Efficacy and Safety of Non-recommended Dose of New Oral Anticoagulants in Patients With Atrial Fibrillation: A Systematic Review and Meta-Analysis

Introduction: The real-world treatment of atrial fibrillation (AF) often involves the prescription of new oral anticoagulants (NOACs) using dosing both lower and higher than recommended guidelines. Our study aimed to evaluate the efficacy and safety of non-recommended dosage of NOACs in AF patients. Methods: A systematic search was performed for relevant studies across multiple electronic databases (PubMed, Embase, Cochrane Library, Clinical Trials Registry) from inception to May 1, 2021. Multicenter randomized trials and observational studies were selected with key reporting measures for inclusion involved efficacy outcomes including stroke or systemic thromboembolism along with safety endpoints assessing major or clinically relevant bleeding events. Results: A total of 11 eligible studies were included involving 48,648 patients receiving recommended dose of NOACs and 50,116 patients receiving non-recommended dosage. Compared to AF patients treated with recommended dose regimens, administration of low dose of NOACs was associated with higher risk of stroke/systemic embolism (RR = 1.24, 95% CI 1.14–1.35, P < 0.00001), but without reducing bleeding risk (RR = 1.18, 95% CI 0.91–1.53, P = 0.21) and a higher risk of all-cause mortality (RR = 1.58, 95% CI 1.25–1.99, P = 0.0001). Moreover, high dose of NOACs was associated with higher risk of stroke and systemic embolism efficacy (RR = 1.71, 95% CI 1.06–2.76, P = 0.03) and a non-significant trend to a greater risk of major or clinically relevant bleeding (RR = 1.57, 95% CI 0.96–2.58, P = 0.07). Conclusions: AF patients treated with low dose of NOACs showed equivalent safety but with worse efficacy compared with recommended dose. High dose of NOACs was not superior to recommended dose regimens in preventing stroke/systemic embolism outcomes in AF patients.


INTRODUCTION
The past two decades have witnessed the gradual implementation of four new oral anticoagulants (dabigatran, rivaroxaban, apixaban, edoxaban) to prevent stroke in patients with nonvalvular AF. These agents have shown greater efficacy and safety compared with vitamin K antagonists (VKAs) (1) and have been widely approved by regulatory bodies including the European Medicines Agency (EMA) (2), the U.S. Food and Drug Administration (FDA) (3), and the Japanese Pharmaceuticals and Medical Devices Agency (PMDA) (4). Nevertheless, in realworld clinical practice, safety concerns among physicians have led to prescribing habits of lower dose of NOACs in patients with high HAS-BLED scores, along with considerations of age, body mass index (BMI), creatinine clearance (CrCl), hepatic function, as well as concomitant disease states (5). Alternatively, patients with significantly high CHA2DS2-VASc scores might be given higher dose of NOACs (6). These scenarios provide concerns that the efficacy and safety of NOACs may be compromised. At present, there are no universal rules for determining the dose regimens in high-risk patients, and no clear risk-benefit analysis was performed to help to adjust NOAC dose regimens. In this study, we performed a meta-analysis of multicenter randomized trials and observational studies to evaluate the efficacy and safety of non-recommended dose compared with recommended dose of NOACs in patients with AF.

MATERIALS AND METHODS
This systematic review and meta-analysis were performed following the recommendations of the Cochrane Handbook for Systematic Reviews of Interventions and Preferred Reporting Items for Systematic Reviews and Meta-Analysis statement (PRISMA).

Search Strategy
Two investigators independently performed comprehensive searches for all relevant articles against the following databases: PubMed, Embase, Cochrane Library databases and Clinical Trials Registry (www.clinicaltrials.gov). English language articles were searched from inception to May 1, 2021. The search terms keywords used were as follows: atrial fibrillation; new oral anticoagulants, non-vitamin K antagonist, NOAC, oral thrombin inhibitors, factor Xa inhibitors, apixaban, rivaroxaban, edoxaban, dabigatran; non-recommended dose, off-label dose, low-dose, reduced-dose, underdosing, high dose, overdose; Abbreviations: AF, atrial fibrillation; NOACs, new oral anticoagulants; VKA, vitamin K antagonists; EMA, European Medicines Agency; FDA, U.S. Food and Drug Administration; PMDA, Japanese Pharmaceuticals and Medical Devices Agency; BMI, body mass index; CrCl, creatinine clearance; PRISMA, Preferred Reporting Items for Systematic Reviews and Meta-Analysis statement; NOS, Newcastle-Ottawa Scale; SE, systemic embolism; ISTH, International Society on Thrombosis and Hemostasis; RR, risk ratios; OC, observational cohorts; RCT, randomized controlled trials; NA, not available; MI, myocardial infarction; TIA, transient ischemic attack; CNS, central nervous system; CV, cardiovascular; GI, gastrointestinal; MACE, major adverse cardiovascular events; M ± SD, mean ± standard deviation; IQR, interquartile range; RD, recommended dose; LD, low dose; HD, high dose; CI, confidence interval. stroke; bleeding. Additionally, the reference lists of related review articles were also reviewed to source additional publications of relevance.

