Warfarin, Low-Risk Clusters, and Resource-Limited Settings Authors: Ken Okumura Published: 2026 ABSTRACT Corresponding AuthorKey words: atrial fibrillation, DOAC, resource, stroke, warfarin Vitamin K antagonists (VKAs), predominantly warfarin, have long been the principal drugs for stroke prevention in patients with atrial fibrillation (AF). A paradigm shift occurred in the early 2010s with the development and introduction of the 4 direct oral anticoagulants (DOACs)—dabigatran, rivaroxaban, apixaban, and edoxaban—for nonvalvular AF. All these DOACs were shown to have at least noninferior efficacy to warfarin for the prevention of thromboembolic events, with an approximately 50% reduction in the risk of intracranial hemorrhage. A meta-analysis of the 4 phase III randomized controlled trials (RCTs) of DOACs demonstrated that, compared with warfarin, standard-dose DOACs were associated with a reduced risk of stroke or systemic embolism (HR: 0.81; 95% CI: 0.73 to 0.91; P < 0.0001), mainly by a reduction in hemorrhagic stroke (HR: 0.49; 95% CI: 0.38-0.64; P < 0.0001), and all-cause mortality (HR: 0.90; 95% CI: 0.85-0.95; P = 0.0003), with a trend toward reduced major bleeding (HR: 0.86; 95% CI: 0.73-1.00).1 Furthermore, all DOACs markedly reduced the risk of intracranial hemorrhage compared with warfarin (HR: 0.48; 95% CI: 0.39-0.59; P < 0.0001),1 which had been a major limitation of long-term warfarin therapy for stroke prevention in patients with AF. These findings have been consistently supported by numerous real-world studies evaluating the effectiveness and safety of the 4 DOACs compared with warfarin. Accordingly, current major guidelines—the 2023 American College of Cardiology/American Heart Association/American College of Clinical Pharmacy/Heart Rhythm Society Guideline,2 the 2024 European Society of Cardiology Guidelines,3 and the 2021 Focused Update of the 2017 Asia Pacific Heart Rhythm Society Guidelines4—uniformly recommend DOACs over warfarin as first-line anticoagulant therapy for patients with nonvalvular AF and with thromboembolic risk. In addition, switching from warfarin to a DOAC is recommended for eligible patients who fail to maintain an adequate time in therapeutic range (TTR) >70% while receiving warfarin.3 What, then, is the role of warfarin in patients with AF? It is important to note that warfarin remains the only oral anticoagulant option for AF patients with mechanical heart valves or moderate-to-severe mitral valve stenosis, as clearly stated in current guidelines.2, 3, 4 Also, in frail elderly patients (≥75 years) with nonvalvular AF who have been treated with well-controlled warfarin (TTR >70%), continuation of warfarin may be reasonable, as switching to DOACs has been shown to increase the incidence of major or clinically relevant nonmajor bleeding events.5 Shared decision-making is therefore necessary in such situations. Another important consideration is the cost of each anticoagulant: In Japan, originator DOACs at standard doses cost approximately $62 to 75 per month, whereas warfarin (3 mg/day) costs only about $5 per month (based on the U.S. Dollar/Japanese Yen exchange rate in May 2026). These differences in drug costs may represent an important consideration depending on the type of insurance and the economic circumstances of individuals, regions, and countries. In fact, warfarin is still widely used, particularly in resource-limited settings, because DOACs are less accessible owing to their higher cost and limited availability.6 Under such circumstances, should warfarin still be considered a first-line anticoagulant option for patients with nonvalvular AF? If so, are there any shared characteristics among patients for whom warfarin therapy may remain appropriate? Identifying patient subgroups in whom warfarin may provide a net clinical benefit comparable to that of DOACs is of considerable clinical importance. In this issue of the journal, Said et al7 provide important insights into the characteristics of patients in whom first-line and continued warfarin therapy may remain justified, particularly in resource-limited settings. The study aimed to explore the heterogeneity of treatment effect (HTE) in patients with nonvalvular AF using individual-patient data from phase III RCTs comparing DOACs and VKAs, using classical and modern machine learning based approaches for assessing multivariable HTE, and identify distinct patient subgroups through cluster analyses, assessing HTE within these subgroups and evaluating individual treatment benefits based on clinical variables. In addition, results were implemented into a simplified risk model and an interactive online tool to help health care providers select patients who would gain the most benefit from DOACs. The authors used individual participant data from COMBINE AF,8 a data set comprising 4 phase III RCTs comparing DOACs with warfarin in patients with nonvalvular AF: RE-LY for dabigatran, ROCKET AF for rivaroxaban, ARISTOTLE for apixaban, and ENGAGE AF–TIMI 48 for edoxaban. The outcome of the analysis was a net clinical outcome, defined as a composite of all-cause death, disabling or fatal stroke, and intracranial or fatal bleeding. Baseline characteristics included age, sex, race (Asian vs non-Asian), body mass index (BMI), creatinine clearance (CrCl), type of