department of Internal Medicine, National Cheng Kung University Hospital, Tainan
    2. Institute of Clinical Pharmacy and Pharmaceutical Sciences, College of Medicine, National Cheng Kung University, Tainan
    Search for more papers by this author
  • T.-C. LIN,

    1. Institute of Clinical Pharmacy and Pharmaceutical Sciences, College of Medicine, National Cheng Kung University, Tainan
    Search for more papers by this author
  • C.-Y. YANG

    1. Department of Orthopedics, National Cheng Kung University Hospital, Tainan, Taiwan
    Search for more papers by this author

Yea-Huei Kao Yang, Institute of Clinical Pharmacy and Pharmaceutical Sciences, College of Medicine, National Cheng Kung University, University Road, Tainan, Taiwan.
Tel.: +886 6 2353535 ext. 5688; fax: +886 6 2373149.
E-mail:yhkao@mail.ncku.edu.tw.

Abstract

Summary. Background: Thromboprophylaxis should be universally administered in major orthopedic surgery. However, epidemiology of venous thromboembolism (VTE) following major knee surgery in Asia is scarce. Objective: To describe the use of thromboprophylaxis and calculate the incidence and risk factors of symptomatic VTE following major knee surgery in Taiwan. Methods: We used Taiwan’s National Health Insurance Research Database to retrospectively identify patients (≧45 years) who underwent major knee surgery from 1998 to 2007 and collected the medical records within 3 months after the discharge. Logistic regression analysis was used to determine the risk factors of symptomatic VTE after the surgery. Results: We identified 113 844 patients (mean age, 69.0 ± 7.7 years; female, 75.2%) receiving major knee arthroplasties. The mean length of stay was 9.1 ± 3.3 days. The overall pharmacological thromboprophylaxis rate was 2.2%. The 3-month cumulative incidence of procedure-related symptomatic VTE was 0.46% (95% CI, 0.42–0.50%). The median time to the first post-operation VTE was 7 days, with 85.4% occurring within 2 weeks after the discharge. Logistic regression analysis showed that previous VTE, malignancy, heart failure and neurologic disorder with extremity paralysis or pararesis were independent risk factors (P < 0.05) for symptomatic VTE following major knee arthroplasties. Conclusions: The thromboprophylaxis rate is low, which may be due to the very low incidence of symptomatic VTE after the surgery in Taiwan. Most symptomatic VTE occurred within 2 weeks after the surgery. Universal thromboprophylaxis for knee arthroplasties may not be necessary in Taiwan, but it should be considered in some high-risk populations.

Introduction

Venous thromboembolism (VTE) is a major public health concern in Western countries. The estimated annual incidence is 71–117 cases per 100 000 persons [1–3]. Generally, the incidence of VTE is believed to be lower in Asia than in Western countries [4–6]. Our recent nationwide population-based study shows that the crude incidence of VTE is only 15.9 cases per 100 000 persons and patients with VTE history have a higher probability of suffering recurrent VTE if they underwent major orthopedic surgery without pharmacological thromboprophylaxis [7]. Major orthopedic surgery has been identified as a uniformly high-risk event for VTE. The American College of Chest Physicians (ACCP) [8] guideline published in 2008 states that the incidence of hospital-acquired VTE is approximately 40–60% following major orthopedic surgery without thromboprophylaxis. Therefore, routine pharmacological thromboprophylaxis is recommended in major orthopedic surgery. However, The American Association of Orthopedic Surgeons (AAOS) recommends routine preoperative assessment for risk of pulmonary embolism (PE) and bleeding and stratifies recommendations for thromboprophylaxis according to whether patients have a standard risk or high risk of PE and bleeding [9]. The AAOS does not consider deep vein thrombosis (DVT) as a major cause of PE. Even though the ACCP guidelines are widely adopted in Western countries and worldwide, pharmacological thromboprophylaxis in Asia is not routine, as the incidence of VTE is generally thought to be low and wound bleeding is a major concern. However, several studies recently challenged this view. In these studies, the incidence of venographically detected DVT in Asian patients undergoing major orthopedic surgery without pharmacological thromboprophylaxis ranged from 11.3% to 76.5% after total knee replacement [10–16]. One international study (SMART study) [17,18] in Asia disclosed that asymptomatic VTE following major knee replacement was up to 49% and symptomatic VTE was 1.0%. However, these results were obtained from small-scale studies, conducted on heterogeneous and often poorly characterized populations. We therefore performed a nationwide population-based study to determine the incidence of symptomatic VTE during hospitalization and 3-month follow-up in Taiwanese patients undergoing major knee surgery, including knee replacement and revision of knee replacement. Additionally, the secondary objective was to determine risk factors for VTE events among these populations.