Inclusion and Exclusion Criteria
Manually selected studies with the assistance of EndNote software according to the prespecified PICOS criteria. Studies that met the following criteria were selected for inclusion involved: (1) P: patients with atrial fibrillation; (2) Intervention: treated with non-recommended dose (low or high dose) of new oral anticoagulants (NOACs: dabigatran, rivaroxaban, apixaban, edoxaban); (3) Comparison: with recommended dose of NOACs; (4) Outcomes: reported at least one of the following adverse outcomes: stroke/systemic embolism, major bleeding, all-cause mortality, cardiovascular cause of death by dose subgroups; (5) Study design: RCTs and observational studies published as full articles.
The exclusion criteria were as follows: (1) Compared anticoagulation strategies using NOACs vs. warfarin; (2) Other types of studies, including case reports, animal experiments, meta-analysis, reviews, comments, editorials, and conference abstracts; (3) Reported exposure or endpoints not suitable for our analysis.

Data Extraction and Quality Assessment
Two investigators independently extracted data from the eligible studies. Baseline characteristics of the patients (age, sex, creatinine clearance, BMI, CHADS2/CHA2DS2-VASc score, HAS-BLED score), study design, follow-up periods and the prespecified adverse outcomes were retrieved. In addition, they also assessed the quality of the RCTs and observational studies using the Revised Jadad's Scale and Newcastle-Ottawa Scale (NOS), respectively. Any disagreements or uncertainties between the two reviewers in the processes of study selection, data extraction and quality assessment were resolved by discussion with the senior investigator (HL).

Definitions and Outcomes
The criteria of standard dose approved by guidelines lately as well as the definitions of recommended/low/high dose of NOACs in our selected studies are listed in Table 1 (5,(7)(8)(9)(10)(11)(12)(13)(14)(15)(16)(17)(18)(19). The primary efficacy outcomes were stroke or systemic embolism (SE) and the primary safety outcomes were major or clinically relevant bleeding. For trials reporting only major bleeding, the same data were used for major or clinically relevant bleeding. The definition of major bleeding was in accordance with the International Society on Thrombosis and Hemostasis (ISTH) criteria (10,20). The secondary outcomes included allcause mortality, cardiovascular cause of death, if reported data were available.

Statistical Analysis
Comparison of the treatment effects of non-recommended dose of NOACs (low/high dose) vs. recommended dose of NOACs was performed used risk ratios (number of events or the incidence in each treatment group) and respective 95% confidence intervals. The heterogeneity across the studies was assessed by Cochran's Q-test (P < 0.1 was regarded as statistically significant) and I 2 statistics, which estimate heterogeneity quantitatively (I 2 value < 25% indicates no or mild heterogeneity, I 2 >75% indicates high heterogeneity). If I 2 ≤ 50% and P ≥ 0.1, the fixed-effects model was used. If I 2 > 50% or P < 0.1, data were pooled used random-effects, according to the Mantel-Haenszel model, and the cause of heterogeneity was sought.
Publication bias assessment was made through visual inspection of the asymmetry in funnel plots. All statistical analyses were performed using Review Manager software (Rev-Man) version 5.3 (Cochrane Collaboration 2014, Nordic Cochrane Center, Copenhagen, Denmark). Two-tailed P-values < 0.05 were considered significant.

