Thromboembolic Events While Taking Direct Oral Anticoagulants: An Analysis of Post-market WHO Database Reports from 2012 to 2020.

Abstract

Several cases of venous thromboembolism in patients treated with direct oral anticoagulants (DOACs) have been reported in the literature, but a quantative analysis of postmarketing reports is lacking. The objective of this study was to determine the post-marketing odds ratio (OR) and reporting odds ratio (ROR) of venous thromboembolism in patients receiving DOACs compared among each other and to vitamin K antagonists (VKAs). The OR and ROR were used to determine the ratio of reports for deep vein thrombosis and pulmonary embolism between 1 January, 2012 and 15 November, 2020 using the World Health Organization VigiLyze database. This was performed using all venous thromboembolism events in which a DOAC or a VKA was the suspected medication. The OR and ROR including 95% confidence intervals were calculated for each DOAC drug in comparison to all VKAs as a group. The OR of deep vein thrombosis was highest for rivaroxaban compared with dabigatran and apixaban [2.63 (2.41-2.89); 1.84 (1.72-1.97)]. The OR of deep vein thrombosis was lowest for edoxaban compared with dabigatran, apixaban and rivaroxaban [0.44 (0.32-0.61); 0.31 (0.22-0.42); 0.17 (0.12-0.23)]. The OR of pulmonary embolism was also highest for rivaroxaban compared with dabigatran and apixaban [2.59 (2.37-2.83); 1.79 (1.68-1.92)]. The OR of pulmonary embolism was also lowest for edoxaban compared with dabigatran, apixaban and rivaroxaban [0.77 (0.60-0.97); 0.59 (0.41-0.67); 0.30 (0.23-0.37)]. Comparing RORs of various DOACs with VKAs, rivaroxaban had the highest RORs for deep vein thrombosis/pulmonary embolism, in comparison to apixaban, dabigatran and edoxaban. Our findings may indicate a higher association between rivaroxaban therapy and venous thromboembolism as compared with apixaban, dabigatran and edoxaban. These findings are uncertain owing to the reliability of a post-marketing registration system that is negatively influenced by a high level of under-reporting. However, based on pharmacodynamics, we cannot exclude the possibility that there is a real effect that may be driven by non-adherence.