Methods

Database

This study used claims data from the 1997–2007 National Health Insurance Research Database (NHIRD) provided by the National Health Research Institute in Taiwan. The NHIRD includes data on every inpatient admission covered under the NHI program, which has enrolled nearly 99% of the Taiwanese population (23 million residents) and contracted with 97% of hospitals and clinics throughout the nation [19]. The Bureau of NHI performs auditing reviews on a random sample of one per every 100 ambulatory claims and one per 20 inpatient claims quarterly, and false reporting of diagnostic information results in a severe penalty from the Bureau [20,21]. The databases used in this study included all inpatient and outpatient medical claims between 1 January 1997 and 31 December 2007. From the databases, we can retrieve medical information regarding disease diagnosis, prescription drugs, procedures and surgery incurred during a hospitalization or at an outpatient visit. For electronic processing by the National Health Insurance in Taiwan, all the health care service providers are requested to submit the diagnosis information using the International Classification of Disease-Clinical Modification, ninth revision (ICD-9-CM) together with service claims.

Study design

For calculating the percentage of pharmacological thromboprophylaxis in patients receiving major knee arthroplasty surgery and the 3-month cumulative incidences of procedure-related VTE, we performed a retrospective review of outcomes using a nationwide NHIRD database to identify 113 844 adult patients who were hospitalized for knee replacement or revision between 1 January 1998 and 30 September 2007 (Fig. 1); this admission for knee surgery was defined as the index hospitalization. The inclusion criteria were all patients ≥ 45 years of age with the treatment codes of ICD-9-CM 81.54 and 81.55. These patients had been followed for 3 months after the index hospitalization and data were censored at the date of the VTE event, the date of death, or the end of the follow-up period. The exclusion criteria included (i) patients who stayed in the hospital for fewer than 3 days or longer than 21 days and (ii) incomplete electronic medical records. To avoid underestimation of fatal events from pulmonary embolism, patients who received knee replacement surgery and died within 3 days without obvious etiologies were enrolled and regarded as fatal pulmonary embolism possibly related to procedure.

Figure 1.

 The flow diagram of the population-based retrospective cohort study.

Comorbid diseases and potential risk factors of VTE

For each patient, the comorbidities for VTE were retrieved from both the inpatient and outpatient claims databases for 1 year before and during the index hospitalization. The comorbidities including previous VTE and procedures predisposing patients to VTE were recorded using ICD-9-CM codes (Table S1). History of VTE was defined as being hospitalized due to VTE before the index hospitalization. Chronic lung disease included emphysema, chronic bronchitis, bronchiectasis, other obstructive pulmonary disease, and chronic respiratory failure. For neurological diseases, we only recorded serious illness, including stroke or other central and peripheral nervous disease associated with extremity paresis or paralysis. Hormone therapy included the use of estrogen and/or progesterone in hormone replacement therapy and oral contraceptives. In this study, pregnancy and hormone therapy were categorized as potential risk factors only if documented within 3 months preceding the index surgery. Prior VTE and malignancy were also regarded as potential risk factors.

Exposure to drugs

We used prescription records to ascertain the status of drug use. In Taiwan, anticoagulant drugs included warfarin, unfractionated heparin, and low-molecular-weight heparin (LMWH). In our analysis, whether the medication was used for prophylaxis or treatment primarily depends on whether there was an accompanied VTE diagnosis. Namely, patients were classified as having acute VTE when they had the discharge diagnosis of VTE in the same operation course, received warfarin and heparin (intravenous heparin or subcutaneous LMWH) or thrombectomy, received duplex or computed tomography examination during the same hospitalization, and patients should continue receiving oral warfarin after the discharge. On the other hand, when patients only received warfarin or heparin during the operation course without discharge diagnosis of VTE and anticoagulant drugs within 2 weeks after discharge, they were defined as receiving pharmacological prophylaxis (including warfarin or heparin). Because it is hard to titrate warfarin, it is rare to use warfarin for VTE prophylaxis in orthopedic surgery in Taiwan. In this study, the presence of other drugs of interest, including statins, antiplatelet agents and analgesics, was recorded only if they were documented within 1 month preceding the surgery. The antiplatelet agents included aspirin, ticlopidine and clopidogrel. The analgesic agents included opioids and non-steroidal anti-inflammatory drugs (NSAIDs). Hormone therapy included the use of estrogen and/or progesterone in hormone replacement therapy and oral contraceptives.

