A comparison of the effects of NOAC and VKA therapy on the incidence of dementia in patients with atrial fibrillation: A systematic review and meta‐analysis

Abstract Atrial fibrillation (AF) patients are more susceptible to dementia, but the results about the effect of oral anticoagulants (OACs) on the risk of dementia are not consistent. We hypothesize that OAC is associated with a reduced risk of dementia with AF and that nonvitamin K antagonist oral anticoagulants (NOAC) are superior to vitamin K antagonists (VKA). Four databases were systematically searched until July 1, 2022. Two reviewers independently selected literature, evaluated quality, and extracted data. Data were examined using pooled hazard ratios (HRs) and 95% confidence intervals (CIs). Fourteen research studies involving 910 patients were enrolled. The findings indicated that OACs were associated with a decreased risk of dementia (pooled HR: 0.68, 95% CI: 0.55–0.82, I 2 = 87.7%), and NOACs had a stronger effect than VKAs (pooled HR: 0.87, 95% CI: 0.79–0.95, I 2 = 72%), especially in participants with a CHA2DS2VASc score ≥ 2 (pooled HR: 0.85, 95% CI: 0.72–0.99). Subgroup analysis demonstrated no statistical significance among patients aged <65 years old (pooled HR: 0.83, 95% CI: 0.64–1.07), patients in “based on treatment” studies (pooled HR: 0.89, 95% CI: 0.75–1.06), or people with no stroke background (pooled HR: 0.90, 95% CI: 0.71–1.15). This analysis revealed that OACs were related to the reduction of dementia incidence in AF individuals, and NOACs were better than VKAs, remarkably in people with a CHA2DS2VASc score ≥ 2. The results should be confirmed by further prospective studies, particularly in patients in “based on treatment” studies aged <65 years old with a CHA2DS2VASc score < 2 or without a stroke background.


| INTRODUCTION
Atrial fibrillation (AF) is an increasingly prevalent arrhythmia that represents a substantial load on citizens and healthcare systems across the world; additionally, it causes various complications, including heart failure and vascular thromboembolic events. [1][2][3] Dementia can result in cognitive impairment rather than a specific illness; the predominant forms of dementia consist of Alzheimer's disease (AD) and vascular dementia. 4 Microembolism developed in AF patients may lead to intracranial circulation disturbance, while decreased cardiac output may worsen cerebral hypoperfusion. Additionally, anticoagulation therapy may heighten the risk of hemorrhage and cause microbleeds inside the brain.
All of these factors, with the addition of stroke, are strongly associated with cognitive disorders. 5,6 Increasing evidence has shown that the dementia risk is 1.4 times greater in AF patients than in non-AF patients, 2.7 times greater in AF patients with stroke, 7 and 27% greater in AF patients without stroke. 8 Accordingly, additional investigations are necessary to examine whether appropriate management of AF might have the potential to impede or postpone the emergence of dementia. Some guidelines recommend the use of oral anticoagulants (OACs) to prevent stroke among AF patients. 9 Additionally, there is evidence that OACs are linked to a decreased prevalence of dementia, thus indicating that early anticoagulation therapy might be beneficial for cognitive protection. [10][11][12][13] Compared with traditional OACs, such as warfarin, a kind of vitamin K antagonist (VKA), the nonvitamin K antagonist oral anticoagulant (NOAC) has been shown to have the advantages of a stable anticoagulant effect, rapid onset of action, fewer individual differences, and a lower risk of intracranial hemorrhage complications. A more current study demonstrated that NOAC was more effective than VKA in reducing the overall incidence of dementia. 14 However, it should be noted that the results of this study were derived from numerous randomized controlled trials (RCTs), in which dementia was reported as an adverse effect and the definition was unclear. Furthermore, several large cohort studies did not reveal a notable disparity between NOACs and VKAs. 13,[15][16][17] Thus, it remains unclear whether NOAC or VKA is superior for preventing dementia in AF patients.
To date, doctors may choose to temporarily offer OACs to those with normal cognition, which might increase the number of dementia diagnoses soon after the patient's registration. Therefore, dementia diagnosed at the beginning of follow-up, known as the "blank period," should be excluded. Unfortunately, the pooled risk ratio of studies with an observational blank period was not statistically significant. 18 Moreover, varying definitions of anticoagulation were used when grouping in a number of cohort studies, which may directly affect the statistical results of the study. Regretfully, there was no comprehensive study being conducted to examine this matter.
Therefore, in light of the seven most recent studies, a systematic review and meta-analysis were carried out to explore the beneficial effects of anticoagulant medication for predicting dementia in AF patients and to determine whether NOACs are more effective than VKAs. In addition, the effects were evaluated via subgroup analyses based on age, CHA 2 DS 2 -VASc score, the inclusion or exclusion of a blank period, the inclusion or exclusion of stroke patients, and the definition of anticoagulation.

| METHODS
The Preferred Reporting Items for Systematic Reviews and Mete-Analyses (PRISMA) standards were followed for conducting this study, 19 as shown in Supporting Information: Table 1. It was not registered in PROSPERO.

| Data sources and search strategy
To find all qualifying studies from their beginning to July 1, 2022

| Selection criteria
Two authors (W. W. and W. F.) separately searched for articles using a standard procedure. In the event of discordance between them, another researcher (Y. S.) provided suggestions until a consensus was reached.
The retrieved articles were arranged using Endnote Version x9.1.
The inclusion criteria were as follows: (a) AF patients older than  Low, moderate, and excellent quality are represented by scores ranging from 0 to 3, 4 to 6, and 7 to 9, respectively. Two authors (W. W. and W. F.) separately completed quality assessments, with discrepancies resolved by another author (Y. S.).

