The current analysis examined patients enrolled in the registry of Acute Coronary Syndrome of the University Hospital of Vigo focused on Atrial Fibrillation (CardioCHUVI-FA). This retrospective registry included all patients diagnosed with AF in the health care area of Vigo between January 1, 2014 and January 1, 2018. To create this registry, patients were first identified by using administrative databases of both inpatients and outpatients. Specific patient groups were identified using the Complex Information Analysis System of the Galician Health Service for both primary care and inpatient care and with the codes 427.31 of the International Classification of Diseases-Ninth Edition and K78 of the International Classification of Primary Care. Of a total of 16 975 patients, we selected a subgroup of patients aged ≥ 85 years. In the second phase, the medical records of all patients were reviewed to confirm the diagnosis of AF (electrocardiographically documented) and to collect data on baseline clinical variables, therapeutic strategy, and events during follow-up.

Patients with a mechanical prosthesis or moderate-to-severe mitral stenosis were excluded. Also excluded were patients receiving chronic therapy with low-molecular-weight heparin.

Because less than 2% of the values were lost for each variable analyzed (n < 70), no specific method was applied to adjust for these missing values.

The study was performed in accordance with the principles of the Declaration of Helsinki and was approved by the Autonomous Research Ethics Committee of Galicia (code HAC-ACO-2018-01, record 2018/258).

The main clinical endpoint of this study was total mortality. Secondary endpoints were embolic and bleeding events. Embolic events comprised strokes (ischemic stroke and transient ischemic attack), pulmonary embolism, and peripheral embolism. Bleeding events comprised clinically relevant bleeding according to the definition of the International Society on Thrombosis and Haemostasis.8,9 Follow-up was completed for all patients at the time of their death or at the last date that they were proven to be alive. The anticoagulant decision for each patient was based on the clinical judgment of the patient's treating physician. Because the oral anticoagulation variable was time-dependent, patients who underwent treatment changes (from no anticoagulation to oral anticoagulation or vice versa) were classified according to the treatment approach at the time of their inclusion.

Our study specifically analyzed patients ≥ 85 years old with a diagnosis of “nonvalvular AF” according to European recommendations. Accordingly, patients with a mechanical prosthesis or mitral stenosis ≤ 1.5cm2 were not included. Patients were classified into 2 groups: with and without moderate-to-severe dementia, defined as cognitive decline between stages 5 and 7 of the Reisberg Global Deterioration Scale,10 which itself corresponds to stages of the Functional Assessment Staging scale ≥ 5.11 Accordingly, patients with mild cognitive decline or incipient/mild dementia were not included; at these stages, the patients have memory problems but are functionally autonomous for activities of daily living. The diagnosis of ischemic stroke was confirmed using concomitant imaging studies, including computed tomography and magnetic resonance. Transient ischemic attack was defined as temporary neurological dysfunction resulting from focal cerebral, spinal cord, or retinal ischemia, with no acute infarct lesion. Bleeding was defined according to the classification of the International Society of Thrombosis and Haemostasis and all clinically relevant events were included (both major and minor).8,9

Continuous variables are expressed as mean±standard deviation and were compared using the Student t test. Categorical variables are expressed as percentages and were compared using the chi-square test. The impact of oral anticoagulation on mortality, embolisms, and bleeding were evaluated using Cox regression analysis with robust estimation of variance; the variable oral anticoagulation was time-dependent. These analyses were adjusted by the variables associated with events in univariate analysis (table 1 of the supplementary data) or by those whose association with clinical events has been consistently shown in previous studies. Thus, multivariate analyses were adjusted by age, sex, hypertension, diabetes mellitus, history of ischemic heart disease, heart failure, previous stroke, chronic obstructive pulmonary disease, previous hospitalization for bleeding, anemia, glomerular filtration rate measured by the CKD-EPI equation (Chronic Kidney Disease Epidemiology Collaboration), left ventricular ejection fraction ≤ 40%, CHA2DS2-VASc and HAS-BLED scores, and treatment with antiplatelet agents, beta-blockers, digoxin, angiotensin-converting enzyme inhibitors or angiotensin receptor blockers, and proton pump inhibitors. The proportional hazards assumption was tested using the Schoenfeld residual test. The results are expressed as hazard ratios (HRs) with their 95% confidence intervals (95%CIs). Differences with P <.05 were considered statistically significant. The results are graphically displayed using Kaplan-Meier curves.