Study outcome

The primary study outcome was the incidence of symptomatic VTE (defined as deep vein thrombosis, pulmonary embolism, or both). The secondary study outcome was the composite of VTE and overall mortality within the 3-month follow-up. In our study, symptomatic VTE was identified from the inpatient and outpatient claims database by an ICD9-CM code of 451.1x; 451.2; 451.83; 453.1; 453.2; 453.4x; 453.8; 453.9; 415.1x. To avoid misdiagnoses, we only selected inpatients who met the following criteria: (i) the discharge diagnosis was DVT or PE; (ii) the patient received a course of subcutaneous or intravenous anticoagulation therapy with unfractionated heparin or surgical thrombectomy during hospitalization and continued oral warfarin therapy after discharge; and (iii) a length of stay of at least 3 days, unless the patient died. We also selected outpatients who met the following criteria: (i) the principle diagnosis was DVT or thrombophlebitis; and (ii) the patient received a course of subcutaneous anticoagulation therapy with LMWH and continued oral warfarin therapy. During the follow-up period, a new VTE event was defined as hospitalization with a discharge diagnosis of VTE or outpatient diagnosis of DVT plus the above-mentioned criteria.

Statistical analysis

During the study period, the population of Taiwan was about 22.4 million (16.8 million adults). Demographic data were expressed as means (± SD) or percentages. The cumulative incidence of DVT and PE was determined within 3 months of major knee surgeries performed between 1 January 1998 and 30 September 2007. The primary outcomes were a principal or secondary diagnosis of DVT or PE within 3 months (91 days) of the day of surgery. We also calculated the cumulative rates of the secondary outcomes (VTE and all-cause mortality) within 3 months after the surgery. For subgroup analysis, we calculated the incidence of primary and secondary outcomes following the surgery separately. The risks of documented symptomatic VTE at hospital discharge and at 3-month follow-up were analyzed by multiple logistic regression. Univariate analysis was performed to screen potential variables for inclusion in the final multivariate model. Variables significant at the P < 0.05 level in the Wald chi-square test in the final multivariate analysis were regarded as independent predictive factors. We assessed the robustness of our results based on the different lengths of hospital stay (up to 42 days) and repeated all the same analyses including multiple logistic regression analysis. The hazard ratio and associated 95% CI for the various potential risk factors were calculated. We used SAS statistical software (version 9.1; SAS Institute Inc., Cary, NC, USA) for the claims data conversion and analysis.

Results

Baseline characteristics and pharmacological thromboprophylaxis rate

Of the total 113 844 hospitalized patients (≧ 45 years) undergoing major knee surgery between January 1998 and September 2007, the mean age (± SD) was 69.0 (± 7.7) years old and 57 126 (50.2%) were over 69 years old. Among these patients, 85 611 (75.2%) were women and 107 810 (94.7%) received total knee replacement and the others received revision knee arthroplasty. Within 1 month before surgery, 14 117 (12.4%) patients received antiplatelet agents, 99 272 (87.2%) received non-steroidal anti-inflammatory drugs, and 797 (0.7%) received opioid agents. The crude thromboprophylaxis rate was 2.2%, including subcutaneous or intravenous anticoagulant therapy (2.04%) and oral warfarin (0.16%). There were significantly lower percentages of antiplatelet and NSAID use in patients with new VTE events compared with those without new VTE events. However, we did not observe a significant difference in thromboprophylaxis rate between these two groups (Table 1).

Table 1.   Overall clinical characteristics of patients receiving major knee surgery from January 1998 to September 2007
CharacteristicsOverall n = 113 844VTE
n = 520
Non-VTE n = 110 075P value
  1. Data are presented as number (percentages) or mean values ± SD.

  2. P value compared with VTE and non-VTE groups using a Student’s t-test or chi-squared test.

  3. VTE, venous thromboembolism; DVT, deep vein thrombosis; PE, pulmonary embolism; NSAID, non-steroidal anti-inflammatory drugs.

