Report of the Subcommittee of Control of Anticoagulation on the determination of the anticoagulant effects of rivaroxaban


Job Harenberg, Department of Clinical Pharmacology Mannheim, Medical Faculty Mannheim, University of Heidelberg, Maybachstr. 14, D-68169, Mannheim, Germany.
Tel.: +49 6213839623; fax: +49 6213839622.


To cite this article: Harenberg J, Marx S, Weiss C, Krämer R, Samama M, Schulman S, on behalf of the working party: methods to determine rivaroxaban of the Subcommittee on Control of Anticoagulation of the ISTH. Report of the Subcommittee of Control of Anticoagulation on the determination of the anticoagulant effects of rivaroxaban. J Thromb Haemost 2012; 10: 1433–6.


Rivaroxaban is an oral direct factor Xa inhibitor that has been demonstrated to be effective for the prophylaxis of postoperative venous thromboembolism in orthopedic surgery and atrial fibrillation and for the treatment of established venous thromboembolism [1–3]. Rivaroxaban is given at fixed daily doses without the need for adjustment according to laboratory results. However, the anticoagulant effect of rivaroxaban may need to be measured in some patient populations, such as in the elderly or children, in patients with renal impairment, during bleeding or thrombotic episodes, prior to surgery, and to assess compliance [4]. The concentration of rivaroxaban is determined by HPLC methods that are not suitable for clinical routine practice [5]. Rivaroxaban prolongs clotting times in many coagulation assays, and inhibits FXa in FXa-specific chromogenic substrate assays [6–14]. The Subcommittee on Control of Anticoagulation of the ISTH decided to analyze the anticoagulant effects of rivaroxaban in several coagulation tests using plasma samples spiked with the anticoagulant in an international collaborative study. This investigation was carried out with a panel of plasmas spiked with rivaroxaban to evaluate the performance of different coagulation tests.

Material and methods

Platelet-poor plasma (PPP) was prepared from blood collected in 3.8% citrate (blood/citrate, v/v, 9 : 1; kindly provided by Technochrome, Vienna, Austria) from 20 healthy volunteers. Rivaroxaban was extracted from commercially available Xarelto (Bayer Health Care, Wuppertal, Germany) by treating an aqueous suspension with dichloromethane at room temperature. The results of electrospray analyses, elemental analysis and NMR analysis (Table S1) were in agreement with published data [15]. A stock solution was prepared containing 100 μg mL−1 rivaroxaban in dimethylsulfoxide (DMSO). The final concentration of DMSO in the plasma samples ranged from 0.02% to 0.15%. Four PPP samples were prepared by spiking with rivaroxaban at disclosed concentrations of 0, 50, 150 and 500 ng mL−1. Three PPP samples (A, B, and C) were prepared by spiking with rivaroxaban concentrations of 100, 350 and 750 ng mL−1, which were unknown to the participants. All plasma samples were placed into sealed glass vials, freeze-dried, and sent to each investigator (kindly performed by Technoclone Immuno, Vienna, Austria). The sample containing no rivaroxaban served as a control as well as for dilution of samples if required.

Thirteen laboratories agreed to participate in the study, one laboratory did not provide results, and one laboratory returned two different sets of data. All participants were from clinical or institutional laboratories of academic institutions.

The organizers of the study decided to include the prothrombin time assay, the prothrombinase-induced clotting time (PiCT) assay and the FXa-sensitive chromogenic substrate assay in the study. The organizers preselected the rivaroxaban-sensitive reagents from the literature. The producers of the reagents were asked to participate in the study by providing the reagents. The activated partial thromboplastin time assay was not chosen for the study, because of its lower sensitivity [10]. The PiCT assay was included because of its sensitivity to rivaroxaban when performed with one incubation step [10,7]. One producer of a prothrombin time reagent, the producer of the PiCT assay and four producers of FXa-specific chromogenic substrate assays agreed to provide the reagents. The participants used their own in-house methods or the methods of the manufacturers to determine rivaroxaban, using several analyzers.

Coagulation assays

The prothrombin time assay was performed with the thromboplastin reagent STA Neoplastin Plus (Roche Diagnostics, Mannheim, Germany). The PiCT assay was performed with the one-stage clotting procedure (reagents from Pentapharm, Basel, Switzerland).

Chromogenic substrate assays

Four chromogenic substrate assays were included in the study. The assays were performed according to local methods or according to the description of the manufacturer. Coamatic Heparin (Instrumentation Laboratories, Haemochrom, Essen, Germany), STA Rotachrom-Heparin (Diagnostica Stago, Asnier sur Seine, France), STA Liquid anti-Xa (Diagnostica Stago), and Technochrome anti-Xa (Technoclone Immuno,Heidelberg, Germany and Vienna, Austria) were used. Table S2 summarizes the coagulation analyzers used by the participants in the study.