Study Selection and Characteristics of the Eligible Studies
Our search strategy identified 1,316 potentially relevant studies, of which 670 records were retrieved after duplicates were removed. After assessing titles and abstracts, a total of 36 full-text articles were eligible for further screening. Of these, 25 records were excluded for not meeting the eligibility criteria, leaving 11 articles for further consideration (5,(9)(10)(11)(12)(13)(14)(15)(17)(18)(19). The study selection flowchart is presented in Figure 1.
Overall assessment of the 11 eligible studies showed data for 100,131 total patients to be included in the meta-analysis. Of these, 48,648 received recommended dose of NOACs whereas 50,116 received non-recommended dose of NOACs. Among the latter group, 49,384 patients were prescribed lower than recommended dose while 732 patients were given higher dose. Articles were published from 2011 to 2021, and most of them were non-randomized, observational and comparative studies. Importantly, the 11 identified trials were rated good for methodological quality according to assessment by NOS and Revised Jadad's scales. The general features of the 11 studies are summarized in Table 2 with corresponding reported endpoints and follow-up periods listed in Table 3.

Baseline Characteristics of Patients Enrolled in the Eligible Studies
The baseline demographic characteristics of AF patients treated with low/high dose of NOACs vs. recommended dose are presented in Tables 4, 5, respectively. The patients receiving nonrecommended dose tended to be elderly, more likely be female, have low body weight, and show high HAS-BLED or CHA2DS2-VASc scores.

Efficacy Outcomes
The results of our pooled indicated that the risk of stroke/systemic embolism in AF patients using nonrecommended dose of NOACs was significantly higher than patients using the recommended dose (RR = 1.25, 95% CI 1.15-1.36, P < 0.00001). Moreover, similar results were also observed when either low dose (RR = 1.24, 95% CI 1.14-1.35, P < 0.00001) or high dose groups (RR = 1.71, 95% CI 1.06-2.76, P = 0.03) were compared against the recommended dose group (Figure 2).
We next conducted subgroup analyses comparing the individual NOAC agents separately between the nonrecommended dose and recommend dose categories. Notably, we found that patients prescribed low dose of NOACs, no matter dabigatran, apixaban or edoxaban, showed a higher risk of stroke/systemic embolism than patients administered the recommended dose. In contrast, patients receiving nonrecommended dose of rivaroxaban (low/high) were not at greater risk of stroke/systemic embolism than the standard dose ( Table 6).
Subgroup analyses revealed that the risk of all-cause mortality in patients receiving low dose of dabigatran, rivaroxaban or apixaban, was significantly higher than the recommended dose

Cardiovascular Cause of Death
Patients treated with non-recommended dose of NOACs were not more likely to experience a higher risk of cardiovascular death compared to patients receiving standard dose (RR = 1.24, 95% CI 0.76-2.03, P = 0.39), and this result was consistent when the nonrecommended dose patients were divided into low and high dose groups (RR = 1.27, 95% CI 0.76-2.12, P = 0.36; RR = 1.52, 95% CI 0.80-2.87, P = 0.20, respectively) ( Figure 5).

Publication Bias and Sensitivity Analysis
Based on a visual inspection of funnel plots (Supplementary Figures 1-4), there was no obvious evidence of publication bias for the selected studies for either reporting efficacy or safety outcomes. But performance bias in these studies inevitably existed. Random effects model showed similar results to the fixed-effects model. AF patients receiving low dose dabigatran, showed high major or clinically relevant bleeding risk in the sensitivity analysis after excluding Ezekowitz 2016 (I 2 = 0%, RR = 1.46, 95%   (1,21), showed low dose of dabigatran was associated with reduced bleeding risk (RR = 0.83,95% CI 0.74-0.91), suggesting that use of low dose dabigatran may reduce long term bleeding risk (Supplementary Figure 8). With regard to all-cause mortality, the pooled results remained stable after sensitivity analysis.

