Deep Venous Thrombosis and Risk of Consequent Sepsis Event: A Retrospective Nationwide Population-Based Cohort Study

Deep vein thrombosis causes several acute and chronic vessel complications and puts patients at risk of subsequent sepsis development. This unique study aimed to estimate the risk of sepsis development in DVT patients compared with non-DVT patients. This population-based cohort study used records of a longitudinal health insurance database containing two million patients defined in Taiwan’s National Health Insurance Research Database (NHIRD). Our study included patients aged over 20 years with a new diagnosis of DVT with at least two outpatient department visits or an admission between 2001 and 2014. Patients with a diagnosis of sepsis before the index date were excluded. Propensity score matching (PSM) was used to homogenize the baseline characteristics between the two groups. To define the independent risk of the DVT group, a multivariate Cox proportional hazard model was used to estimate the hazard ratios. After PSM, the DVT group (n = 5753) exhibited a higher risk of sepsis (adjusted hazard ratio, aHR, 1.74; 95% CI, 1.59–1.90) compared with non-DVT group (n = 5753). Patients with an increased risk of sepsis were associated with being elderly aged, male, having diabetes, chronic kidney disease, chronic obstructive pulmonary disease, stroke, malignancy, and use of antibiotics. In conclusion, this population-based cohort study demonstrated an increased risk of sepsis in DVT patients compared with non-DVT patients. Thus, early prevention and adequate treatment of DVT is necessary in clinical practice.


Introduction
Deep vein thrombosis (DVT) is a clinical manifestation of venous thromboembolism (VTE) and can contribute to severe morbidity, mortality, and economic burden as well. Even though often underestimated, there were more than 200,000 deep vein thrombosis Int. J. Environ. Res. Public Health 2021, 18, 7879 2 of 11 cases annually in the United States [1] and about 500,000 in European countries [2]. In Taiwan, the crude incidence of VTE was 15.9 per 100,000 person-years, with a higher risk of recurrence in people with a history of VTE, cancer, major extremity trauma, or undergoing major surgery [3]. The economic burden of the initial VTE is approximately $3000-9500, equivalent to $33,000 for each individual [4] and estimated to be $7-10 billion in the United States [5].
The major theory illustrating the pathogenesis of DVT, known as the Virchow's triad, includes three main mechanisms of venous stasis, endothelial damage, and a hypercoagulation state [6]. Clinical presentations of DVT are nonspecific, and many patients are asymptomatic. However, DVT should be suspected in patients who present with classical symptoms of pain, warmth, leg swelling, and erythema. The common risk factors of DVT include older age, history of DVT, obesity, active cancer [7], hormone replacement therapy [8], surgery [9], trauma, hospitalization, nursing home residence, and paralysis [10].
On the other hand, when dealing with patients with sepsis, especially those complicated by hypotension and septic shock in clinical practice, DVT was often considered one of those troublesome complications [12,13]. Kaplan et al. performed a multicenter prospective study and showed a 37.2 (95% CI, 28.3-46.8) incidence of VTE in patients with severe sepsis and septic shock, despite using guideline-recommended medications for thromboprophylaxis [14]. Aside from the acting complications of sepsis, infection and sepsis resulting from DVT is an important issue but infrequently mentioned and studied.
DVT causes both acute and chronic complications associated with sepsis development. The consequence of chronic venous insufficiency resulting from DVT, manifesting as mild limb swelling, intractable edema, to severe leg venous ulceration, is known as postthrombotic syndrome (PTS) [15]. Even under adequate anticoagulant medications, PTS develops in about 20% to 50% of DVT patients [16]. Chronic venous ulceration could further develop a local infection, cellulitis, and even sepsis. In contrast to the chronic complication of DVT, acute DVT could also cause an emergent condition called phlegmasia cerulea dolens (PCD) [17]. PCD is a life-threatening complication of acute DVT, with cascading symptoms from extremities swelling and pain, pale, tissue ischemia, cyanosis to necrosis and sepsis, especially when accompanied with compartment syndrome [18].
Nevertheless, most studies focused on the risk stratifications of DVT and prevention strategies rather than assessment risk of subsequent infection or even sepsis caused by DVT. Presently, no studies have conducted an evaluation of the risk of sepsis in patients with DVT or subgroups linked with a higher risk and poor prognosis due to sepsis. Hence, this study examined whether DVT increases the risk of sepsis and their complications by executing a present nationwide population-based cohort study using the National Health Insurance Research Database (NHIRD).

