Association of Anticoagulant Therapy with Bleeding in Patients with Chronic Liver Disease: A Case-Cross over Study Using the National Veterans Health Administration Database

Background: Patients with chronic liver disease (CLD) are at a higher risk of venous thromboembolism (VTE) i.e., deep venous thrombosis or pulmonary embolism {Ambrosino, 2017}, and cardioembolic stroke {Kuo, 2017} than people with normal liver function. Anticoagulant therapy is the cornerstone of management for VTE and atrial fibrillation (Afib) with additional stroke risk factors. CLD patients, however, have fragile hemostasis that could tip towards bleeding from anticoagulant therapy. The effects of anticoagulant therapy on risks of variceal and non-variceal bleeding in CLD remain unclear with studies reporting conflicting outcomes {Lee 2015, Chokesuwattanaskul 2019}. CLD patients were excluded from trials of anticoagulants and retrospective studies could underestimate anticoagulant effects due to confounding by contraindication (i.e., CLD patients at considerable risk of bleeding are less likely to have received anticoagulant therapy). To minimize selection bias, we performed a case-cross over study in a national cohort of patients with CLD to quantify the increment in risk of variceal and non-variceal bleeding after exposure to anticoagulant therapy. Aims: To quantify the odds of variceal and non-variceal bleeding in individuals with CLD exposed to anticoagulant therapy for management of VTE or Afib. Methods: We used the national Veterans Administration (VA) data to identify CLD patients (using Nehra 2013 algorithm) with an incident diagnosis of VTE or Afib. We selected those with clinically relevant bleeding events (variceal or non-variceal) within 12 months of Afib/VTE diagnosis, using a validated algorithm combining International Classification of Diseases (ICD) 9/10 codes and positions to identify bleeds. We excluded patients with bleeding events within 5 years prior to Afib/VTE diagnosis. Exposure of interest was ≥ 30-day outpatient prescription for anticoagulant therapy (low-molecular-weight heparin, warfarin, or direct oral anticoagulants). Using a case-cross over design, we defined the case-period as 30 days before the first bleed (whether variceal or non-variceal) and control period as 30 days starting 6 months before the first bleed. We used conditional logistic regression to estimate the odds ratios (ORs) and corresponding 95% CIs for the risk of bleeding associated with anticoagulant prescription in the case versus control period. As each case patient served as his/her own control, we only adjusted for statistically significant time-varying covariates identified in this cohort i.e., antiplatelet therapy and surgery. We performed similar analyses for those with 1) any variceal bleed, 2) any non-variceal bleed during the one-year follow-up after Afib/VTE diagnosis. Results: Our final cohort included 827 CLD patients with incident Afib/VTE who experienced either a variceal or non-variceal bleed within one-year. Median age of the cohort was 68 years, most (98%) were male. Half of the cohort (49%) had alcoholic liver disease and MELD score averaged 10.5. Anticoagulant therapy was prescribed for 32% of the patients and included warfarin (67%), low-molecular-weight heparin (<1%), and direct oral anticoagulants (33%). When comparing the case-period to the control-period, exposure to anticoagulant therapy doubled the risk of bleeding (OR 2.01, 95% CI 1.33-3.03). Use of antiplatelet therapy was not associated with higher risk of bleeding (OR 1.48, 95% CI 0.91-2.40) whereas surgery increased the risk of bleeding (OR 7.94, 95% CI 4.34-14.51), consistent with prior literature. Analysis of 719 CLD patients who experienced at least one variceal bleed within one-year after Afib/VTE diagnosis revealed comparable results for association between anticoagulant exposure and bleeding (OR 1.97, 95% CI 1.15-3.40). In this cohort, 393 patients experienced at least one non-variceal bleeding event within the one-year after Afib/VTE diagnosis. A similar association between anticoagulant therapy and odds of non-variceal bleeding events was observed (OR 1.82, 95% CI 1.05-3.13). Conclusions: Exposure to anticoagulant therapy doubles the odds of variceal and non-variceal bleeding in CLD patients with Afib or VTE. This information can guide clinical decisions and future research focused on anticoagulant optimization.