DISCUSSION
In routine clinical practice, a balance is struck between the risks of stroke and bleeding in AF patients, highlighting the need   (22). The ORBIT-AF study showed improved safety outcomes in AF patients with renal indications for NOACs  dose reduction where major bleeding outcomes were reduced although there were no significant differences encountered for stroke risk relative to patients receiving recommended dose (23). Yu et al. also suggested that Asian AF patients treated with higher dose of NOACs were at increased risk of stroke or systemic embolism, major bleeding, and all-cause mortality compared with recommended regimens (24). As shown in our meta-analysis, we found that treating patients with high dose of NOACs produced a trend of higher incidence of stroke or systemic embolism compared with AF patients under standard treatment. This difference may be explained considering the reason underlying the administration of higher dose was likely the associated ischemic comorbidities of these patients. Non-recommended (low or high) dose of NOACs were also significantly associated with a higher risk of death along with composite endpoints (stroke/systemic embolism, major bleeding and death) compared to patients receiving recommended dose. However, after adjustment for baseline characteristics, the composite outcomes, including death risk, were no longer significant, suggested that the higher rates of events associated with non-recommended dose were probably related to the underlying diseases rather than the dose itself (5). As part of our meta-analysis, we considered the effects of the individual agents as subgroup analyses. We found that patients treated with low dose dabigatran had higher associated risks of stroke/systemic embolism, but nevertheless, exhibited equivalent safety, compared with recommended dose. Consistently, Beyer-Westendorf et al. showed that low dose dabigatran produced increased risks of bleeding and stroke compared with AF patients receiving 150 mg bid (25).
Regarding rivaroxaban, we found patients who received nonrecommended dosing showed comparable efficacy to standard regimens, no matter low or high. Nevertheless, low dose rivaroxaban was associated with higher bleeding risk in patients, whereas the higher dose group showed equivalent safety compared to the recommended dose. This contrasts with the findings of the XAPASS study where the incidence rate of stroke/systemic embolism was significantly higher in patients who received low dose rivaroxaban. Moreover, after adjustment for baseline characteristics, the authors reported that the rates of major bleeding were similar in for patients receiving either low or recommended dose of rivaroxaban (26). Other comparisons can be made with the EXPAND and SAKURA AF registry studies where the incidence rates of stroke/systemic embolism and major bleeding were both comparable between rivaroxaban interventions using low and recommended dose (6,27).
Our subgroup analysis of apixaban-treated patients showed lower dose were associated with significantly higher risk of stroke and systemic embolism, major or clinically relevant bleeding compared to patients receiving the recommended dose. Our data are partly consistent with a prior study showing that low dose apixaban in patients with no renal indication for dose reduction was associated with significantly higher risk of stroke but without a reduction of major bleeding compared to patients taken standard dose of apixaban (23).
Lastly, the risk of stroke or systemic embolism in AF patients receiving low dose edoxaban was higher than patients treated with the recommended dose, but nonetheless this regimen had a superior safety outcome. This notion is supported by a published study reporting that the rate of ischemic stroke was higher with a low dose edoxaban regimen, but with greater risk reduction in bleeding compared to warfarin (28).
In summary, our pooled meta-analysis showed low dose of NOACs had worse efficacy for stroke/systemic embolism prevention, but without reducing in major bleeding, and with increased all-cause mortality compared to standard use. Moreover, high dose of NOACs did not show better outcomes in stroke/systemic embolism prevention in AF patients. Therefore, non-recommended dose of NOACs should be carefully prescribed since off-label use of any NOAC may cause different clinical outcomes. Furthermore, there may be subtle differences among the individual NOACs, such as pharmacologic properties, frequency of use, dose, interaction between drugs or other currently undefined aspects to consider. So, direct comparative studies are warranted to determine whether these are real differences in clinical efficacy and safety outcomes between individual NOACs.

LIMITATIONS
Some limitations should be acknowledged of this metaanalysis. First, pooled data were collected from different trials with differences in their design, including baseline participant characteristics, agents and dose regimens, definitions of efficacy and safety outcomes, along with differences in the length of follow-up time. Any combination of these variables could account for the source of moderate to high degree of statistically heterogeneity in our meta-analysis. This point needs further validation and it is also important to consider that heterogeneity can occur by chance and would almost certainly be found with meta-analysis involving many large studies (29). Furthermore, most of the studies included in our meta-analysis were nonrandomized observational studies and were limited by potential selection and ascertainment bias. Moreover, the efficacy or safety outcomes in some of the included studies were not fully adjusted with medical records or baseline characteristics. Therefore, clinical events might be associated with off-label use but not causally related to dosing, thus introducing a possible risk of bias and misclassification of the end points (12), indicating the need for caution in interpreting outcomes.

CONCLUSIONS
Based on the available data, our study concludes that AF patients treated with non-recommended dose of NOACs may be at higher risk of stroke/systemic embolism and all-cause mortality compared with recommended dose, albeit the benefit of without reducing in major bleeding. Of particular interest, dabigatran, rivaroxaban, apixaban, edoxaban all differed in their efficacy and safety outcomes with non-recommended dose regimens.

DATA AVAILABILITY STATEMENT
The original contributions presented in the study are included in the article/Supplementary Material, further inquiries can be directed to the corresponding author/s.