Low incidence of thromboembolism with Fiix-monitored warfarin compared to conventional warfarin and DOACs in patients with AF

Pivotal randomized clinical trials (RCTs), reported that in patients with nonvalvular atrial fibrillation (AF), direct oral anticoagulants (DOACs) had at least a similar efficacy to traditionally prothrombin time–based international normalized ratio (PT-INR)-monitored warfarin (hereafter PT-warfarin) in reducing total stroke and systemic embolism.¹ In these trials, superior efficacy of DOAC drugs was reported at sites at which PT-warfarin control groups had time within target range (TTR) of <70%.^2,3 The trials also consistently reported lower incidence of intracranial hemorrhage with DOACs.^1,3-6 Based on these RCTs and convenience of administration, DOACs are now recommended as first-line anticoagulants over PT-warfarin in most patients with AF.⁷ 

While the DOAC trials were being conducted, our group developed a new method for warfarin monitoring, called the coagulation factor II (FII) and FX test (Fiix test; pronounced “fix test”; also known as Fiix-PT). A Fiix normalized ratio (Fiix-NR) is calculated in analogous fashion to PT-INR. By design, Fiix-NR is affected only by reductions in FII and FX,^8,9 the 2 factors that primarily bring about the antithrombotic effect of warfarin.^8,10,11 By ignoring FVII reductions, less-variable NR and dosing are achieved.^12-14 A modified warfarin dose initiation protocol is used to prevent initial overdose but maintenance management is based on PT-INR protocols.¹⁵ We conducted an investigator-initiated, blinded mixed-population RCT between 2012 and 2014 that demonstrated a 48% reduction in total thromboembolism (TE) without increased bleeding with Fiix-NR monitored warfarin (Fiix-warfarin) compared with conventional PT-warfarin; these findings were statistically noninferior.¹³ After publication of the results, PT-INR monitoring of warfarin was replaced with Fiix monitoring at our institution in July 2016. We then confirmed in a before-and-after interrupted time series study that the benefit of Fiix-warfarin over PT-warfarin extended into clinical practice.¹⁴ 

In light of the prior results, we hypothesized that Fiix-warfarin would have more favorable clinical outcomes in real-world practice than PT-warfarin, apixaban, dabigatran, and rivaroxaban.

This observational study compared rates of total TE, all-cause death, and major bleeding in all patients with AF who underwent anticoagulation long term (>180 days) with apixaban, dabigatran, rivaroxaban, PT-warfarin, and Fiix-warfarin in the Greater Reykjavik, Iceland, area during a 5-year period. Propensity score weighting (inverse probability of treatment weighting [IPTW]) was used to adjust treatment groups for possible drug prescription bias.¹⁶ Only patients residing in the Greater Reykjavik area (defined by postal codes) were included because they are all referred for medical emergencies to a single acute-care university hospital system that uses the same electronic hospital chart. Therefore, TE, major bleeding, as well as death rates could be accurately captured. Beginning in the middle of the study period, on 1 July 2016, Fiix-NR monitoring of warfarin was adopted instead of PT-INR monitoring. The study was approved by the National Bioethics Committee of Iceland (VSN-16-057-V4) and was conducted according to the Declaration of Helsinki.

The study used data from the nationwide Icelandic OAC database.^16,17 In short, all patients residing in the Greater Reykjavik area that were prescribed an OAC agent from 1 March 2014 to 28 February 2019 were identified using anatomic therapeutic chemical (ATC) codes from a national registry containing all outpatient drug prescriptions in Iceland.¹⁷ The unique personal identification numbers of these patients were linked to the electronic medical records of the Landspitali National University Hospital of Iceland in Reykjavik, Iceland’s 4 smaller regional hospitals serving rural communities outside the capital area, and primary care databases.