| Statistical analysis
The HR and 95% CIs for each risk factor were aggregated and evaluated using Stata 17.0. Heterogeneity was taken into account while choosing the pooled data for the model and was evaluated using I 2 statistics and p values. A random-effects model was used to examine data with high heterogeneity (classified as I 2 > 50% or p < .1), whereas a fixed-effects model was employed in other situations. 21 To investigate the possible effect of heterogeneity, subgroup analyses were conducted based on age stratification (<75 vs. ≥75 years and <65 vs. ≥65 years), CHA 2 DS 2 -VASc score (<2 vs. ≥2), presence of a blank period, exclusion of stroke, the definition of anticoagulation (studies in which patients received anticoagulants for more than a certain proportion of time during the follow-up period were classified as "based on treatment," while studies in which patients were classified according to their initial prescription were classified as "intention to treat"), and specific types of NOAC. Publication biases were assessed with funnel plots, examining with Begg's and Egger's tests.

| Study selection
Above all 850 entries were found in the online database. Following deleting 212 duplicate records, reading the headings and abstracts, and applying screening criteria, 27 articles were found to be eligible.
Therein, some relevant RCTs were excluded because dementia was reported as an adverse side effect rather than as the primary outcome, as shown in Supporting Information: Table 3 filtering process is displayed in Figure 1.
Reporting Items for Systematic Reviews and Meta-analyses; HR, hazard ratio.

| The effect of OACs on the risk of dementia in AF patients
As shown in Figure 2, the dementia risk in AF patients was compared between patients with OACs and non-OACs in 8 stud-

| The effect of NOAC on the risk of dementia in AF patients compared with VKA
As shown in Figure 3, the prevalence of dementia in AF individuals was compared between NOAC and VKA in nine studies 11,13,[15][16][17][26][27][28]31 (n = 809 467). The pooled analysis indicated that people receiving NOAC treatment were related to a potential reduction of dementia risk than those treated with VKA (pooled HR: 0.87, 95% CI: 0.79-0.95). A random-effects model was employed since there is a considerable variability among research (I 2 = 72%, p < .001).

| Subgroup analyses
Several subgroup analyses were performed because of the substantial level of heterogeneity when comparing OACs with non-OACs and NOACs with VKAs ( proven to double the danger of dementia. 38 Obviously, the result was more credible when examining participants without a stroke background. However, it was revealed that those with AF might have a greater dementia danger, which was independent of strokes or acute ischemic events. 39 Thus, further investigations are required to corroborate the outcomes.
Furthermore, our subgroup analysis showed that NOAC had a stronger protective effect than VKA regardless of whether a blank period was included, which was different from the findings of a previous meta-analysis. 18 Indeed, dementia cases were more likely to appear within the initial year of monitoring, suggesting that the participants were already in the prodromal stage of dementia development before they were enrolled. 40 Thus, the blank period may help to minimize the overestimation of pharmacological protective effects. In light of this, the pooled HR with a blank period might offer a more accurate representation of the real effect of OACs or NOACs on dementia danger.
Moreover, regardless of the definition of anticoagulation used, the reduction of OACs on dementia risk in AF individuals was confirmed. However, this protective effect did not reach statistical significance when comparing NOAC to VKA in the "based on treatment" group, which guaranteed the duration of action of the drug and was more likely to reflect the actual effect of the drugs in reducing dementia risk in AF patients.
We excluded some RCTs because they reported dementia as an adverse side effect rather than the primary outcome. Additionally, there are several ongoing RCTs related to this topic that may provide further insights and clarify our uncertain conclusions, as listed in Supporting Information: Table 5. We will continue to monitor these studies and update our findings as necessary.

| Limitations
Our research still had some limitations that should be acknowledged.
First, while we used random effects models, most of the analyses within each group had a high level of heterogeneity, which could be attributed to differences in screening criteria, sample sizes, demographics of the study population, types of drugs studied, follow-up times, unmeasured confounding factors, and other variables. The factors utilized in every trial for statistical adjustment might have also attributed to the significant level of heterogeneity. Additionally, the present included data were observational studies with several potential biases that we were unable to control or measure. For instance, whether or what kind of medication the patient chooses might be influenced by other factors such as socioeconomics, which will affect the incidence of dementia.
Furthermore, due to the limited duration of follow-up in most studies, it cannot be guaranteed that the incidence of dementia was accurately assessed. Dementia is a slow-progressing disease, the development of which is influenced by age as a standalone risk factor, which could be unrelated to the diagnosis of AF. This makes establishing a logical link between them challenging. These limitations should be considered when interpreting our findings.

| CONCLUSIONS
The results of this systematic review and meta-analysis, which comprised 14 retrospective cohort studies, suggested an association between OACs and a lower risk of dementia in AF patients.
Additionally, NOAC was more effective than VKA, regardless of whether a blank period was included in the analysis, as well as in persons with a CHA 2 DS 2 -VASc score ≥ 2. However, the impact could not be determined, particularly in subpopulations such as those in "based on treatment" studies, those aged <65 years old, those with a CHA 2 DS 2 -VASc score < 2, or those without a stroke history.
Therefore, NOACs should be considered a preferred first-line option in AF individuals to minimize the dementia danger unless other factors outweigh the risk in favor of VKA. Accordingly, more highquality RCTs and cohort studies are expected to clarify the dementia risk in AF patients receiving various anticoagulants and to provide better guidance for clinical practice.