In addition, within the dementia patient group, the analyses were complemented with propensity score matching to balance the patients’ baseline characteristics with and without anticoagulation. This analysis used a 1:1 matching algorithm without replacement, with a 0.2 caliper for standard deviations and with the nearest neighbor matching method. The independent variables used to obtain the propensity score were age, sex, hypertension, diabetes mellitus, ischemic heart disease, previous stroke or embolism, previous heart failure or left ventricular ejection fraction ≤ 40%, history of bleeding, anemia, CKD-EPI <60mL/min/1.73 m2, CHA2DS2-VASc score, HAS-BLED score, antiplatelet therapy, and treatment with beta-blockers, angiotensin-converting enzyme inhibitors or angiotensin receptor blockers, digoxin, statins, and proton pump inhibitors. The statistical analyses were performed with SPSS version 25.0 and Stata MP64 version 15.0.

In total, we identified 3595 patients aged ≥ 85 years with a confirmed diagnosis of AF. Excluded from this group were those with a mechanical prosthesis or moderate-to-severe mitral stenosis (n=10) and those receiving chronic therapy with low-molecular-weight heparin (n=36). Accordingly, the final study population comprised 3549 patients (figure 1); 2382 were women. The mean age of the overall group was 88.9±3.2 years. Of these patients, 221 had moderate-to-severe dementia (6.2%): 68.3% had Alzheimer disease, 20.8% had vascular dementia, 7.2% had mixed vascular-degenerative dementia, and 3.6% had other causes of dementia. The differences in the baseline characteristics of the patients with and without dementia are shown in table 1. Patients with dementia were older and had higher rates of heart failure history and previous ischemic stroke and higher CHA2DS2-VASc and HAS-BLED scores.

Figure 1.

Flow chart of study patients.

(0.29MB).

Of the 3549 patients, 2522 (71.1%) were receiving oral anticoagulant therapy (figure 1). Fewer patients with dementia were receiving oral anticoagulants than patients without dementia (60.2% vs 71.8%; P <.001). The most frequently used anticoagulants were vitamin K antagonists (VKAs) (75.9% and 79.6% in patients with and without dementia, respectively; P=.598) and the 2 groups had a similar time in therapeutic range (> 60%; 45.5% vs 46.3%; P=.883). Table 2 shows the baseline characteristics of the patients with dementia according to whether they were receiving anticoagulant therapy or not.

During follow-up, 29 patients with dementia (14.2%) underwent changes in antithrombotic therapy (1 started anticoagulation, 10 discontinued anticoagulation, and 18 switched from VKAs to direct oral anticoagulants [DOACs]; no patients switched from DOACs to VKAs). Analysis revealed that 41.4% of the changes were due to poor control of the international normalized ratio in patients under treatment with VKAs, 20.7% due to high perceived bleeding risk, 10.3% due to high perceived embolic risk, and 27.6% for unknown reasons. Of the patients without dementia, 471 (13.1%) changed their antithrombotic therapy (16 started anticoagulation, 73 discontinued anticoagulation, 381 switched from VKAs to DOACs, and 1 switched from DOACs to VKAs). The results indicated that 44.4% of the changes were due to poor control of the international normalized ratio in patients under treatment with VKAs, 18.9% due to high perceived bleeding risk, 10.9% due to high perceived embolic risk, and 25.9% for unknown reasons (figure 1 of the supplementary data).