Data Sources
The longitudinal health insurance database is regulated by the Health and Welfare Data Science Center (HWDC) in Taiwan. The database contains two million beneficiaries that were randomly sampled from the 2000 registry for beneficiaries out of the entire population. The database contains all outpatients' and inpatients' medical claims, as well as details on drugs prescribed, medical operations, procedures, and medical expenses from 2000 to 2015. The cause of death data contains the code and date of death and has been validated by previous research [19]. The study was approved by the ethical review board of the Chung Shan Medical University Hospital (CS1-20056).

Study Design and Outcome
This study is a retrospective cohort study design with a study population based on new diagnoses of deep vein thrombosis (DVT; ICD-9-CM codes = 453.8) and age ≥ 20 years from 2001 to 2014. In order to ensure the accuracy of the diagnoses, outpatient visits of ≥ 2 times or hospitalizations of ≥ 1 time were imposed. The index date was established as the first diagnosis date of deep vein thrombosis. The positive predictive value of DVT codes used in this study was about 90%, as shown in a previous validation study [20]. Diagnoses of sepsis (ICD-9-CM = 038, 995.91, 995.92) before the index date were excluded to ensure new DVT onset patient samples. The non-DVT subjects were defined with no diagnosis of deep vein thrombosis from 2000 to 2015. The outcome variable was defined as a diagnosis of sepsis (ICD-9-CM = 038, 995.91, 995.92) from emergency or admission. A previous validation study also showed the codes used to define sepsis in this study obtained a positive predictive value of 75% (95% CI, 53.0-90.2%) and a negative predictive value of 99.4% (95% CI, 98.3-99.8%) [21]. Both groups were followed up until the onset of sepsis, death, or 31 December 2015, whichever occurred first.

Covariates and Matching
The baseline characteristics were age, sex, hypertension (ICD . These comorbidities were defined before the index date within one year and at least two outpatient visits or one hospitalization. In addition, antibiotics use before the index date, within one year and ≥ 4 times orders were included. The accuracy and positive predictive value of the diagnosis of covariate diseases in the NHIRD has previously been validated [22]. First, a 1:4 matching by age and sex was used to provide an index date for the non-DVT subjects that had the same starting point. Then, propensity score matching was performed by age, sex, comorbidities, and antibiotics use between the two groups ( Figure 1). The propensity score was a probability that was estimated through logistic regression. The binary variable was the DVT and non-DVT groups. By matching the propensity score, it could balance the heterogeneity of the two groups.

Statistical Analysis
Comparison of the DVT group and non-DVT group was performed using absolute standardized differences (ASD). When the absolute standardized difference was less than 0.1, it was defined as the characteristics of both groups being similar [23]. The relative risk (RR) and 95% confidence intervals (CI) were calculated via the Poisson regression model. Kaplan-Meier analysis was used to calculate the cumulative incidence of sepsis among the two groups. The log-rank test was used to test the significance. To identify the independent risk of the DVT group, a multivariate Cox proportional hazard model was used to estimate the hazard ratios. We performed E-value to define the minimum strength of the association for an unmeasured confounding effect between deep venous thrombosis and sepsis [24]. The statistical software was SAS version 9.4 (SAS Institute Inc., Elizabeth City, NC, USA). diagnosis of deep vein thrombosis from 2000 to 2015. The outcome variable was defined as a diagnosis of sepsis (ICD-9-CM = 038, 995.91, 995.92) from emergency or admission. A previous validation study also showed the codes used to define sepsis in this study obtained a positive predictive value of 75% (95% CI, 53.0%-90.2%) and a negative predictive value of 99.4% (95% CI, 98.3%-99.8%) [21]. Both groups were followed up until the onset of sepsis, death, or 31 December 2015, whichever occurred first.

