Association of Anemia with Venous Thromboembolism in Acutely Ill Hospitalized Patients: An APEX Trial Substudy.

Background: Anemia is a common finding and independent predictor for adverse outcomes in hospitalized patients with medical illness. It remains unclear whether anemia is a risk factor for venous thromboembolism and whether the presence of anemia can refine risk assessment for prediction of venous thromboembolism, thereby adding incremental utility to a validated model.

Methods: In the Acute Medically Ill Venous Thromboembolism Prevention with Extended Duration Betrixaban trial (APEX), 7513 hospitalized medical patients were randomized to receive either betrixaban or standard-of-care enoxaparin for thromboprophylaxis. Baseline hemoglobin concentrations were obtained in 6861 patients, with a follow-up of 77 days. Symptomatic venous thromboembolism events, including symptomatic deep vein thrombosis, pulmonary embolism, and venous thromboembolism-related mortality, were compared between low-hemoglobin and normal-hemoglobin groups (normal range: 12.5-17.0 g/dL for males and 11.0-15.5 g/dL for females). The relationship between anemia and venous thromboembolism events was assessed by fitting a univariable and multivariable logistic regression model composed of thromboprophylaxis and risk factors. Venous thromboembolism risk refinement by hemoglobin measurement was evaluated in the International Medical Prevention Registry on Venous Thromboembolism (IMPROVE) risk assessment model.

Results: Low hemoglobin at baseline was associated with a greater risk of symptomatic venous thromboembolism (relative risk [RR] 1.94 [95% confidence interval, 1.27-2.98]; P = .002), symptomatic deep vein thrombosis (RR 2.29 [1.12-4.68]; P = .019), and nonfatal pulmonary embolism (RR 2.63 [1.22-5.65]; P = .010) but not venous thromboembolism-related mortality (RR 1.47 [0.71-3.04]; P = .30). After adjusting for thromboprophylaxis, history of previous venous thromboembolism, intensive or coronary unit admission, and D-dimer, low hemoglobin (as a categorical or continuous variable) remained associated with an increased likelihood of venous thromboembolism (adjusted odds ratio 1.71 [95% confidence interval, 1.09-2.69]; P = .020). Low hemoglobin also improved risk discrimination and reclassification after inclusion in the IMPROVE model.

Conclusions: Anemia was independently associated with a greater risk of symptomatic venous thromboembolism among acutely ill medical patients despite the provision of thromboprophylaxis. Hemoglobin measurement also improved risk stratification by the IMPROVE venous thromboembolism risk score.