  4. *Stroke includes ischemic and hemorrhagic stroke.

  5. Neurologic disease means stroke or other central and peripheral nervous disease with associated extremity paresis or paralysis.

  6. Thromboprophylaxis means in-hospital use of unfractionated heparin, low-molecular-weight heparin or warfarin.

Mean age (± SD), years69.0 ± 7.769.2 ± 8.169.0 ± 7.70.511
Female885 570 (75.2)379 (74.4)85 191 (75.2)0.722
History of DVT2282 (2.0)120 (23.1)2162 (1.9)< 0.001
History of PE330 (0.3)12 (2.3)318 (0.3)< 0.001
Cancer3769 (3.3)42 (8.1)3727 (3.3)< 0.001
Hypertension73 856 (64.9)369 (71.0)73 487 (64.8)0.004
Heart failure19 588 (9.3)81 (15.6)10 507 (9.3)< 0.001
Coronary heart disease32 376 (28.4)171 (32.9)32 205 (28.4)0.025
Renal insufficiency12 528 (11.0)86 (16.5)12 442 (11.0)< 0.001
Chronic lung disease29 868 (26.2)147 (28.3)29 721 (26.2)0.294
Diabetes mellitus30 686 (27.0)148 (28.5)30 538 (26.9)0.457
Stroke*7710 (6.8)42 (8.1)7668 (6.8)0.254
Neurologic disease15 745 (13.8)95 (18.3)15 650 (13.8)0.004
Varicose vein1786 (1.6)23 (4.4)1763 (1.6)< 0.001
Recent statin use3328 (2.9)13 (2.5)3315 (2.9)0.606
Antiplatelet use14 088 (12.4)101 (19.4)13 987 (12.3)< 0.001
NSAID use99 246 (87.2)425 (81.7)98 821 (87.2)< 0.001
Narcotic use816 (0.7)4 (0.8)812 (0.7)0.792
Thromboprophylaxis2491 (2.2)6 (1.2)2485 (2.2)0.076
Length of hospital stay, days9.1 ± 3.311.0 ± 3.89.0 ± 3.3< 0.001

Clinical outcomes

During the 113 708 patient-months follow-up, the overall incidence of symptomatic VTE events (n = 520) was 0.46% (95% CI, 0.42–0.50%) within 3 months after surgery, which included 73.8% deep vein thrombosis, 21.1% pulmonary embolism, and 5% coexisting deep vein thrombosis and pulmonary embolism. Of symptomatic VTE events, 148 (28.5%) happened during the hospitalization and eight died of presumed PE within 3 days. Among all patients (n = 113 844) receiving the major knee surgery, the median length of stay was 9.1 days. However, among those patients who developed VTE events after the discharge, the median time to developing VTE events was 7 days after the discharge. Three hundred and seventy-two (71.5%) developed VTE after discharge and most VTE events (n = 323, 86.9%) occurred within 15 days after surgery (Fig. 2). The overall incidence of VTE was 0.47% (95% CI, 0.40–0.55%) in men and 0.45% (95% CI, 0.40–0.50%) in women. While the incidence was 0.47% (95% CI, 0.41–0.52%) in the elderly (aged more than 69 years), it was 0.26% (95% CI, 0.01–0.55%) in those younger than 50 years. In subgroup analysis, we found that the overall cumulative rate of VTE events within 3 months after surgery was 5.26% (95% CI, 4.35–6.17%) in patients with previous DVT, 3.64% (95% CI, 1.62–5.66%) in previous PE, 1.11% (95% CI, 0.77–1.45%) in cancer, and 0.77% (95% CI, 0.60–0.94%) in heart failure (Fig. 3). The overall cumulative rate of VTE in other disease-specific populations was 0.49% in chronic lung diseases, 0.50% in hypertension, 0.48% in diabetes mellitus, 0.53% in coronary heart disease, and 0.69% in renal insufficiency. The cumulative VTE incidence in patients undergoing knee replacement was similar to those receiving revision knee arthroplasty (0.45% vs. 0.50%, P = 0.632). The mortality rate was 0.1% (n = 119) during the initial hospitalization, and the cumulative mortality rate was 0.12% (n = 136) within 3 months after surgery. Among these 520 VTE patients, the overall mortality was 9.0%, with 3.4% in DVT and 24.9% in PE. Patients with history of PE (0.9%), cancer (0.3%), heart failure (0.33%), renal insufficiency (0.45%), coronary heart disease (0.32%) and new PE events (24.9%) had higher mortality rates than overall populations (0.12%) (Fig. 3).