The patients were followed up for 2.8±1.7 years (2.9±1.7 years in the dementia group and 2.4±1.6 years in the group without dementia). Compared with patients without dementia, patients with dementia had higher mortality (annual incidence, 25.9% vs 13.7%; P <.001) but similar rates of embolic (annual incidence, 4.7% vs 3.6%; P=.219) and bleeding (annual incidence, 5.6% vs 5.3%; P=.765) events. Table 3 shows the baseline characteristics of the patients with dementia according to events during follow-up (mortality, embolisms, and bleeding). In addition, table 2 of the supplementary data shows the bleeding location according to the presence of dementia and whether anticoagulant therapy was prescribed or not.

After multivariate analysis, oral anticoagulation was not associated with lower mortality in patients with dementia (HR=1.04; 95%CI, 0.63-1.72; P=.880), in contrast to patients without dementia (figure 2). However, anticoagulant therapy was associated with a significant reduction in the rate of embolic events in patients with dementia (HR, 0.36; 95%CI, 0.15-0.84; P=.018), at the expense of a significant increase in the risk of bleeding events (HR, 2.44; 95%CI, 1.04-5.71; P=.041) (table 3 of the supplementary data). For the composite events, anticoagulant therapy was not associated with different rates of mortality, embolic events, and total bleeding in patients with dementia (adjusted HR, 0.83; 95%CI, 0.45-1.53; P=.541).

Figure 2.

Forest plot for the adjusted hazard ratios of anticoagulant therapy (vs no anticoagulation) for mortality, embolisms, and bleeding. 95%CI, 95% confidence interval; HR, hazard ratio; ISTH, International Society on Thrombosis and Haemostasis; PE, pulmonary embolism.

(0.18MB).

Of the 133 patients with dementia initially treated with anticoagulant therapy, 6 discontinued the anticoagulation during follow-up with no precipitating event and 4 discontinued the therapy after a bleeding event (figure 2 of the supplementary data). On the other hand, of the 88 patients with dementia not initially anticoagulated, 1 began anticoagulant therapy during follow-up with no precipitating embolic or bleeding event (figure 3 of the supplementary data). The analyses were repeated after the exclusion of the 11 patients with treatment changes during follow-up but similar results were obtained (adjusted HR for mortality, 1.09; 95%CI, 0.67-1.77; P=.731; adjusted HR for embolisms, 0.43; 95%CI, 0.21-0.95; P=.033; adjusted HR for bleeding, 2.35; 95%CI, 1.01-5.97; P=.047).

Finally, the analyses were complemented with propensity score matching, with patients with dementia paired according to anticoagulant therapy. Two groups of 26 matched patients were obtained based on the use or nonuse of anticoagulant therapy (table 2). In the patients with AF and dementia, anticoagulation was not associated with mortality (HR, 0.91; 95%CI, 0.45-1.83; P=.785) or the composite of mortality, embolisms, and bleeding (HR, 1.02; 95%CI, 0.52-2.00; P=.962).

Despite the interest generated by our results, the following limitations of our work must be considered. First, the retrospective design of the study, despite the meticulous collection of the data and their consistency, always allows for the possibility of errors during the process of categorizing patients into different groups. The retrospective design would be associated with, on the one hand, a risk of underestimating the true percentage of patients with dementia (possibly because a certain percentage of patients with dementia was not identified by their physicians with this diagnosis in their clinical history) and, on the other hand, an inability to accurately determine why the treating clinicians decided not to anticoagulate these patients. In addition, the sample size precludes subgroup analyses that would be of considerable interest. Despite the analysis of over 3500 patients, only 221 had moderate-to-severe dementia and only 133 of these were receiving oral anticoagulation. Although it is true that the event rate in the dementia patients was high (> 50% mortality and > 10% rate of embolisms and bleeding), the power of the study is suboptimal to perform analyses according to type of anticoagulant. Additionally, we did not determine whether the DOAC dose adjustment was performed correctly and in accordance with the data sheet, which is why our cohort might exhibit an underdosing related to the feeling of frailty and bleeding risk conveyed by these patients. Nonetheless, we believe our results to be highly valuable because they provide, for the first time, evidence on anticoagulant therapy in elderly patients with dementia and AF and lay the groundwork for future studies (ideally randomized) that may confirm our findings.