Only patients with AF treated long term, defined as having received OACs for at least 180 consecutive days, were included because fewer and fewer new warfarin referrals were received during the study period. If patients had been continuously on any OAC for >180 days when the study started, they were included as long-term patients on day 1 of the study but others entered the study once they reached their 181st treatment day. Patients were excluded if they had end-stage kidney disease (defined by use of ICD-10 [International Classification of Diseases 10th Revision] codes), a mechanical heart valve, antiphospholipid antibody syndrome, or underlying mitral stenosis. Follow-up was continued until patients experienced a primary outcome, discontinued OAC, switched to a different OAC, died, moved out of the Greater Reykjavik area, or once 28 February 2019 was reached, whichever came first. Patients who were switched between anticoagulants at any time during the study period reentered the study as new patients, that is, patients having received long-term anticoagulation who switched from one OAC to another during the study period without anticoagulation interruption were observed as being on long-term anticoagulation with the new agent immediately after the switch. When these patients reentered the study as new patients, their baseline characteristics were updated to reflect their baseline status at the date of study reentry (ie, date of treatment switching).

DOAC dosing was as prescribed in clinical practice. All warfarin dosing was managed by experienced staff at the 2 centralized anticoagulation management services (AMS) in Reykjavik: the Landspitali National University Hospital and the Cardiology Clinic. The DAWN anticoagulation management software (4-S Ltd, Penrith, England) was used for dosing. Monitoring tests were performed using citrated plasma obtained by phlebotomy. Only PT-INR (using the prothrombin-proconvertin time [P&P] test or Owren´s PT test) was measured from 1 March 2014 until 30 June 2016. Thereafter, Fiix-NR replaced the PT-INR at the Landspitali National University Hospital but PT-INR was in continued use at the Cardiology Clinic.¹⁴ The warfarin anticoagulation target range was the same with both warfarin-monitoring methods, that is, an NR of 2 to 3 for both PT-INR and Fiix-NR. The Fiix observation period started on the day of each warfarin-treated patient’s first Fiix-monitoring test. Switching occurred for all Landspitali National University Hospital patients within 3 months of 1 July 2016.

The exposure of interest was long-term anticoagulation with apixaban, dabigatran, rivaroxaban, PT-monitored warfarin, or Fiix-monitored warfarin. The planned comparisons were the rates of principal effectiveness outcomes, that is, total TE (composite of ischemic stroke, transient ischemic attacks [TIAs], systemic embolism, acute myocardial infarction, and venous TE), all-cause death, rate of the principal safety outcome (total major bleeding), and rate of a composite of the principal effectiveness outcomes (total TE or all-cause death) in the separate treatment groups using Fiix-warfarin outcome rates as reference. Albeit similar, the composite principal effectiveness outcome definition differs from that of the INVICTUS trial in that hemorrhagic stroke is not included.¹⁸ Major bleeding was defined a priori according to the International Society on Thrombosis and Haemostasis criteria for nonsurgical bleeding, that is, fatal bleeding; bleeding leading to a decrease in hemoglobin level of ≥20 g/L from the patient’s baseline hemoglobin; transfusion of ≥2 units of packed red blood cells; or symptomatic bleeding into a closed compartment such as intracerebral, subdural, intraspinal, intraocular, retroperitoneal, intrapericardial, or intraarticular.¹⁹ 

For patients who were treated during both the Fiix and PT periods, we calculated patient number of annual monitoring tests, testing interval, NR distribution, NR between-test variability (variance growth rate B2 formula),²⁰ and mean TTR by the Rosendaal method.²¹ 

Electronic medical records and AMS records were searched for vascular events using ICD-10 codes indicating bleeding or TE of any kind (eg, cerebral infarction, TIA, peripheral arterial embolism, myocardial infarction, and venous TE [deep vein thrombosis or pulmonary embolism], refer to the supplemental Table 1).^16,17 We included all searched major vascular events or deaths occurring while on an OAC and within 2 days of discontinuing a DOAC or within 5 days of discontinuing warfarin without emergency reversal of effect. Events occurring during temporary discontinuation of anticoagulation treatment (eg, in the setting of surgery) were excluded according to the same rules. All events identified by ICD-10 code search were then adjudicated independently by 2 coauthors of this report (A.B.I. and P.T.O.) that separately reviewed the patient charts. Discordant results were adjudicated by a third independent and blinded reviewer, also a coauthor of this report (E.S.B.). All-cause mortality was obtained using Iceland’s National Death Index. Causes of death were retrieved from hospital charts and the Icelandic Cause of Death Registry at the Directorate of Health in Iceland.