Figure 2.

 Time of readmission for symptomatic venous thromboembolism in 509 patients within the 3-month period following major knee surgery.

Figure 3.

 Cumulative incidences of readmission for venous thromboembolism and death among different patient groups within the 3-month period following major knee surgery. DVT denotes deep vein thrombosis; PE, pulmonary embolism; HF, heart failure; ND, neurologic disorder with extremity paralysis or pararesis; elderly, age ≧ 70 years.

Predictors for operation-related venous thromboembolism

Potential VTE risk factors were prior DVT (2.0%), prior PE (0.3%), malignancy (3.3%), serious neurologic disease (13.8%), heart failure (9.3%) and varicose veins (1.6%). In the research population, the percentage of primary hypercoagulable states, including activated protein C resistance, antithrombin III deficiency, factor V Leiden mutation, lupus anticoagulant, protein C deficiency, protein S deficiency and prothrombin gene mutation, was very low (0.4%). Univariate analysis showed that there were significantly higher percentages of underlying diseases such as history of VTE, cancer, cardiovascular diseases, renal insufficiency, serious neurologic diseases and varicose veins in patients with new VTE events than those without VTE events (Table 1). In primary analysis, several baseline characteristics and potential risk factors were compared between the VTE and non-VTE groups by multiple logistic regression analysis (Table 2). In our study, gender, hypertension, coronary heart disease, chronic lung disease, stroke, diabetes mellitus, varicose veins and hypercoagulable states were not associated with procedure-related VTE. For women, there was no difference in the prevalence of hormone therapy including estrogen and/or progesterone use (6.1% vs. 6.0%, P = 0.926) between VTE and non-VTE groups. Finally, multivariate logistic regression analysis shows that previous DVT (hazard ratio [HR], 13.16; 95% confidence interval [CI], 10.53–16.39), previous PE (HR, 3.15; 95% CI, 1.69–5.92), malignant neoplasm (HR, 2.23; 95% CI, 1.61–3.08), heart failure (HR, 1.41; 95% CI, 1.09–1.81) and serious neurologic diseases (HR, 1.26; 95% CI, 1.03–1.63) were independent risk factors for developing VTE (Table 2).

Table 2.   Results of multivariate logistic regression analysis of variables associated with symptomatic venous thromboembolism
VariablesPrimary analysisAdditional analysis*
Hazard ratio (95% CI), P valueHazard ratio (95% CI), P value
  1. *We repeated multiple logistic regression analysis based on the different lengths of hospital stay (up to 42 days).

  2. Neurologic disease means stroke or other central and/or peripheral nervous disease with associated extremity paresis or paralysis.

  3. Pharmacological VTE prophylaxis means oral warfarin or unfractionated heparin or low-molecular-weight heparin during the hospitalization.

Deep vein thrombosis history13.16 (10.53–16.39), < 0.00112.11 (9.68–15.89), < 0.001
Pulmonary embolism history3.15 (1.69–5.92), < 0.0012.65 (1.73–3.98), 0.001
Cancer2.23 (1.61–3.08), < 0.0012.34 (1.49–3.77), 0.002
Varicose veins1.46 (0.94–2.26), 0.0911.41 (0.90–2.37), 0.087
Heart failure1.41 (1.09–1.81), 0.0091.39 (1.19–1.86), 0.015
Neurologic disease1.26 (1.03–1.63), 0.0281.33 (1.07–1.72), 0.027
Hypertension1.11 (0.91–1.36), 0.2911.13 (0.83–1.58), 0.182
Renal insufficiency1.23 (0.96–1.56), 0.0971.21 (0.79–1.17), 0.236
Coronary heart disease1.11 (0.90–1.35), 0.2871.09 (0.88–1.37), 0.428
Pharmacological VTE prophylaxis0.84 (0.65–1.23), 0.3280.83 (0.65–1.19), 0.338

The percentages of clinical morbidities did not change appreciably when we redefined the length of hospital stay extending to 42 days. As expected, the results of predictors for operation-related VTE were virtually unchanged when we performed multivariate logistic regression analysis (Table 2).

