Cirrhosis is Associated with an Increased 30-Day Mortality After Venous Thromboembolism

Deep venous thrombosis (DVT) is a common medical event with 30-day mortality between 3 and 30%, depending on whether pulmonary embolism (PE) develops.1 By contrast, portal vein thrombosis (PVT) is less common, but a potential serious condition with 30-day mortality varying from ~3 to 50%.2 Patients with venous thromboembolism (VTE) often have underlying comorbidities that may increase their risk of dying from a thrombotic event. In a large population-based cohort study of patients with DVT or PE, we recently examined the effect of several comorbidities on mortality after thrombosis.1 In stratified analyses, only presence of cancer, diabetes, and chronic liver disease yielded higher mortality rates after the thrombotic event, compared with absence of these factors.1 Among patients with PVT, prevalent cancer or cirrhosis are predictors of increased mortality.2, 3 Patients with cirrhosis are at increased risk of DVT and PE compared with the general population.2, 4, 5, 6 This increased risk of thrombosis is likely due to a combination of external factors among cirrhosis patients (immobilization, surgical procedures, severe infections and a high comorbidity burden)7, 8 and intrinsic factors (disturbance of the coagulation system and increased estrogen levels).7, 8, 9 In addition, local factors may result in venous stasis (e.g., compression by a solid tumor, abscess, or by hepato- or splenomegaly) causing PVT. Cirrhosis in itself has a grave prognosis because of cirrhosis-related complications and comorbidities.10, 11 In case of venous thrombosis in patients with cirrhosis, initiation of standard treatment with anticoagulant medications may be impeded considering their increased bleeding tendency. Therefore, it is important to know whether cirrhosis affects mortality after venous thrombosis. We undertook this nationwide cohort study to examine whether cirrhosis affects 30-day mortality after DVT, PE, or PVT, clarifying the clinical course of venous thrombosis among patients with cirrhosis. The DNPR can be linked to the Danish Civil Registration System, which, in addition to issuing civil registration numbers, has monitored deaths and emigration from the country since 1968.15 We used the Danish Register of Causes of Death16 to obtain information on causes of death for patients with VTE. The register contains information from all Danish death certificates since 1943, coded according to the Danish version of the International Classification of Diseases (ICD-8 from 1972 through 1993, and ICD-10 from 1994 through 2011). Based on medical history preceding a hospital contact for VTE or on status at the time of this contact, as recorded in the DNPR, we identified patients with cirrhosis and patients without cirrhosis registered (comparison cohort). Cirrhosis was further classified as alcoholic, biliary (primary, secondary, and non-specified biliary cirrhosis), and other or non-specified cirrhosis. Because of substantial differences in baseline characteristics among patients with cirrhosis and patients in the comparison cohort, we matched the VTE patients with and without cirrhosis by age, gender, calendar year of VTE diagnosis, and type of VTE. We were able to match 96% (n=713) of patients with cirrhosis with five patients each in the comparison cohort. We also examined diagnoses or conditions related to VTE and/or cirrhosis that may affect 30-day mortality after VTE. These were heart failure, chronic pulmonary disease, ulcer disease, diabetes, alcoholism-related disease, psychiatric disorders, and obesity.22, 23, 24, 25 Patients with cirrhosis are prone to infections because of their compromised immune system,26 and prevalent infections have a strong impact on prognosis.27 We therefore also included infections diagnosed during the VTE-related hospital contact (i.e., pneumonia, urinary tract infections, and skin, soft tissue, and bone infections). To further characterize VTE patients with cirrhosis, we collected information on previous or concurrent diagnoses of gastroesophageal varices with and without bleeding. Finally, we retrieved information on post-discharge use of VKA and LMWH from the prescription database. We followed the patients from the date of their first VTE-related hospital contact until date of death from any cause, 30 days of follow-up, emigration, or censoring on 31 December 2011, whichever came first. The Kaplan–Meier survival method was used to compute 30-day mortality risk after DVT, and/or PE, and PVT among patients with and without cirrhosis. We used Cox proportional hazards regression to compute 30-day mortality rate ratios (MRRs) and 95% confidence intervals (CIs) for VTE patients according to presence of cirrhosis. Using log–log plots, we visually confirmed proportionality of hazards for DVT and PE throughout the 30 days of follow-up, whereas there was non-proportionality of the overall follow-up for PVT. Therefore, we divided follow-up after PVT into 0–7 days and 8–30 days. In accordance with the matched design, we also used a stratified Cox regression model (which revealed similar results). We adjusted for gender, age categories, and calendar-year periods (by study design), in addition to the classical risk factors (as described above) and other comorbidities (heart failure, chronic pulmonary disease, ulcer disease, diabetes, alcoholism-related disease, and concurrent infections). We also conducted analyses stratified according to type of cirrhosis (alcoholic, biliary, and other or non-specified), comorbidity level, and cancer (potential effect modifiers). To quantify whether patients with cirrhosis were less likely than their comparisons to receive treatment with anticoagulant medication post discharge, we used χ2-test for homogeneity of proportions. We calculated the proportions of deaths due to PE among patients with and without cirrhosis. For patients with cirrhosis, we described the prevalence of immediate causes of death within 30 days following VTE diagnosis. Statistical analyses were performed using STATA 12.0 (StataCorp LP, College Station, TX, USA) and SAS 9.2 (SAS Institute Inc., Cary, NC, USA). The study was approved by the Danish Data Protection Board (record number 1-16-02-1-08 and 2012-41-0793). Danish registry data generally are available for research purposes, and use of the