Investigating interassay variability between direct oral anticoagulant calibrated anti–factor Xa assays: a substudy of the perioperative anticoagulation use for surgery evaluation (PAUSE) trial

Background: Direct oral anticoagulant calibrated anti-factor Xa (FXa) assays can assess residual anticoagulant levels in patients requiring urgent procedures or surgery. However, previous studies have shown variability between anti-FXa levels determined by different instrument-reagent combinations. This may be related to use of lyophilized samples, direct oral anticoagulant-spiked plasma, or interlaboratory variation. Objectives: 1) Determine the interassay variability in anti-FXa levels using 3 common instrument-reagent combinations. 2) Determine if differences between these combinations are clinically relevant. Methods: Seventy apixaban and 59 rivaroxaban samples from participants in the Perioperative Anticoagulation Use for Surgery Evaluation trial were simultaneously tested using Biophen reagents on the BCS XP analyzer (Siemens), HemosIL reagents on the ACL TOP analyzer (Werfen), and Stago reagents on the STA CompactMAX analyzer (Diagnostica Stago). Interassay correlations were analyzed at the predetermined cutoff of 30 ng/mL and compared with median anti-FXa levels. Results: Anti-FXa levels showed moderate-to-very strong correlations for apixaban (r = 0.7271-0.9467) and rivaroxaban (r = 0.6531-0.9702). Anti-FXa levels were also significantly different between all instrument-reagent combinations in the < 30 ng/mL group. In the ≥ 30 ng/mL group, apixaban was significantly different in all combinations, while rivaroxaban only differed between Biophen/BCS XP and Stago/STA CompactMAX. 7.8% (10/129) of samples were discrepantly classified across the 30 ng/mL threshold. Conclusions: Anti-FXa levels determined by 3 common instrument-reagent combinations show moderate-to-very strong correlations with each other. Although there are statistically significant differences between median anti-FXa levels these differences are not clinically significant, and result in discrepant classification across the 30 ng/mL threshold in only 7.8% of samples.