The Charlson Comorbidity Index (CCI) and the CHA₂DS₂-VASc score were used to evaluate the comorbidity burden and stroke risk of patients, respectively. The CHA₂DS₂-VASc score is a point-based system used to stratify the risk of stroke in AF patients. The acronym CHA₂DS₂-VASc stands for congestive heart failure, hypertension, age ≥75 years (doubled), diabetes, stroke (doubled), vascular disease, age 65 to 74 years, and sex category (female).^22,23 The ICD-10 codes used for the CCI were based on a previously validated algorithm.²⁴ Treatment indication was identified using AMS’ records and ICD-10 codes. Comorbidities and treatment indication were determined based on data recorded up until the day of study entry. Manual chart review was performed if treatment indication was missing or ambiguous. Information on concomitant drug use within 6 months of study entry was retrieved in the same manner as for OACs using anatomic therapeutic chemical codes, see supplemental Table 2.

Rates of adverse outcomes were compared in patients receiving each of PT-warfarin, apixaban, dabigatran, and rivaroxaban, with Fiix-warfarin as reference. To account for potential selection bias in different treatment groups, IPTW was used to yield balanced study groups. IPTW assigns weights to patients based on propensity scores calculated from potential confounders, thereby creating a balanced pseudopopulation that includes the whole study population. The IPTW model accounted for age, sex, all variables in the CCI (except for AIDS, which was too sporadic), hypertension, bleeding or coagulation disorders, history of venous TE or gastrointestinal bleeding, treatment indication, and concomitant drug use (antihistamines, antiplatelet agents, corticosteroids, nonsteroidal anti-inflammatory drugs, proton pump inhibitors, and selective serotonin reuptake inhibitors). Standardized mean differences were used to measure balance between study groups before and after IPTW; values of <0.1 indicate ideal balance and values of <0.2 indicate acceptable balance.^25,26 The primary effectiveness and primary safety outcomes were analyzed independently, that is, patients who experienced 1 of the principal outcomes were not censored from the analysis of the other principal outcomes. Repeated events of the same principal outcome were not included. We did not adjust for multiple comparisons, and we considered secondary outcomes to be exploratory.^27,28 

Study outcomes were compared using propensity score–weighted Cox regression, and the data were visualized using propensity score–weighted Kaplan-Meier survival curves. The proportional hazards assumption of the Cox models were assessed with the Schoenfeld residual test using the cox.zph function in R. We compared surrogate parameter data in warfarin-treated patients with nonparametric tests, the Mann-Whitney U test for unpaired data, and the Wilcoxon matched-pairs signed-rank test for paired data. We performed sensitivity analyses after stratifying patients based on whether they had a known ischemic heart disease or not and whether they had a venous TE or not, and after excluding PT-warfarin–treated patients monitored at the private Cardiology Clinic.

Statistical analysis was performed with R, version 4.2.2 (R Foundation for Statistical Computing) using RStudio, version 2023.06.2+561. The “twang” package was used to calculate propensity score weights for nonequivalent groups (https://cran.r-project.org/package=twang), and the “survey” package was used to fit the propensity score–weighted Cox regression models.²⁹ All statistical tests were 2-tailed. Missing data were <5% and no imputation was performed. Although P values <.05 or a 95% confidence interval (CI) excluding 1.0 for the hazard ratio (HR) is arbitrarily defined as statistically significant, we stress that results should also be judged in relation to their effect size and CIs.^27,28 

