Prevalence, risk characteristics, and prediction of low-dose edoxaban treatment in hospitalized patients: a multicenter, observational cohort study

Shujuan Zhao^1,2,3*Hengfen Dai⁴Jiaxin Chen⁵Ming Ni⁶Wenxing Peng⁷Xiaoyu Li⁸Fen Li⁹Boya Chen¹Haixia Cai¹Yinping Liu¹Song Du¹⁰

• ¹Department of Pharmacy, Henan Provincial People’s Hospital, Zhengzhou, Henan Province, China
• ²People's Hospital of Zhengzhou University, Zhengzhou, Henan Province, China
• ³School of Clinical Medicine, Henan University, Zhengzhou, Henan Province, China
• ⁴Department of Pharmacy, First Affiliated Hospital of Fujian Medical University, Fuzhou, Fujian Province, China
• ⁵Department of Pharmacy, Ningde Municipal Hospital Affiliated to Ningde Normal University, Ningde, China
• ⁶Department of Pharmacy, Fuwai Central China Cardiovascular Hospital, Zhengzhou, Henan Province, China
• ⁷Department of Pharmaceutical, Beijing Anzhen Hopital, Capital Medical University, Beijing, China
• ⁸Department of Pharmacy, Xinxiang Central Hospital, Xinxiang, China
• ⁹Department of Pharmacy, The First People's Hospital of Xianyang, Xianyang, Shanxi Province, China
• ¹⁰Department of Cardiology, Henan Provincial People’s Hospital, Zhengzhou, Henan Province, China

Background: Treatment with a low-dose non-vitamin K antagonist oral anticoagulant (NOAC) is common among hospitalized patients, and a model to predict the need for such treatment would support individualized interventions. This study evaluated the prevalence of low-dose edoxaban treatment and developed and evaluated a model to predict low-dose administration of edoxaban among hospitalized patients.Methods: This study included 1208 inpatients with non-valvular atrial fibrillation (NVAF) or venous thromboembolism (VTE) who were treated with edoxaban. Univariate and multivariate analyses identified variables significantly associated with low-dose edoxaban therapy. Least absolute shrinkage and selection operator (LASSO) regression was used for data dimension reduction and selection of the best variables. A nomogram was built based on the predictive variables for easy visualization. Model performance was evaluated, and the model was further validated internally with 1000 bootstrap resamples.The prevalence of low-dosing edoxaban treatment was 65.98% (797/1208). The predictors of edoxaban included in the final nomogram were age, weight, surgery or operation, anticoagulation indication, the use of antiarrhythmic drugs, anemia, and bleeding history. The model showed good discrimination with an area under the curve value of 0.792. The Hosmer-Lemeshow test showed that the model had satisfactory goodness of fit (χ 2 =10.757, P=0.2158). The calibration curve showed good agreement between predicted and actual probabilities.The developed predictive model for low-dose edoxaban use among hospitalized patients was built using seven readily available variables and showed good performance. This study provides an empirical basis for early detection and intervention using a low-dose NOAC.