Discussion

It is impossible to conduct a large-scale cohort study to observe the clinical course of VTE after major knee surgery among patients with rare pharmacological thromboprophylaxis because pharmacological thromboprophylaxis should be routinely given in these populations. This study was a 10-year nationwide population-based observational cohort study to estimate the pharmacological thromboprophylaxis rate in Taiwanese patients undergoing major knee arthroplasties and the incidence of surgery-related symptomatic VTE within a 3-month follow-up. The main findings were as follows: (i) the rate of pharmacological thromboprophylaxis among patients undergoing major knee surgery was very low (2.3%); (ii) among the 113 844 patients, the 3-month cumulative incidence of surgery-related symptomatic VTE was only 0.46% and the median time to the first post-discharge VTE events was 8 days, with 86.9% occurring within 2 weeks after discharge; and (iii) histories of VTE or malignant neoplasm, serious neurologic diseases with extremity paralysis, and longer hospital stay, were associated with higher risks of VTE occurrence following major knee arthroplasties.

It is widely accepted that deep vein thrombosis frequently develops after major orthopedic surgery (hip and knee replacement) if patients do not receive pharmacological prophylaxis [22,23]. Previous study shows that there is a close relationship between asymptomatic DVT and symptomatic DVT and there is an association between DVT and PE [24,25]. Because it is difficult to predict whether PE will develop in a DVT patient and there is a high prevalence of deep vein thrombosis, ACCP guidelines recommend routine thromboprophylaxis in patients undergoing major orthopedic surgery [8]. In a meta-analysis of 987 patients undergoing general or major orthopedic surgery, LMWH compared with placebo reduced any DVT by 69% and any PE by 61% [23]. In another meta-analysis of 3999 patients undergoing major orthopedic surgery, LMWH compared with placebo reduced venographic DVT by 52%, symptomatic DVT by 59%, and any PE by 57% [26]. However, these studies were mostly conducted among Caucasian populations in Western countries. By contrast, the incidence of VTE in Taiwan and other Asian countries is much lower than that in Western countries [4–7] and thromboprophylaxis is not yet a routine practise in Asia even in high-risk surgical situations [10–17]. One prospective Asian SMART study [17,18] showed that the pharmacological thromboprophylaxis rate was only 3.6% in patients undergoing major orthopedic surgery. Although the incidence of asymptomatic VTE was 35.6% during the hospitalization, the incidence of symptomatic VTE and sudden death within a 1-month follow-up was only 1.5%. Based on the SMART study, most asymptomatic VTE events developing after the major orthopedic surgery possibly spontaneously regress in Asian patients without pharmacological intervention. Similar data from the USA suggests a much lower incidence of VTE among Asian/Pacific Islanders, and that risk stratification tools and recommendations for prophylaxis should incorporate race/ethnicity. However, we also need more information from other populations such as the Japanese, Koreans and residents of India.

A noteworthy finding in this study was the time course of developing symptomatic VTE in patients undergoing major knee arthroplasty. We found that 71.5% of symptomatic VTE events were diagnosed and readmitted after a discharge. The median time to the first post-operational VTE event among these patients was 7 days and 86.9% occurred within 15 days after the discharge. Although the incidence of symptomatic VTE was much lower in Taiwan than that in Western countries, the natural course of developing symptomatic VTE in these populations was very similar. White et al. and Cha et al. [27,28] also reported that the median time of diagnosis of symptomatic VTE was about 7 days in patients undergoing major knee surgery. These findings may suggest that orthopedic surgeons should be aware of patients’ signs of VTE within 14 days after discharge. Additionally, in high-risk Taiwanese patients, the use of pharmacological thromboprophylaxis for major knee surgery should be extended from the hospital stay to at least 2 weeks after the discharge.

