Discussion
The present systematic review and meta-analysis found that OAC use significantly decreased the risk of developing dementia in patients with AF by 28% compared with no treatment. But in the subgroup analyses with observational window, there is no statistical difference of the risk of dementia between OAC use and no treatment. The high heterogeneity and low SOE may impact the confidence of the results.
The main finding of this study, i.e. the effect of OAC use on the incidence of dementia, is in accordance with the previous two meta-analyses that demonstrated the protective effect of OACs on the incidence of dementia, but high heterogeneity and low SOE were the common limitation.8, 9 The present meta-analysis improved the quality of included studies by excluding the cross sectional studies and the dementia-included studies, and updated four studies. The results of the comparison of VKA use and non-OAC use, and VKA use and NOAC use on the risk of dementia, were consistent with the recent meta-analysis.9 However, in this study, NOAC use was not associated with the reduced risk of dementia compared with non-OAC use, which was not in line with the general perception. A recent meta-analysis, including 6 RCTs and 2 observational studies, found that NOACs might decrease the risk of cognitive impairment in comparison to VKAs/acetylsalicylic acid, with marginally significant level of synthetic effect for random-effects model (HR [95%CI] =0.77 [0.53, 1.01], I2=39.4%).32 The results ask for further RCTs for direct comparison of NOACs and VKAs on the risk of dementia. In the latest clinical guideline,1 the high risk of stroke, i.e. CHA2DS2-VASc score ≥ 2, is still the strong indication to prescribe OACs. In the present study, we found the similar protective effect of OACs on the incidence of dementia in AF patients with CHA2DS2-VASc score ≥ 2.
It is known that advanced age, shared risk factors, clinical and silent cerebral infarcts may link AF and dementia.2, 5 One of the leading potential mechanism is the silent cerebral infarcts, which are very common in patients with AF and are associated with reduced cognitive functioning. The AF-related silent cerebral infarcts are most prevalent in the frontal lobes which play an important role in executive functioning. Therefore, cumulative silent infarcts of the frontal circuit components may contribute to the development of dementia. Considering the antithrombotic role of OACs, it is a reasonable belief that the effective OAC treatment for AF-related infarcts may prevent the occurrence of dementia. This study may support the validity of this potential hypothesis. It is worth mentioning, the pooled RR is not stable in the subgroup analysis with the observational window. In the non-RCT study, physicians prefer to prescribe OACs to patients with good cognitive performance, which may cause the increased diagnosis of dementia within the short time after inclusion. It was seen that dementia cases occurring in the first year during follow-up were more prevalent in non-OAC users.10 The observational window may partly prevent the overestimation of the protective effects of OACs. To this end, the pooled RR with the observational window may be closer to the real effect of OACs on the risk of dementia. Therefore, although this meta-analysis indicated a significant association between OAC use and reduced incidence of dementia, the following limitations of the included studies need to be noted: i) the observational design for all included studies was the primary limitation, which may lead to the different decision of OACs prescription and the different proportion of complications between OAC use and no treatment patients. Although all included studies were adjusted for potential baseline covariates, the maximum adjusted covariates were varied, and some confounding factors may still remain; ii) it is a long progression from the normal cognitive performance to the mild cognitive impairment and the more serious abnormality of dementia.2 The long period may result in the varied effects of studies with different follow-up duration and different baseline age. In addition, the worse cognitive status prior to the stage of dementia exactly associated with the no prescription of OACs, which overestimated the positive effect of OACs on risk of dementia. The result of the subgroup analysis with the observational window presented the lower and insignificant pooled RR. Our study excluded the studies included patients with moderate to severe cognitive impairment, while much studies didn’t report the data of baseline cognitive status. iii) the adjusted covariates were based on the baseline data, but several covariates like medication management, complications, AF types, were time-varying, which result in the deviation from the true status of patients during follow-up. iv) the AF burden was significantly associated with the progression of AF-related complications, and a few study found that persistent/permanent rather than paroxysmal AF had a high burden to cognitive impairment,33 but all included studies didn’t report the data of AF types, let alone AF burden.