Results
Out of 146 patients recruited for the study, 134 patients (91.8%) were included (97 patients taking rivaroxaban, 37 patients taking apixaban). Patients who were lost to follow-up, had their DOAC changed, stopped or down-titrated were excluded (Fig. 1).
Baseline demographics of all 146 patients are reported in Table 1. Patient characteristics were similar for rivaroxaban and apixaban patients. Three patients out of 134 included for analysis experienced SSE events (2.2%), while 33 experienced bleeds (24.6%).
We compared the characteristics of patients with and without bleeding events for both DOACs. For the rivaroxaban cohort, CrCl and concomitant use of simvastatin were statistically significant factors associated with bleeding (p = 0.047 and 0.024 respectively) (Table 2). Further information on the frequencies of patients who bled in each individual class is described in Supplementary Table S1. We did not carry out further analyses for the apixaban cohort because there were no significant factors associated with bleeding in the cohort (Supplementary Table S2).
Table 3 shows the multivariable analysis for risk of bleeding with rivaroxaban. After controlling for BMI, CrCl and previous SSE, concomitant use of simvastatin (compared with patients not on statins) remained as the only significant predictor of bleeding [Adjusted OR = 6.14 (95% CI: 1.18 – 31.97), p = 0.031]. Although not statistically significant, patients with previous SSE events seemed to have a four-fold lower risk of bleeding [Adjusted OR = 0.25 (95% CI: 0.05 – 1.27), p = 0.094].
Comparing rivaroxaban plasma concentrations in patients with or without concomitant simvastatin use, median peak rivaroxaban plasma concentration in simvastatin patients was significantly higher compared to patients not on simvastatin (Table 4). Trough plasma concentrations were not significantly different between groups.
There was no statistically significant difference in median rivaroxaban plasma concentrations with or without concomitant use of atorvastatin (Supplementary Table S3). There were too few patients with concomitant use of rosuvastatin for comparison to be done.
Table 5 describes the trough concentrations and CHA2DS2-VASc scores of the three patients with SSE. These patients had low trough plasma concentrations of between 8.1% and 26.8% of the populations’ median concentrations for the respective DOACs. No further analyses were carried out to compare these patients as there were too few events for substantial comparison.