The main finding of the current study is that the prevalence of DOAC increased gradually over time, and use of DOAC was not associated with stroke/SE or major bleeding events, compared with warfarin, on Cox proportional hazards modeling incorporating change in OAC status as a time-dependent covariate.
In the real world, patients had gone through several transitions of OAC status, with the net use of DOAC growing over time. In the era of warfarin, OAC was overused for low-risk patients, and also underused for high-risk patients.9
After the release of DOAC, the total prevalence of OAC increased gradually, but use of OAC increased predominantly in low-risk patients. This suggests that the issues in the era of warfarin have not been overcome and there is still discordance between the guidelines and real-world clinical practice.
The efficacy and safety of DOAC (dabigatran, rivaroxaban, apixaban, and edoxaban) compared with warfarin were shown in such RCT as the RE-LY, ROCKET AF, ARISTOTLE, and ENGAGE AF–TIMI 48 trials, respectively.2–5
Recently, meta-analysis of these RCT reported that DOAC significantly reduced stroke/SE events by 19%, compared with warfarin, mainly driven by a reduction in hemorrhagic stroke, and DOAC were similar to warfarin with regard to major bleeding events.19
Observational studies in general practice settings, however, warned that some DOAC may have a higher risk of gastrointestinal bleeding than warfarin, especially in older patients.20,21
In a recent report from the Danish nationwide database, there was no significant difference between DOAC and warfarin in terms of ischemic stroke, and the risk of major bleeding was significantly lower for apixaban and dabigatran compared with warfarin, therefore, DOAC seem to be safe and effective alternatives to warfarin in a routine care setting.22
Similarly, in the current study, the use of DOAC was not statistically significant different to warfarin with regard to the incidence of stroke/SE, and there was a tendency toward fewer major bleeding events, although this was not statistically significant.
There are several possible reasons for why the present results do not exactly match the results of RCT. First, frequent use of under-dose DOAC may have had some influence on unfavorable clinical outcome. With regard to US cardiologist prescription data, approximately 20% of DOAC prescriptions were for a reduced dose.23
Use of under-dose DOAC is common in real-world clinical practice, and although it may have a more favorable bleeding profile, it may also result in insufficient stroke prevention.19
In the current study, more than half of the DOAC prescriptions were for reduced dose in 2015. Off-label under-dosing of DOAC in the USA is also associated with an increased risk for cardiovascular hospitalization.24
Second, OAC adherence may have some influence on unfavorable clinical outcomes. In primary care practices in Germany, only approximately 50–60% of DOAC users had good adherence.25
In contrast, a US Veterans Affairs cohort had good adherence to dabigatran (72%), but poor adherence to dabigatran was associated with increased adverse outcome.26
In patients with CHA2DS2-VASc score ≥2, poorer adherence was associated with higher stroke risk and a relatively small decrease in bleeding risk.27
OAC adherence is often suboptimal in real-world clinical practice, and may be associated with poor clinical outcome. The half-lives of DOAC are shorter than that of warfarin, and the anticoagulation effects of DOAC decline more rapidly when scheduled doses are not taken. Therefore, the impact of poor adherence on adverse effects may be greater for DOAC than warfarin.
This study has several limitations. First, this was an observational study with the limitations inherent to observational study design such as selection bias and unmeasured confounders. Therefore, the Cox regression modeling may still be biased and should be interpreted with caution. Second, OAC status was assessed annually in the current study, and we made an assumption regarding change in OAC status. To compensate for the weakness of the statistical analysis, we conducted an additional sensitivity analysis and confirmed that the results did not change. Third, we investigated neither time in therapeutic range in the warfarin group during follow-up nor OAC adherence, and therefore the effects of quality of warfarin control and of OAC adherence on outcome are unknown. Fourth, the number of patients on DOAC was relatively small, and we could not analyze the impact of DOAC on outcome due to insufficient statistical power. Fifth, although frequent use of under-dose DOAC might have some influence on clinical outcome, we could not conduct further analysis due to the small number of DOAC users and follow-up period. Sixth, data on medical therapy prior to events were not readily available, therefore the impact of this on outcome is not known. And seventh, although blood pressure control is an important factor, we did not investigate this.