Characteristics of the Study Subjects
We identified 6932 patients who were newly diagnosed with DVT via outpatient visit or admission from 2001 to 2014. After applying exclusion criteria and 1:4 age and sex matching, there were 6141 DVT patients versus 24,564 non-DVT patients. The DVT groups showed a higher proportion of elderly and female patients. Compared with the non-DVT group, the DVT group had a higher proportion of comorbidities, including hypertension, hyperlipidemia, diabetes, ischemic heart disease, chronic kidney disease, stroke, malignancy, and peripheral arterial occlusive disease. The DVT group also had a higher proportion of antibiotics use. We further performed PSM and conclusively included 5753 DVT patients with 5753 control cases for examination. A comparison of the characteristics of the DVT and non-DVT patients is presented in Table 1. Age was on the borderline with no significant difference between the PSM-derived DVT and non-DVT groups (ASD = 0.10). Hypertension was the comorbidity with the highest prevalence in both groups, followed by diabetes, ischemic heart disease, and hyperlipidemia. Sex, comorbidities, and antibiotics use were not significantly different between the PSM-derived DVT and non-DVT groups.

Risk of Sepsis between DVT and Non-DVT Group
After PSM, the incidence of sepsis was 37.25 (95% CI, 35.20-39.41) and 23.59 (95% CI, 22.06-25.24) in the DVT and non-DVT cohorts, respectively. The cohort with DVT presented an increased incidence of sepsis (RR, 1.58; 95% CI, 1.45-1.72) ( Table 2). The Kaplan-Meier survival analysis showed a significantly higher cumulative incidence of sepsis in the DVT group (Log-rank, p <0.01) (Figure 2). Compared with the patients in the non-DVT group, those with DVT presented an increased risk of sepsis (aHR, 1.74; 95% CI, 1.59-1.90). Other significant risk factors of sepsis included older age, male, diabetes, chronic kidney disease, COPD, stroke, malignancy, and use of antibiotics (Table 3). In contrast, hyperlipidemia showed a protective effect on sepsis. As shown in Table 4, subgroup analysis performed on the association of DVT with a significantly higher risk of sepsis included the stratification of age, sex, hypertension, hyperlipidemia, diabetes, ischemic heart disease, chronic kidney disease, COPD, stroke/TIA, malignancy, and antibiotics use. For patients aged 20-39 years in the non-hypertension and non-antibiotic subgroup, DVT subjects performed a significantly higher risk of sepsis than other subgroups (p for interaction <0.01, <0.01, 0.03, respectively). The E-value of the analyses was 2.29 and performed the minimum strength of an unmeasured confounder.