Previous studies showed that prior VTE [29], female gender [27], age over 70 years [1,3], body mass index (BMI) over 25 [30] and longer duration of immobilization [31] are important risk factors for development of symptomatic VTE among patients undergoing major hip arthroplasty. Instead, Asian/Pacific Island ethnicity [32] and the use of extended medical prophylaxis with either a LMWH or warfarin were found to be associated with a reduced incidence of symptomatic VTE [33]. However, few studies have specifically examined risk factors for VTE after knee arthroplasty, except prior VTE and advanced age [27,29]. In our study, histories of VTE, malignant neoplasm and neurologic diseases with extremity paralysis were associated with higher risks of VTE development following major knee arthroplasties. In addition, NSAID use during the perioperative period seemed to decrease the hospital stay, which might have contributed to fewer VTE events. Although the beneficial effect of pharmacological thromboprophylaxis in these populations was not observed in this study, we attributed the neutral effect of pharmacological thromboprophylaxis to the small sample size of patients received the prevention because the proportion of patients received the prevention in the VTE group was nearly half of that among the non-VTE group. We attributed the neutral effect of pharmacological thromboprophylaxis to the small sample size of patients receiving the prevention. Further study is warranted to examine the effect of pharmacological thromboprophylaxis in Taiwan’s populations.

Our findings have several clinical implications for Asian populations. First, although the rate of pharmacological thromboprophylaxis was very low in patients undergoing major knee arthroplasty in Taiwan, symptomatic VTE rarely developed. One prospective Asian study showed that the pharmacological thromboprophylaxis rate was only 3.6% in patients receiving major orthopedic surgery, but the incidence of VTE and sudden death at a month’s follow-up was only 1.5% [17]. However, our findings suggest that adequate thromboprophylaxis around major knee arthroplasties in high-risk populations, especially those with VTE history, may reduce surgery-related VTE incidence. Because most VTE events develop within 2 weeks after the surgery, patients should be educated and be mobilized as soon as possible and surgeons should be alert if persistent swelling of the diseased extremity exists beyond 1 week after the discharge.

There are several limitations in the present investigation. First, the healthcare claims data inherently contain potential disease misclassification bias. However, the auditing mechanism conducted by the Bureau of National Health Insurance [20,21] would help to minimize the diagnostic uncertainty and misclassification in claims databases. Second, whether the patient received pneumatic compression after the surgery was indeterminate. Third, there are no ICD-9 codes for new and prior VTE, separately. Therefore, we established the strict criteria of new and prior VTE in this study. Using the definition of new VTE might include a few patients with recent VTE before the major knee surgery and who were taking anticoagulants. This would overestimate the rate of VTE. If we excluded patients with prior VTE in this study, the incidence of symptomatic VTE was only 0.34%, which was amazingly low. Besides, the true incidence of VTE following major knee surgery in Taiwan may be underestimated because some patients with PE may die suddenly without accurate diagnosis and some symptomatic VTE patients may be treated as surgery-related leg edema. To avoid underestimating fatal pulmonary embolism, we also calculated the all-cause mortality within 3 months after the discharge. Given the mortality rate was only 0.12%, the bias should be low. Source of bias, such as acute asymptomatic VTE was not coded as new symptomatic VTE, might exist in this study. However, physicians would not prescribe duplex or computed tomography CT examination if patients had no symptoms of VTE because major knee replacement was regarded as case-payment procedure in Taiwan. Therefore, this bias could be neglected. Finally, information about travel, smoking history and body mass index was not available in the databases.

In conclusion, the rate of pharmacological thromboprophylaxis is very low in Taiwan, which may be related to the low incidence of symptomatic VTE after major knee arthroplasties. Most symptomatic VTE events occurred within 2 weeks after the surgery. Although universal pharmacological thromboprophylaxis is not recommended generally in Taiwan, it should be considered in high-risk populations that include a history of VTE, cancer, heart failure, and neurologic diseases with extremity paralysis.

Addendum

C. H. Lee initiated and designed the study, prepared the data, conducted the analysis and interpretation, and wrote the first draft of the paper. L. J. Lin contributed to the development of the protocol, design, analysis and interpretation, and to drafting the paper. C. L. Cheng contributed to the design of the study, writing the protocol, interpreting results and drafting the article. T. C. Lin, C. H. Chang undertook the data processing and statistical analysis. Y. H. Kao Yang and C. Y. Yang coordinated the execution of the study, established the strategy of data processing of the claims data and participated in the manuscript preparation. All authors have approved the final draft.

Acknowledgements

This study was funded by the Department of Health, Taiwan (DOH098-TD-D-113-098002) and Multidisciplinary Center of Excellence for Clinical Trial and Research, Department of Health, Executive Yuan, Taiwan (DOH100-TD-B-111-002). The funding organizations did not play a role in the design, conduct or analysis of this study or decision to submit the manuscript for publication.

Disclosure of Conflicts of Interests

The authors state that they no conflict of interest.

Ancillary