data does not require informed consent according to Danish law. This is the first nationwide population-based cohort study to report the impact of cirrhosis on 30-day mortality following DVT, PE, or PVT. We found that patients with cirrhosis had higher absolute mortality risks after any thromboembolic event than their matched comparisons, but the risk difference was more pronounced for PE than for DVT and PVT. Patients with cirrhosis also had a higher relative mortality rate after DVT and PE than matched patients without cirrhosis, whereas it was not clear whether cirrhosis patients had higher mortality after PVT than patients in the comparison cohort. PE was the most frequent cause of death within 30 days among patients with cirrhosis and VTE, and most of the deceased had alcoholic cirrhosis. Clearly, the site and extension of a venous thrombosis impact on mortality risk.28 Presence of underlying chronic comorbidities among patients with VTE is also a prognostic factor for mortality after a thrombotic event. We recently examined the effect of several comorbidities on mortality among 128,223 patients with DVT or PE and 640,760 persons from the general population.1 Extensive stratified analyses revealed that among numerous considered comorbidities, only presence of cancer, diabetes, and chronic liver disease resulted in a higher mortality after VTE, compared with absence of these factors.1 In general, patients with cirrhosis have a substantial excess short-term mortality.27, 29, 30, 31 This increased mortality likely stems from a high comorbidity burden,11 increased susceptibility to bacterial infections,32, 33 and complications of cirrhosis.34 In addition, the patients with cirrhosis had a high prevalence of classical risk factors for VTE, but also other comorbidities, particularly alcoholism-related complications. Their risk profile may therefore have impacted on the course of VTE including the choice of treatment. There is still inadequate evidence regarding effectiveness and safety of anticoagulant treatment in patients with cirrhosis and VTE,35, 36, 37, 38 and the establishment of a risk–benefit ratio for pharmacological VTE prophylaxis, and treatment therefore remains a critical problem. Most of the evidence regarding treatment with anticoagulants stems from studies including PVT patients. Treatment patterns with anticoagulants within the first month after splanchnic venous thrombosis were described in a multinational cohort study including 244 patients with isolated PVT.39 Although 81 (33%) patients did not receive treatment, 143 (59%) were treated with LMWH and 77 (32%) patients with VKA, alone or in combination. A larger proportion of patients with active cancer or cirrhosis received prolonged LMWH, which is in agreement with the guidelines for DVT and PE.17, 39 As the frequencies of patients treated with anticoagulants were not provided separately for patients with and without cirrhosis, our results are not comparable. The main strengths of our registry-based study were its size and setting within the uniformly organized Danish health-care system, permitting a nationwide population-based design. A number of limitations must also be considered, including the accuracy of VTE and cirrhosis diagnoses in the patient registries and the ability to control for confounders such as underlying comorbid conditions. The VTE diagnosis in the DNPR has been found to have a positive predictive value of 71% for DVT and 82% for PE18 compared with strict diagnostic criteria (including a combination of typical clinical symptoms in combination with confirmatory diagnostic imaging test results). However, any misclassification of thrombosis diagnoses should not differ between patients with and without cirrhosis (i.e., it is non-differential). The positive predictive value of cirrhosis codes in the DNPR was previously found to be 85%, using either the diagnostic criteria for cirrhosis40 or through comparison with medical charts.41 In regard to confounder control, the positive predictive values of other diseases41 and surgical procedures42, 43 are also high. Another study limitation is that we could not classify patients according to cirrhosis severity because the data necessary for severity scoring are not available in the patient registries. Instead, we stratified patients broadly by type of cirrhosis. Of note, we had only a few cases of hepatitis C-associated cirrhosis, as hepatitis C virus is rare in Denmark.44 Patients with cirrhosis may be frail persons who likely have a high mortality when admitted with any acute illness, and confounding by baseline risk may have impacted our results. We performed comprehensive adjustment for potential confounders, which clearly attenuated the relative VTE mortality risks. Still, we cannot rule out residual confounding that could lead to overestimation of the association between cirrhosis and mortality following VTE. The impact of cirrhosis on relative mortality was more pronounced for DVT than PE, which may reflect a high mortality after PE per se, regardless of underlying disease. In conclusion, during 30 days of follow-up after a diagnosis of DVT, PE or PVT, we found higher mortality risk and rates in DVT and PE patients with cirrhosis than in VTE patients without cirrhosis. PE was the main cause of death among patients with cirrhosis, but the proportion of deaths due to PE was similar to that of other VTE patients. Guarantor of the article: Kirstine Kobberøe Søgaard, MD. Specific author contributions: Study concept and design, analyses, interpretation of data, manuscript writing, manuscript revision, editing, and decision to publish: Kirstine Kobberøe Søgaard; study concept and design, supervising in analyses, interpretation of data, revision of manuscript, editing, and decision to publish: Jonathan Montomoli, Hendrik Vilstrup, Henrik Toft Sørensen; study concept and design, participated in data analysis, interpretation of data, revision of manuscript, editing, and decision to publish: Erzsébet Horváth-Puhó. All authors had full access to all data. Financial support: The study was funded by the Clinical Epidemiology Research Foundation at Aarhus University Hospital and by a grant from the Aarhus University Research Foundation. Potential competing interest: None. Supplementary Information accompanies this paper on the Clinical and Translational Gastroenterology website