Discussion
This is the first nationwide population-based cohort study conducted to clarify the risk of sepsis in DVT patients, and the findings revealed that DVT is associated with a higher risk of sepsis (aHR, 1.74; 95%CI; 1.59-1.90). In the past decades, DVT was considered a troublesome complication of sepsis and septic shock [11,12]. We discovered an inverse relationship between DVT and sepsis and provided an alternative thinking process for clinical practice. Since most previous studies have emphasized prevention and early medication of DVT with other diseases [25], our study showed the importance of the prevention of infection and sepsis in DVT patients.
Our findings are compatible with the pathophysiology of DVT to PTS, extending to sepsis caused by venous leg ulcers [26,27]. It is estimated that more than 20% of DVT patients develop a chronic clinical presentation of local or segmental edema, cramping pain, redness, hyperpigmentation of the painful area, and eventually ulceration [28]. Two mechanisms are responsible for venous hypertension and the progression from DVT to PTS, including venous obstruction and valvular incompetence due to vein valves damage [15]. PTS is the long-term complication of DVT and causes chronic socioeconomic morbidity [29].
For clinical diagnosis and severity evaluation of PTS, the Villalta scale has been adopted by the International Society on Thrombosis and Haemostasis [30] and is applied worldwide [31] due to its validity and reproducibility throughout the disease process. The scale contains five symptoms and six clinical signs, with each component scored from 0 to 3. Recently, several potential novel biomarkers have been proposed for improving the diagnosis of DVT and PTS [32], including intracellular adhesion molecule-1(ICAM-1) [33], P-selectin [34], and cell-free DNA [35].
Malignancy is a risk factor of both DVT and sepsis, as described and proved in a previous study [36,37]. Our study showed a similar result that malignancy was associated with a higher risk of sepsis (aHR, 1.75; 95%CI, 1.55-1.99). Moreover, our study performed subgroup analysis and found that for both non-malignancy and malignancy groups, patients with DVT in both subgroups obtained a higher risk of sepsis (aHR, 1.74, 2.15, respectively). Compared with non-malignancy patients, malignancy patients with DVT were associated with a slightly greater but not significantly higher risk of sepsis (p for interaction = 0.11).
On the other hand, peripheral arterial occlusive disease (PAOD) is another common vascular disease of distal extremities, causing local pain, ischemia, necrosis, wound infection, spies, and potentially results in amputation. Studies have shown a positive correlation between PAOD and infection [38]. However, PAOD in our study showed a borderline increased risk of sepsis in the univariate model (aHR, 1.41; 95%CI, 1.13-1.76) but a decreased risk of sepsis in the multivariate model (aHR, 0.97; 95%CI, 0.77-1.21). One possible explanation could be the improvement of diagnostic tools, medication, and endovascular intervention of PADO, resulting in less severity of disease progression.
Our study also demonstrated that younger DVT patients were susceptible to a higher risk of sepsis (p for interaction = 0.0093). While the incidence of DVT in young adults is rare [39], the development of DVT at a younger age is related to thrombophilic disorders and anomalies of the inferior vena cava (IVC) [40]. IVC anomalies could affect the blood return of the lower extremities and increase the condition of venous stasis and venous hypertension [41], while thrombophilia is a hypercoagulation state due to the over-activity of coagulation factors. However, the higher correlation between younger DVT patients and sepsis compared with older DVT patients may be due to the different distribution of co-infection between the age subgroups, and more studies are required for a plausible mechanism and confirmation. In addition, this study result showed the protective effect of hyperlipidemia against sepsis, which could be explained as the statins prescriptions for hyperlipidemia patients. Previous studies have demonstrated a similar correlation that statins use was associated with a lower risk of developing sepsis [42][43][44].
Based on the results of our study, the early management of DVT and prevention for sepsis are essential in clinical practice. Multidisciplinary treatment and consultation are necessary, including anticoagulant therapy of low-molecular-weight heparin and non-vitamin K antagonist oral anticoagulant (NOAC), elastic compression stockings [45], infection controls, adequate wound dressing, and a surgical or endovascular approach [46]. This observational study had some limitations. First, the NHIRD did not obtain information on the severity of DVT, patients' activity ability or patients' personal behaviors, such as smoking and obesity, or factors that influence wound healing and risk of thrombosis. Second, the database also lacked the details of initial treatment of DVT and therapy procedures, such as medication treatment or indicated surgical intervention. Third, coexisted infection was not excluded in the study population; however, we used PSM for antibiotics use in the DVT and non-DVT group analysis to minimize potential bias. Fourth, PTS was a set of signs and symptoms and was not defined in the ICD-9-CM codes system, resulting in a lack of PTS data in our study. However, DVT patients developed sepsis mainly via PTS. Since the correlation mechanism between DVT, PTS, and sepsis were defined, the lack of PTS data in our results did not affect the main outcome of the study. Finally, this observational study extracted data from Taiwan's national database. More precisely designed and planned studies with randomized control trials are warranted for further analysis and evaluation of the risk of DVT from sepsis before worldwide application.

Conclusions
In summary, this population-based cohort study with PSM demonstrated that DVT is associated with an increased risk of sepsis, with a protective effect against sepsis in patients of older age, with hypertension, and under antibiotics use. Early intervention and prevention of sepsis complications are essential for clinical practice when caring for DVT patients.
Informed Consent Statement: Patient consent was waived by both the National Health Insurance Administration and the Institutional Review Board of Chung Shan Medical University Hospital due to the database-processing nature of the current study.
Data Availability Statement: Restrictions apply to the availability of these data. Data were obtained from the National Health Insurance database and are available from the authors with the permission of the National Health Insurance Administration of Taiwan.