In this real-world, retrospective, observational study, we evaluated the effectiveness and safety of apixaban versus warfarin in very elderly patients with AF in Japan. The primary results indicated that apixaban was associated with a significantly lower risk of stroke/SE and major bleeding compared with warfarin in very elderly patients with AF. Additionally, the secondary analyses demonstrated that the risk of ischemic stroke, embolic stroke, cerebral infarction or SE, intracranial bleeding, gastrointestinal bleeding, or intraocular bleeding were significantly lower with apixaban than with warfarin. The real-world findings of this study in elderly patients align with and complement the results of the ARISTOTLE trial [4], a large-scale RCT that compared the safety and effectiveness of apixaban versus warfarin. While this study has certain limitations, its significance lies in the ability to extrapolate the ARISTOTLE results to a real-world patient population that is typically predominantly elderly, alongside other real-world evidence obtained from various databases across different countries [13, 18,19,20].

In this study, patients aged ≥ 80 years were selected for the analysis, with the mean age of the s-IPTW cohorts being approximately 85 years. This contrasts with the ARISTOTLE trial, an RCT in which the mean age of the participants was 70 years [4], as well as another database study in Japanese patients in which the mean age was 76 years. Evidence in the elderly population is limited because RCTs, such as the ARISTOTLE trial, often include younger populations. However, elderly patients are at high risk for both stroke and bleeding, necessitating careful treatment decisions. A key finding of this study is that apixaban significantly reduced the risk of stroke/SE and bleeding compared with warfarin, even in this older population. These results are also consistent with those of the ANAFIE study [21], a registry study of elderly patients with AF in Japan (mean age, 81–82 years). Although ANAFIE was a voluntary registry with a somewhat ambiguous follow-up process—yielding relatively lower efficacy and adverse event rates compared with those observed in the current study—our findings confirm these results on a broader basis, with a more advanced age cohort and secure outcome recording under Japan’s universal healthcare system.

Due to the limitations of the database used in this study, laboratory data such as the international normalized ratio (INR) values were available for only a small subset of patients, making it difficult to directly assess anticoagulation control in the warfarin group. However, patients with AF are typically managed by specialists, such as cardiologists or neurologists. In fact, the ANAFIE registry, which included patients with NVAF primarily managed by cardiologists, reported a mean INR of 1.97 at enrollment [22]. Similarly, INR levels in international clinical studies have generally been maintained within or close to guideline-recommended ranges [23, 24]. Since patients in the current study were also treated at acute care hospitals by specialists like those registered in the ANAFIE registry [22], the INR levels were similar to those in the Japanese registry and were likely managed within or near the recommended range. Nonetheless, given the advanced age of the study population, INR control may have skewed slightly lower, toward the lower end of the therapeutic range, due to concerns about bleeding risk. While the quality of anticoagulation control in the warfarin group could have influenced the outcomes, this study reflects routine clinical practice and thus provides meaningful real-world evidence.

In general, the physical condition of elderly patients differs substantially, even among those of the same age, and for this reason, determination of a patient’s condition should not be based on age alone. In this study, we conducted subgroup analyses to examine whether the treatment effect was influenced by the patients’ condition using the CCI and ADL indices. No significant interaction between the treatment (apixaban or warfarin) and CCI or ADL was observed in any of the analyses (all P values for the interaction were greater than 0.05), suggesting that CCI or ADL had no impact on treatment efficacy. These results suggest that even in patients with high CCI values or low Barthel Index scores, the balance between the risks of stroke and bleeding should be carefully evaluated. Apixaban appears to offer a greater benefit compared with warfarin in patients requiring anticoagulation, and our data provide valuable information to support shared decision-making.

A considerable number of elderly patients are in a state of so-called polypharmacy, wherein a large number of drugs are taken owing to a high number of complications [25, 26]. Treatment guidelines recommend that patients with AF with polypharmacy minimize the number of medications while giving due consideration to the need for cardiovascular disease prevention with the goal of reducing bleeding risk. This includes, but is not limited to, avoiding the use of drugs that may increase the risk of bleeding, such as antiplatelet agents and NSAIDs, as much as possible [27]. While polypharmacy may affect the usefulness of drugs through drug interactions and poor medication adherence, the number of drugs taken did not significantly affect the treatment effect in our analysis. However, this analysis focused solely on the number of medications taken and did not examine potential drug interactions or the use of medications that could elevate bleeding risk, which warrants further investigation.

Several studies have evaluated the safety and effectiveness of DOACs in elderly patients, and the findings of the present study are consistent with these reports, demonstrating that DOACs are associated with better safety and effectiveness profiles than warfarin even in older adults [28,29,30]. Although comparative studies with different DOACs have been conducted, definitive conclusions about the optimal treatment choice remain elusive, particularly in elderly patients with additional complexities, including comorbidities, polypharmacy, or impaired renal or hepatic function. Tailored treatment decisions remain essential, and further research is warranted to clarify the comparative effectiveness of different DOACs in this population.

One of the strengths of this study is that it was conducted in Japan, which has an aging population ahead of other developed countries, and thus was able to collect data for analysis under real-world clinical conditions of a larger number of very elderly patients ≥ 80 years of age with AF who were taking apixaban or warfarin in comparison to previous studies. We believe that these data will be useful in other countries where the number of elderly patients is expected to increase in the future as well.

As with many other database studies, the inherent limitation due to analysis using insurance claims data exists in this study as well. First, the current study used information provided by hospitals applying the DPC, flat-fee payment system, which are mostly large hospitals responsible for acute care. Therefore, a significant number of the patients included were likely in poorer health than the average population, possibly with more comorbidities, limiting the generalizability of the study. Second, due to the nature of the claims data, laboratory test results were not available for all patients (only 15–20% of patients had results available) because of this we were unable to consider these variables in the calculation of the propensity score, and, consequently, there is no certainty that patient characteristics were fully balanced after s-IPTW. Third, this study used hospitalization due to cardiovascular events or bleeding as the primary outcome, rather than including all events regardless of hospitalization. This definition was necessary to ensure the identification of new and clinically relevant events, as the claims database does not provide detailed or reliable information on nonhospitalized events. Consequently, the incidence of both cardiovascular events and bleeding may have been underestimated, and the generalizability of our findings to broader clinical settings may be limited. Fourth, no information regarding follow-up loss was available as the database does not include information on patients who visit a different hospital or clinic. This could have led to an underestimation of the incidence of stroke/SE or major bleeding events. Fifth, as INR and time-in-therapeutic-range (TTR) data were not available for most patients in the warfarin group, the quality of anticoagulation control could not be assessed. This lack of information substantially limits the validity of the comparison with warfarin treatment and is therefore one of the major limitations of the study. Sixth, although apixaban dosage information was available, most (approximately 90%) patients received a reduced dose, and a dose-stratified analysis was not conducted. Furthermore, the database does not contain detailed clinical information regarding the reasons for dose reduction on a per-patient basis. This limitation precluded further evaluation of whether outcomes may have differed according to dosing appropriateness or physician judgment. Finally, primary endpoints were defined as events leading to hospitalization, which is different from the adjudicated endpoints commonly used in RCTs, and did not include mortality.