All reagents were sent together with the plasma samples to the participants of the study as quadruplicates. All assays were performed on all samples on four different days as duplicates. The results of the analyses of the samples with the known concentrations of rivaroxaban were plotted vs. time (coagulation assays) or optical density (chromogenic assays) to calculate the concentration of rivaroxaban in samples A, B, and C each day. The ratios of the coagulation data were calculated centrally for the samples with known concentrations of rivaroxaban.

Statistical analysis

The mean, median, standard deviation (SD), minimum and maximum values were calculated from the data. Differences in the coefficient of variation (CV) for each assay compared across the sites were analyzed statistically with the Maloney–Rastogi test [16]. The level of significance was set at < 0.01. This analysis accounts for the interlaboratory variation and the variation between methods.


The instruments used by the centers are summarized in Table S2. A descriptive analysis of results was not performed, because some of the assays were used in only one center.

Samples with known concentrations of rivaroxaban

The ratios of the prothrombin time assay showed a linear relationship between the known concentrations of rivaroxaban and the mean values of the coagulation times (Fig. S1). The ratios of the PiCT assay were prolonged more and with a higher SD than those of the prothrombin time assay (Fig. S2). The results of the chromogenic assays could not be analyzed, because laboratories used continuous or endpoint assessment of the release of the chromogen from the substrate in the assays.

Samples with unknown concentrations of rivaroxaban

The mean, median, SD, CV, minimum and maximum values of samples A, B and C are summarized in Table 1, using the sum of all values and all methods provided by participants.

Table 1.  Results of the coded samples A, B, and C, using the sum of all results of the study for samples A, B, and C, containing the added (true) concentrations of 100, 350 and 750 ng rivaroxaban mL–1 plasma
Sampleng mL−1nxMean (ng mL−1)Median (ng mL−1)SD (ng mL−1)CV (%)Minimum (ng mL−1)Maximum (ng mL−1)
  1. CV, coefficient of variation; SD, standard deviation.


The results of the determination of the concentration of rivaroxaban in samples A, B and C are given in Tables S3, S4, and S5, respectively. The P-values are included in these tables for the analysis of differences of the error variance between two assays.

STA Neoplastin Plus had the lowest SD for samples A (100 ng mL−1; Table S3) and B (350 ng mL−1; Table S4), indicating that this assay had the highest precision. Eight of 10 test results (Tables S3 and S4) were highly significant (P < 0.0001).

The chromogenic assays showed significant differences from each other, but without a consistent pattern for samples A and B. SDs for sample C (750 ng mL−1 rivaroxaban; Table S5) were high with all methods, without differences between the assays.

Discussion and conclusion

The overall results of the interlaboratory comparison show acceptable CVs between 7% and 25% for some assays at 100 ng mL–1 and 350 ng mL–1. These concentrations are expected at trough levels and peak levels after administration of 10 or 20 mg rivaroxaban orally, respectively [17]. High concentrations of rivaroxaban (750 ng mL−1) resulted in high SD and CV values, the latter ranging from 37% to 44%. The high concentration of rivaroxaban was included in the study to obtain information on the deviation of the results from this value for the detection of an overdose of rivaroxaban. Special attention has to be given to this finding. Our findings suggest that very high levels of rivaroxaban may not be measured accurately.

The Maloney–Rastogi test uses the empirical correlation coefficient of the sum and the difference of two measurements, X1 and X2. It can be shown mathematically that the Pearson correlation coefficient of X1 + X2 and X1 – X2 is zero if, and only if, there is no difference between the two error variances [18]. The F-test, which is widely used in order to compare the empirical variances of X1 and X2, was not appropriate in this study, because we were dealing with paired variances. It is noteworthy that the power of the Maloney–Rastogi test decreases with increasing population variance. Studies with the primary goal of precision comparison, such as the present one, may therefore require a largely homogeneous study population [18]. Another option to improve the power of the Maloney–Rastogi test is to perform replicated measurements on each sample under investigation. This latter option was chosen for the present study.

This study has limitations. Assays were chosen according to the available information at the time of planning of the study in the middle of 2010. The study did not use samples from patients treated with rivaroxaban. However, spiking of plasma samples in studies such as this is a well-established methodology, and has been used with known and unknown concentrations of heparin, low molecular weight heparin, hirudin or argatroban in other studies of the SSC on Anticoagulation [19–21]. This contrasts with the standardization procedure for vitamin K antagonists, where the compound does not directly induce any anticoagulant effect.

In summary, of the tested methodologies, STA Neoplastin Plus is the most precise method for determination of the anticoagulant rivaroxaban in human plasma. Other thromboplastin reagents remain to be investigated. Some chromogenic assays precisely determined rivaroxaban in plasma. Efforts are needed to reduce the error variances between assays. Investigations have yet to be performed on the precision of additional assays, the use of ex vivo samples containing rivaroxaban, and other new oral coagulation inhibitors.

Disclosure of Conflict of Interests

The authors state that they have no conflict of interest.