AF, prior ischemic stroke, prior myocardial infarction, diabetes mellitus, and prior VKA use. To provide a robust assessment of HTE and to identify specific population phenotypes for whom warfarin might work as well as DOACs, several approaches for multivariable HTE assessment were employed including unsupervised techniques and supervised techniques based on gradient boosting for effect-based stratification. There were 2 distinct patient subgroups identified through cluster analysis: a high-risk cluster characterized by older age, lower BMI, and reduced renal function, and a low-risk cluster characterized by younger age, higher BMI, and better renal function.7 In the high-risk cluster, DOACs significantly reduced net clinical outcome events compared to warfarin (HR: 0.86; 95% CI: 0.81-0.92; P < 0.001). In the low-risk cluster, the benefit of DOACs was smaller and nonsignificant compared with warfarin (HR: 0.93; 95% CI: 0.85-1.03; P = 0.15, P-interaction, 0.115). The primary predictors of HTE were renal function, age, and BMI, with no treatment-by-subgroup interactions identified. Based on these results, the authors emphasized the importance of individual patient characteristics—particularly renal function, age, and BMI—in determining the optimal anticoagulant therapy for patients with nonvalvular AF. Accordingly, in regions with limited access to DOACs, the use of warfarin, instead of DOACs, for patients considered a low-risk cluster group could be justified, as it may achieve net clinical outcomes comparable to those of DOACs. Furthermore, they developed an interactive risk prediction model for estimating individual patient outcomes based on covariates such as renal function, age, and BMI. This tool is available at TIMI Risk Calculators. This study was carefully conducted using accurate and sophisticated statistical analyses, and the conclusions appear to be well supported. As mentioned previously, contemporary guidelines uniformly recommend DOACs in preference to warfarin for patients with nonvalvular AF at risk of thromboembolism,2, 3, 4 irrespective of baseline characteristics. The results of the present study may provide a new perspective on the use of warfarin in specific patient subgroups while maintaining net clinical outcomes comparable to those achieved with DOACs. This is particularly important in resource-limited settings.6 The interactive risk prediction tool developed based on the results of this study appears to be useful for predicting the risk of net clinical outcomes when treated with DOACs and warfarin.7 Further validation studies are warranted to assess its accuracy in real-world settings. There remain, however, several concerns. The first is how the high- and low-risk clusters can be defined in actual clinical practice for an individual patient. What cutoff values for younger age, higher BMI, and better renal function should be used to identify the low-risk patient group? The study showed that the median age and CrCl (mL/min) of the high-risk cluster were 76.0 years (IQR: 71.0 to 80.0) and 55.9 mL/min (IQR: 46.0-65.0), respectively, whereas those of the low-risk cluster were 65.0 years (IQR: 59.0-70.0) and 93.3 mL/min (IQR: 82.3-112.0), respectively.7 Should patients aged ≤65 years or those with CrCl ≥93 mL/min be considered part of the low-risk group? The second concern is whether an interactive prediction tool is always necessary for risk stratification in an individual patient. Given that CrCl calculated using the Cockcroft-Gault equation incorporates both age and body weight, what are the respective roles of age and body weight in risk stratification? In addition, what is the role of absolute risk reduction in choosing between a DOAC and warfarin? These issues require further investigation. The final concern is that maintaining warfarin therapy is not always straightforward because it requires regular monitoring of prothrombin time-International Normalized Ratio to maintain a TTR >70%, is subject to interactions with foods and concomitant medications, has a slower onset and longer offset of action than DOACs, and carries a higher risk of intracranial hemorrhage than DOACs. In the era of DOACs, warfarin therapy in patients with nonvalvular AF is justified only when these concerns are appropriately addressed. In conclusion, this study provided evidence that identifying a lower-risk patient subgroup that may achieve net clinical outcomes similar to those with DOACs while receiving warfarin could help optimize resource utilization. On the other hand, higher-risk patient subgroups should be prioritized for treatment with DOACs. Although the definition of high- and low-risk groups requires further investigation, well-controlled warfarin may still have a role in patients with nonvalvular AF classified into the low-risk cluster, especially in resource-limited settings. Funding support and author disclosure Dr Okumura received remuneration from Daiichi Sankyo Co., Ltd., Bristol-Myers Squibb K.K., Medtronic Japan Co., Ltd., and Johnson & Johnson K.K. Footnotes The author attests they are in compliance with human studies committees and animal welfare regulations of the author’s institution and Food and Drug Administration guidelines, including patient consent where appropriate. For more information, visit the Author Center. References