Essentials Accurate determination of anticoagulant plasma concentration is important in clinical practice. We studied the accuracy and consistency of anti-Xa assays for rivaroxaban in a multicentre study. In a range between 50 and 200 μg L ^-1 , anti-Xa activity correlated well with plasma concentrations. The clinical value might be limited by overestimation and intra- and inter-individual variation.
Background Determining the plasma level of direct oral anticoagulants reliably is important in the work-up of complex clinical situations. Objectives To study the accuracy and consistency of anti-Xa assays for rivaroxaban plasma concentration in a prospective, multicenter evaluation study employing different reagents and analytical platforms. Methods Rivaroxaban 20 mg was administered once daily to 20 healthy volunteers and blood samples were taken at peak and trough levels (clinicaltrials.gov NCT01710267). Anti-Xa activity was determined in 10 major laboratories using different reagents and analyzers; corresponding rivaroxaban plasma concentrations were measured by high-performance liquid chromatography-tandem mass spectrometry (HPLC-MS). Findings Overall Pearson's correlation coefficient of anti-Xa levels and HPLC-MS results was 0.99 for Biophen ^® Heparin (95% CI, 0.99, 0.99), Biophen ^® DiXaI (95% CI, 0.99, 0.99) and STA ^® anti-Xa liquid (95% CI, 0.99, 1.00). Correlation was lower in rivaroxaban concentrations below 50 μg L ^-1 and above 200 μg L ^-1 . The overall bias of the Bland-Altman difference plot was 14.7 μg L ^-1 for Biophen Heparin, 17.9 μg L ^-1 for Biophen DiXal and 19.0 μg L ^-1 for STA anti-Xa liquid. Agreement between laboratories was high at peak level but limited at trough level. Conclusions Anti-Xa activity correlated well with rivaroxaban plasma concentrations, especially in a range between 50 and 200 μg L ^-1 . However, anti-Xa assays systematically overestimated rivaroxaban concentration as compared with HPLC-MS, particularly at higher concentrations. This overestimation, coupled with an apparent interindividual variation, might affect the interpretation of results in some situations.