Clinical data could be retrieved from electronic medical records for all patients. During the 5-year period from 1 March 2014 to 28 February 2019, nationwide, 16 882 patients were prescribed with any type OAC for all different indications, of whom 6417 patients residing in the Greater Reykjavik area received anticoagulation long term (for >180 days) with OACs for nonvalvular AF, and only those are included in this analysis (n [median observation years]): Fiix-warfarin, 1257 (1.9); conventional PT-warfarin, 1904 (2.1); apixaban, 1171 (1.3); dabigatran, 549 (2.7); and rivaroxaban 1536 (2.0) (Figure 1). Patients had different oral anticoagulation experiences before entering the study as patients on long-term anticoagulation, that is, Fiix-warfarin, median 7.5 years; PT-warfarin, median 5.3 years; apixaban, median 0.5 years; dabigatran, median 0.7 years; and rivaroxaban, median 0.5 years. The following number [percentage] of the long-term patients were new OAC prescriptions during the study period:Fiix-warfarin, 65 (5%); PT-warfarin, 206 (11%); apixaban, 913 (78%); dabigatran, 250 (46%); and rivaroxaban, 1200 (78%). Overall, 17 patients on PT-warfarin (0.9%), 264 on apixaban (22.5%), 49 on dabigatran (8.9%), and 256 on rivaroxaban (16.7%) entered the study after having been switched from a different oral anticoagulant. Almost all patients on Fiix-warfarin entered the study after having previously been on a different oral anticoagulation regimen (first and foremost PT-monitored warfarin), ie, 1190 of 1257 (94.7%) patients.

Other differences are outlined in Table 1 and are accounted for in the IPTW model, including age differences, the DOAC-treated patients being younger at baseline (apixaban [mean, 74.4 years], dabigatran [mean, 71.1 years], rivaroxaban [mean, 70.7 years]) than the PT-warfarin–treated patients [mean, 75.9 years], and Fiix-warfarin–treated patients [mean, 77.7 years]). There were no differences in the baseline presence of ischemic heart disease, peripheral vascular disease, or diabetes mellitus, and there were no differences in antiplatelet agent use between the study groups. As shown in Table 1, IPTW of the 5 patient groups provided balanced baseline characteristics between treatment groups (standard measured difference of <0.2). In the overall study population of 6417 patients, 224 experienced TE, 337 died, and 340 had major bleeding.

The rates of weighted effectiveness outcomes in the separate treatment groups are shown in Figure 2. Only weighted results are discussed in the text. TE occurred at the lowest weighted annual incidence per person-year (py) with Fiix-warfarin, that is, 1.1% vs 1.9% with PT-warfarin (HR, 1.86; 95% CI, 1.19-2.91), 1.9% with apixaban (HR, 1.94; 95% CI, 1.13-3.32), 2.2% with dabigatran (HR, 2.19; 95% CI, 1.18-4.06), and 1.6% with rivaroxaban (HR, 1.58; 95% CI, 0.96-2.61). The lowest all-cause mortality rate was observed in Fiix-warfarin–treated patients (2.0%) vs 2.2% in PT-warfarin–treated patients (HR, 1.14; 95% CI, 0.82-1.58), 2.6% with apixaban (HR, 1.37; 95% CI, 0.93-2.02), 2.7% with dabigatran (HR, 1.30; 95% CI, 0.82-2.06), and 3.0% (HR, 1.48; 95% CI, 1.02-2.14) with rivaroxaban. Similarly, the composite outcome rate of total TE or all-cause death was lowest with Fiix-warfarin at 2.9%, compared with 4.5% for PT-warfarin (HR, 1.31; 95% CI; 1.00-1.72), 4.4% for apixaban (HR, 1.56; 95% CI, 1.11-2.16), 4.6% for dabigatran (HR, 1.57; 95% CI, 1.07-2.30), and 4.5% for rivaroxaban (HR, 1.54; 95% CI, 1.13-2.08; Figure 2). The Kaplan-Meier curves in Figure 3 show the cumulative occurrence of both raw and weighted principal effectiveness outcomes over time. Secondary exploratory outcomes are shown in web supplemental Table 3 that shows that the lower TE rate for Fiix-warfarin seemed to be mostly driven by lower rates of myocardial infarction.

