Discussion
The present study is, to the best of our knowledge, the first study that highlights a six-fold increased risk of bleeding when rivaroxaban and simvastatin were used together. We propose that there could be two possible explanations for the observation of increased bleeding with concomitant simvastatin.
Firstly, approximately 18% and 14% of rivaroxaban is metabolized by CYP3A4 and CYP2J2 respectively [21], while P-gp was reported to contribute to its renal elimination [21-23]. In addition, the human organic anion transporter 3 (OAT3) was demonstrated to play a pivotal role in the renal clearance of rivaroxaban via its basolateral uptake in proximal tubular cells [24]. A previous study predicted that systemic exposure of rivaroxaban and bleeding risk could be increased due to inhibition of OAT3 by benzofuran antiarrhythmic agents [25].
Several studies had been carried out to understand the inhibitory potential of 3-hydroxy-3-methylglutaryl coenzyme A (HMG-CoA) inhibitors (statins) on liver enzymes and drug transporters. Yang et al demonstrated that simvastatin exhibited CYP3A4 inhibition with a half-maximal inhibitory concentration (IC50) value of 3.10 µM [26], while another study demonstrated a greater than 85% reduction of CYP2J2 activity by simvastatin at 30 µM [27]. Simvastatin is also known to inhibit P-gp with IC50values between 8.9 to 49 µM [28, 29] and shown to have some inhibitory effects on OAT3 with IC50 values between 32.3 and 48.1 µM [30-32]. Thus, we postulate that simvastatin could potentially decrease the hepatic and renal clearances of rivaroxaban, potentially leading to more bleeding. This is further supported by significantly higher median peak plasma concentration in patients co-prescribed with simvastatin in this study.
Secondly, lower low-density lipoprotein cholesterol (LDL-C) levels may result in higher bleeding risk [33]. While LDL-C was not measured in our study, the use of simvastatin has been reported to decrease the LDL-C levels by 28.3 to 45.8% [34]. It was posited that low LDL-C levels could be negatively associated with platelet activation [33], and that platelet aggregation might be impaired due to depletion of cholesterol [35]. Taken together, decreased platelet activation and aggregation might have contributed to the observed increased risk of bleeding in patients co-prescribed with simvastatin. Further studies involving LDL-C measurements would need to be carried out to substantiate this postulation.
While the event rate for rosuvastatin was too small for a convincing analysis, a question arises as to why a similar significance was not observed for concomitant use of atorvastatin which also decreases LDL-C levels in patients [34] and was similarly shown to inhibit CYP3A4 [26], P-gp [28] and OAT3 [30, 31] in various studies? One possible explanation is the weaker inhibitory potencies of atorvastatin against CYP3A4-mediated metabolism and P-gp-mediated transport with IC50 values of 48.0 µM [26] and between 271 and 356 µM [28], respectievly. Considering the relatively more potent inhibition of CYP3A4 and P-gp by simvastatin, the interaction with rivaroxaban was more pronounced culminating in higher peak rivaroxaban plasma concentrations observed in this study.
Another finding was that median CrCl was significantly lower in patients taking rivaroxaban who experienced bleeds as compared to those who did not. This is supported by two possible explanations.
Firstly, an increased risk of bleeding independent of anticoagulant use in patients with renal impairment was identified by Del-Carpio et al in a meta-analysis [36]. As CrCl can estimate glomerular filtration rate which is used as a measurement of renal function [37], it may be inferred that patients with lower CrCl have poorer renal function which potentially contributes to bleeding risk.
Secondly, poorer renal function may also have decreased the elimination of some concomitant medications that may have interactions with rivaroxaban, for example, simvastatin which is partially cleared by the renal route [38]. This would translate to a potentially larger extent of inhibition of liver enzymes or drug transporters that are crucial in the elimination of rivaroxaban and may thus contribute to bleeding risk. This may also explain the likely collinearity between use of simvastatin and CrCl, as CrCl was a significant predictor for bleeding in the univariable analysis, but was no longer a significant predictor when modeled with other factors including use of simvastatin, while simvastatin was identified to be the only significant predictor for bleeding after controlling for BMI, CrCl and previous SSE.
Interestingly, patients with a previous SSE were associated with a four-fold lower risk of bleeding, albeit not statistically significant. We postulate that patients who had previously experienced SSE events have a higher risk of clotting at baseline based on the pathophysiology of embolism, and hence have lower bleeding risk.
While statistical tests were not performed to characterize the relationship between trough plasma concentrations and SSE events due to the small sample size of patients who experienced a SSE event during follow-up, we observed that these patients belonged to the lowest class for trough plasma concentrations, with one patient whose trough plasma concentration fell below the LLOQ of the HPLC-MS assessment. This observation is supported by a previous study by Testa et al [13]. DOACs are reversible, competitive inhibitors of factor Xa with short half-lives [39, 40]. Thromboembolism can occur in a period where plasma concentrations are below the necessary thresholds for sufficient inhibitory activity to maintain adequate anticoagulation. This should be verified in larger studies.
We acknowledge that this study has its limitations. Firstly, the ability to detect significant relationship between DOAC plasma concentrations and SSE events is limited by the small sample size. This is especially apparent for the apixaban group which only consisted of 37 patients included for analysis, and only 1 patient experienced an SSE. To counter issues with small sample size, we included patients whose plasma concentrations were below the LLOQ by estimating their plasma concentrations to LLOQ/2 to provide more data points for the analysis. As only a small percentage of recorded plasma concentrations was below the LLOQ, this method of estimation is unlikely to be biased [41]. Future studies could be designed to address the issue of extremely low trough plasma concentrations and how that should be best handled. This would be important in associative studies correlating trough concentrations and risk of SSEs when investigating the acceptable lower limits for trough concentrations. This would have important clinical interpretation and use.
Secondly, we assumed that the recorded plasma concentrations of the DOACs stay constant throughout the 1-year follow-up. The most likely reason for fluctuating plasma concentrations could be attributed to poor adherence to DOACs because a one year timeframe is too short for significant changes in the disposition of DOACs. We made the best possible effort to ascertain medication adherence during the follow-up calls, and were able to ascertain that the adherence rate was over an impressive 90%. Thus we do not think that poor adherence could have contributed to the observed outcome of low plasma concentrations and thus SSEs.