In this analysis of nearly 42,000 NVAF patients (identified from January, 2013 to September, 2016, including both patients with commercial insurance and Medicare coverage) who newly initiated DOACs, 37% initiated apixaban, 51% initiated rivaroxaban, and 13% initiated dabigatran. Our findings were that, after we adjusted for differences in patient characteristics, NVAF patients who received apixaban were significantly less likely to switch or discontinue treatment compared to patients who received either rivaroxaban or dabigatran. In this study, approximately 21% of patients who initiated DOACs switched to another OAC (apixaban 3.6%; rivaroxaban 6.3%; dabigatran 11.1%). The overall DOAC switching rate was similar to that reported by Manzoor et al. who found that, among 34,022 OAC-naïve NVAF patients identified from the MarketScan Commercial and Medicare Supplemental Database during 2009–2013, 19.4% of patients who initiated DOACs switched to another OAC . In the study of Manzoor et al., the NVAF patients who switched from their index DOAC switched to warfarin or another DOAC fairly equally . Among our study population, slightly more NVAF patients who switched from apixaban switched to another DOAC than to warfarin (51.4% vs. 48.7%); of those who switched from rivaroxaban (56.3% vs. 43.8%) and dabigatran (71.1% vs. 28.8%), more patients switched to another DOAC than to warfarin. Because of the limitations of the claims database used for this study, we were unable to establish reasons for switching OAC treatment and further research is warranted on this topic.
In an updated analysis by Brown et al. , the switching rate observed among 15,341 NVAF patients identified between 2013 and 2014 from MarketScan data was higher than observed in our study and also than that observed in the study by Manzoor et al. . In this particular study, of the patients treated with dabigatran, apixaban, and rivaroxaban, 18.8%, 10.5%, and 9.5%, respectively, switched within 9 months of starting DOAC treatment . However, compared to that observed in our study, the DOAC discontinuation rate (gap of at least 30 days) was lower, ranging between 11.9% and 15.9% after 9 months . The study cohorts of Brown et al. were older with a mean age that ranged between 68 and 74 years old and they were at higher stroke risk, which may indicate that older NVAF patients with higher stroke risk are more likely to switch treatment than discontinue DOAC treatment . As prescribing trends change for OACs, it will be important to continue to monitor such switching and discontinuation trends as this information may be useful for healthcare decision-makers and payers for optimizing OAC treatment choices for particular patient groups.
In the current study, the unadjusted discontinuation rates of NVAF patients who initiated apixaban, rivaroxaban, and dabigatran were 52.8%, 60.3%, and 62.9%, respectively, and were generally consistent with the findings of Lip et al., who reported cumulative discontinuation rates after 1 year of 50.5% for apixaban, 57.8% for rivaroxaban, and 64.7% for dabigatran among NVAF patients (n = 29,900 identified from MarketScan data; mean ages 66.5–68.5 years) who newly initiated DOAC treatment . In both the Lip et al. study and our study, discontinuation was defined as having a gap of more than 30 days in prescriptions and may be indicative of patients with an interruption in therapy and not permanent drug discontinuation. Our regression adjusted hazard rates for discontinuation of rivaroxaban vs. apixaban (HR 1.1, p < 0.001) and dabigatran vs. apixaban (1.3, p < 0.001) were also directionally consistent with those of Lip et al., who reported 1.2-fold and 1.5-fold increased risk for discontinuation for rivaroxaban and dabigatran users, respectively, compared to apixaban users .
McHorney et al. conducted a study of NVAF patients identified from the IMS Health claims database during July 2012 to June 2015 . They reported persistence (i.e., no gap of more than 30 days between end of days of supply and the next fill start date) rates at 9 months after NVAF patients initiated DOACs of 68.1% for apixaban, 69.3% for rivaroxaban, and 57.8% for dabigatran . These findings may be presumed to indicate lower discontinuation rates of DOACs among their study population than observed in our study population . After adjusting for patient differences, the findings of McHorney et al. were that NVAF patients who initiated rivaroxaban vs. those who were treated with apixaban and dabigatran were more likely to be persistent and adherent [proportion of days covered (PDC) at least 0.80] to treatment . Noted differences in the study population of McHorney et al. vs. our study population include that NVAF patients were older and they had greater comorbidity with higher stroke risk . Furthermore, unlike in our study in which patients were excluded if they had prior usage of OACs in the 12-month baseline period, between 20% and 35% of the cohorts in the study by McHorney et al. had prior usage of OACs other than their index OAC . A second study by Manzoor et al. looked at persistence and adherence to DOACs when NVAF patients were stratified by whether they were naïve or experienced with OAC usage . In this study of 66,090 patients identified from MarketScan data, those who were naïve to OAC treatment compared to those who were experienced had lower persistence and suboptimal adherence to treatment . This could be related to potentially longer duration and greater severity of NVAF in OAC-experienced patients than OAC-naïve patients. Thus, since our patient population was entirely OAC naïve (at least in the 12-month baseline period), they might be expected to have higher discontinuation rates than observed by McHorney et al. and the findings between the DOACs may differ as well .
A large retrospective claims database analysis by Yao et al. of 64,661 patients with AF (identified from Optum Labs Data Warehouse between November 2010 and the end of 2014) who initiated apixaban (16.4% of study population), rivaroxaban (39.2%), dabigatran (10.3%), or warfarin (34.1%) reported that approximately 47.5% of the patients treated with DOACs were adherent (PDC at least 80%) to treatment, compared with 40.2% of warfarin-treated patients . The proportion of patients adherent to DOAC treatment was similar to that observed by Brown et al., which was also less than half regardless of the DOAC that was initiated . In the study by Yao et al., patients treated with apixaban had the highest unadjusted rate of adherence at 61.9%, followed by patients treated with rivaroxaban (50.5%) and dabigatran (38.5%) . As observed in prior studies [21,22,23], Yao et al. found that patients with a high stroke risk who were nonadherent to DOAC treatment were at significantly increased risk for stroke compared to adherent patients . Moreover, a recent analysis of participants in the ENGAGE AF-TIMI 48 trial found that an interruption of more than 3 days in anticoagulation therapy (edoxaban or warfarin) was associated with a considerable increase in the risk for major cardiac and cerebrovascular events in the 30 days following therapy interruption . Alongside the high discontinuation rates observed in our study and that observed by Lip et al., the findings of Yao et al.  highlight that adherence to anticoagulation treatment is generally poor in routine clinical practice in the USA. Additionally, under the relatively currently used OAC management practices, adherence is only found to be modestly improved with DOACs vs. warfarin treatment . Thus, persistence and continuous adherence to OAC treatment remain a challenge, suggesting more routine outpatient follow-up of NVAF patients is needed to better inform NVAF patients of the importance of adherence to treatment for maintaining stroke prevention and also to address barriers to OAC continuation. The importance of medication adherence is further emphasized in that it is being used as a quality metric for providers and health plans, as well as for making formulary decisions and determining prescribing practices [17, 25, 26].