The weighted major bleeding rate (Figure 2) did not differ between different OACs, that is, Fiix-warfarin 2.7% vs PT-warfarin 2.5% (HR, 0.89; 95% CI, 0.65-1.22), apixaban 2.4% (HR, 0.86; 95% CI, 0.57-1.30), dabigatran 2.2% (HR, 0.77; 95% CI, 0.48-1.23), and rivaroxaban 3.0% (HR, 1.07; 95% CI, 0.76-1.50). The Kaplan-Meier curves in Figure 4 show the cumulative occurrence of both raw and weighted primary safety outcomes over time. Secondary outcomes are shown in supplemental Table 3.

Results did not differ when patients were stratified based on presence or absence of prior ischemic heart disease, presence or absence of venous TE, or when PT-warfarin–treated patients treated at the private Cardiology Clinic were excluded (data not shown).

As shown graphically in Figure 5, compared with PT-warfarin–treated patients, Fiix-warfarin–treated patients had (median, interquartile range) fewer annual monitoring tests (10 [7-18] vs 14 [9-26], P < .0001), longer testing interval (35 [20-56] vs 26 [14-42] days; P < .0001), and lower between-test NR variability (variance growth rate [VGR], 0.08 [0.03-0.16] vs 0.12 [0.06-0.31]; P < .0001). The mean (median [interquartile range]) TTR was 76% (91 [54-100]) in Fiix-warfarin–treated patients vs 71% (83 [50-100]) in PT-warfarin–treated patients; P = .13.

This observational study is the first study to compare Fiix-monitored warfarin with apixaban, rivaroxaban, and dabigatran in addition to conventional PT-monitored warfarin in patients with AF. The study found Fiix-monitored warfarin to be associated with the lowest composite primary effectiveness outcomes (TE or all-cause death). This was explained, first by the lowest total TE rates being observed with Fiix-warfarin although confidence intervals did not exclude a null-effect for rivaroxaban and, secondly, by the lowest all-cause death rates being observed with Fiix-warfarin although the CIs could only exclude a null effect when compared with rivaroxaban. Major bleeding was similar in all study groups.

The preplanned principal outcomes in this study were total TE of any kind, all-cause death, a composite of total TE or all-cause death, and major bleeding. We anticipated lower total TE outcomes in Fiix-warfarin–treated patients compared with PT-warfarin based on the effect size previously observed with the same end points in the per-protocol and post hoc long-term patient analysis of the Fiix RCT,¹³ and the follow-up before-and-after real-world practice study.¹⁴ This study, however, extends those findings by also suggesting better overall effectiveness outcome with Fiix-monitored warfarin in comparison with apixaban, rivaroxaban, and dabigatran. The results of this study and our 2 prior clinical Fiix-monitoring studies^13,14,30 also support the concept that there is a more stable antithrombotic effect present during Fiix-warfarin management than during conventional PT-warfarin management. This is indicated by lower NR variability and consequent fewer annual monitoring tests with Fiix monitoring. Although, unfortunately, warfarin dosing data are lacking in this study, less warfarin dose variation and smaller dose changes were observed with Fiix monitoring in the before-and-after study,¹⁴ suggesting that less-variable dosing was a consequence of reduced NR variability. Finally, and most importantly, our prior studies also found that lower TTR,³⁰ more variable NRs,^14,30 and more variable dosing,^13,30 all indicators of unstable anticoagulation, were associated with higher rates of TE complications and bleeding events. Based on all of this, we contend that the traditional monitoring method of warfarin invented in 1935, the PT-INR, because of its sensitivity to rapid fluctuations in factor VII, confounds warfarin dose management with consequent suboptimal warfarin effectiveness outcomes. Nevertheless, we stress that only an adequately sized RCT can confirm a clinical improvement with Fiix-warfarin over DOACs. However, the importance of achieving stable vitamin K antagonist (VKA) effect may be supported by a recent German insurance claim study including 570 137 patients that found significantly lower mortality in patients undergoing anticoagulation with a long-acting VKA (phenprocoumon) compared with apixaban, rivaroxaban, and edoxaban, a finding that may support our hypothesis that low day-to-day anticoagulation variability may improve VKA clinical outcome.³¹ 

In this study, we included all patients with nonvalvular AF residing in the Greater Reykjavik area receiving long-term oral anticoagulation during a 5-year period. We used IPTW, a statistical method that simulates randomized study groups in observational studies, based on age, sex, comorbidities, and concomitant medication use to yield similar treatment groups. The weighted treatment groups had similar characteristics as evident from the standardized mean differences, which show that the IPTW process was successful. Nevertheless, being an observational study, conclusions can only be considered suggestive, and several limitations must be considered.

First, our study is limited by the treatment groups being small and the principal outcomes being based on relatively few event rates. Lower event rates in the secondary analyses can result in false-positive or false-negative findings. Therefore, we emphasize the preplanned principal outcomes without emphasizing potentially misleading smaller secondary analyses findings that we consider to be exploratory, for example, intracerebral hemorrhage or acute myocardial infarction (MI). For example, a secondary analysis finding of this study suggests that reduced TE is mainly driven by low rate of acute MI in Fiix-warfarin–treated patients. This is in opposite to our previous before-and-after study that found in a secondary analysis that patients with AF treated with Fiix-warfarin had reduced risk of cerebral ischemic events but not of MI when compared with PT-warfarin.¹⁴ The prior study observed a shorter treatment period in the same patient material but used a different definition of long-term OAC treatment (>90 days on OAC), so relatively more patients who had recently started on warfarin were included. However, in both studies and in the Fiix trial,¹³ the principal effectiveness outcomes were reduced to the same degree with Fiix-warfarin compared with PT-warfarin. Could this indicate that ischemic stroke prevention is more evident in patients studied during early-phase anticoagulation but that prevention of acute MI weighs more later on? A much larger study is needed to assess this; and these contrasting findings, in our opinion, illustrate why small secondary analysis findings should only be considered to be hypothesis generating. Be that as it may, because it is known that patients with AF have increased risk of MI^32-34 and that PT-warfarin is effective in both primary³⁵ and secondary³⁶ prevention of MI in patients with AF, we think that it is plausible that Fiix-warfarin with its less-variable anticoagulant effect not only reduces embolic events originating from a stagnant left atrium in patients with AF but also more effectively reduces TE caused by a different pathogenetic mechanisms such as ruptured atheromas.

Second, although we would have preferred studying only OAC-naïve patients, we had to limit our analysis to patients on long-term anticoagulation (prevalent users), having received anticoagulation for at least 180 consecutive days because the number of warfarin- naïve patient prescriptions rapidly decreased after the approval of DOAC drugs. We argue that our long-term definition (>180 days on OAC) is appropriate because studies typically show that major adverse events in patients on anticoagulants occur at the highest rate, in particular for warfarin, during the initial 2 to 3 months of OAC treatment but thereafter at a relatively steady rate.^37-39 Third, because Fiix-warfarin was only used during the second 2.5 years of the study observation period, an unmeasured confounder such as prevalent user bias (eg, gradual attrition of high-risk patients) cannot be excluded although the successful treatment group weighting (IPTW) should reduce the magnitude of such bias.⁴⁰ Fourth, multimorbidity and polypharmacy likely is more prevalent in the oldest patients and those prescribed longest, namely the warfarin-treated patients, and this could affect outcomes.^41,42 However, the magnitude of this bias is likely small given that we accounted for multiple comorbidities and concomitant drug use in our IPTW model and as the weighted mean Charlson index was similar for all groups. Supporting this, the observed absolute outcome rates with both PT-warfarin and DOACs in this study are consistent with those reported in the pivotal DOAC trials.^1-5 Likewise, the observed TE reduction with Fiix-warfarin is quantitatively similar to that found compared with PT-warfarin in the randomized blinded Fiix trial performed at our center¹³ and this may further support the current observations. Fifth, relatively minor events may have been missed more often in patients receiving DOACs because of a “nonreferral bias,” for example, short-lasting TIAs or self-limited but significant nosebleeds may not have been documented or seen at the hospital when there was no perceived need on behalf of patients or physicians to measure or reassess anticoagulation. Sixth, we do not have data on successful AF-reversal therapy during the study period that could affect TE risk. However, we do not expect many patients on long-term anticoagulation to have undergone successful cardioversion or AF ablation, in particular not in the warfarin groups that were older and had been treated longer before entering the study than those in the DOAC groups.

Although our studies have hitherto mainly focused on anticoagulation in moderate-risk patients, prior clinical studies support the notion that OAC antithrombotic mechanisms may differ with different anticoagulation indications. Thus, RCTs have found conventional PT-warfarin to be more effective and safer than DOACs in patient populations with high thrombogenic risk, namely in patients with mechanical heart valves,^43,44 high-risk antiphospholipid antibody syndrome,⁴⁵ and rheumatic heart disease–associated AF.¹⁸ This could indicate that in these high-risk patient groups, warfarin affects pathogenetic mechanisms that differ from those causing embolism in nonvalvular AF.⁴³ Multicentric RCTs would be needed to study whether further benefit could be achieved with Fiix-warfarin in these populations. We suggest that this is possible because we have found Fiix monitoring to reduce INR variability more in those with the high INR target of 2.5 to 3.5 than in those with lower INR targets.¹⁴ 

Finally, we highlight that external validation of our findings remains lacking and studies by independent investigators at other sites would be valuable. An independent RCT comparing Fiix-warfarin with PT-warfarin in patients with mechanical heart valves is in planning. Also, we stress that the Quick PT will remain a valuable diagnostic test of coagulopathies, for example, for the assessment of liver disease, because the Fiix-test is only intended for the purpose of VKA monitoring.

In summary, this propensity score–weighted observational study of patient with AF on long-term anticoagulants provides suggestive real-world evidence that, compared with Fiix-warfarin, significantly higher total TE rates occur with conventional PT-monitored warfarin, apixaban, and dabigatran, and nonsignificantly higher rates occur with rivaroxaban, without an increase in major bleeding. These results, together with reduced warfarin anticoagulation variability, previously reported reduced dose variation, and reduced testing frequency with Fiix monitoring compared with PT monitoring in patients who were monitored head-to-head by both methods suggest that warfarin can be improved and made less inconvenient by changing its monitoring method. Furthermore, the results may suggest that Fiix-warfarin could also be more effective than the more convenient DOAC drugs for patients with AF. Therefore, although only an adequately sized RCTs comparing Fiix-warfarin with other OACs can generate definitive conclusions, warfarin, which may become an improved drug with optimized monitoring, should not yet be brushed aside by physicians.

The authors thank the nursing and laboratory technical staff of the Landspitali Anticoagulation Management Center for their highly valued work during the study.

This study was funded by the Icelandic Centre for Research (RANNIS grant 207113-051) and by the Landspitali University Hospital Research Fund (grant A-2018-012).

The study sponsors had no role in study design; in the collection, analysis, and interpretation of data; in the writing of the report; or in the decision to submit the manuscript for publication.

Contribution: A.B.I., E.S.B., and P.T.O. conceptualized and designed the study; A.B.I., A.S.A., B.R.G., E.R., and L.R.L. acquired data; A.B.I., B.R.G., E.S.B., J.P.H., P.T.O., R.P., and S.H.L. (statistician) analyzed and interpreted data; A.B.I., P.T.O., and R.P. drafted the work; A.B.I., E.S.B., J.P.H., P.T.O., and R.P. critically analyzed the manuscript; and all authors approved the final version of the manuscript.

Conflict-of-interest disclosure: B.R.G. and P.T.O. are inventors of the Fiix-test, which is patented by them, the Landspitali National University Hospital, and the University of Iceland. The remaining authors declare no competing financial interests.

Correspondence: Pall T. Onundarson, Laboratory Hematology, Landspitali National University Hospital, Hringbraut, K-building, Reykjavik 101, Iceland; email: [email protected].