Background

Aging is a known risk factor for atrial fibrillation (AF), and the prevalence of AF increases with age, with more marked increases in people aged over 60 years [1,2,3]. The prevalence of diabetes mellitus (DM) also increases with age, particularly over the age of 65 years [4]. DM is a risk factor for developing stroke and systemic embolic events (SEE) in patients with AF [5], and it is among the factors used for clinical risk stratification for predicting stroke and thromboembolism in AF, namely the CHADS2 or CHA2DS2-VASc scores [6]. Among patients with AF, high proportions present with DM as a comorbidity; such patients tend to have worse outcomes and increased mortality compared with patients who do not have DM [7].

Glycated hemoglobin (HbA1c), an indicator of the average blood glucose level over the previous 2 to 3 months, is an indicator of glycemic control. It has been reported that glycemic control to HbA1c < 7.0% is important for suppressing the development and progression of microvascular complications [8, 9]. HbA1c targets for non-elderly patients with DM are HbA1c < 6.0% for patients thought to achieve glycemic control through diet and exercise therapy alone or those likely to achieve it while undergoing pharmacologic therapy without adverse reactions; HbA1c < 7.0% to prevent complications; HbA1c < 8.0% for those in whom intensive therapy is difficult because of the risk of hypoglycemia, among other reasons. For elderly patients with diabetes, however, the Japan Diabetes Society and Japan Geriatrics Society Joint Committee recommended in 2017 that patients be classified into three categories depending on the patient’s background characteristics and health status, such as age, cognitive function, and basic/instrumental activities of daily living, and that patients in each category be further divided into those receiving and those not receiving drugs potentially associated with severe hypoglycemia, such as insulin, sulfonylureas, and glinides [10]. Furthermore, the guideline specified the upper and lower limits of HbA1c to prevent diabetic complications and severe hypoglycemia. According to this guideline, the HbA1c target is < 8.5% for patients at risk of developing adverse reactions to multi-drug combination therapy or in those with serious comorbidities or poor social support [11].

Several studies have evaluated the incidence rates of adverse clinical outcomes in AF patients with and without DM, and the effects of oral anticoagulant (OAC) therapy on the incidence rates of outcomes [12,13,14]. The ENGAGE AF-TIMI 48 trial reported that, compared with patients with AF and without DM, patients with DM had a similar rate of stroke/SEE (adjusted hazard ratio [aHR] 1.08; P = 0.28) but a significantly higher risk of major bleeding (aHR 1.28; P < 0.001) after multivariate adjustments [12]. In the RELY trial, AF patients with DM had a stroke or SEE more frequently than patients without DM [13]. In the ROCKET-AF trial, an adjusted analysis suggested that AF patients with DM had 1.3-, 1.5-, and 1.9-fold higher 2-year rates of stroke, vascular mortality, and myocardial infarction, respectively, than AF patients without DM [14]. However, most AF patients in these studies were aged < 75 years and tended to be compared by the presence or absence of DM. A recent study investigated the association between HbA1c levels and the risk of stroke/SEE and major bleeding among patients with AF treated with or without OACs as well as the effectiveness and safety of warfarin vs. DOACs in a subgroup analysis stratified by different HbA1c levels [15]. The risk of events increased significantly with HbA1c levels above 6.5%. That study also found that, compared with warfarin, DOACs were more effective and safer in all HbA1c levels. The study indicated that glycemic control to achieve a more stringent target of HbA1c < 6.5% may be warranted in AF patients. However, it is important to note that patients in that study had a mean age between 68.9 and 71.2 years [15] and were younger than the target population of ANAFIE [14]. Thus, few studies have evaluated the incidence of main clinical outcome by HbA1c level and efficacy and safety by OAC status in elderly AF patients, particularly in those ≥ 75 years of age.

The All Nippon AF in the Elderly (ANAFIE) Registry is an observational prospective study conducted at multiple centers in Japan focusing on elderly patients (aged ≥ 75 years) with non-valvular AF (NVAF) to collect real-world data on their clinical status, anticoagulant therapy status, and prognosis with or without anticoagulation [14]. The baseline characteristics [16] and main outcomes at the 2-year follow-up have been reported [17]. The objective of this sub-analysis of the ANAFIE Registry was to investigate the association between HbA1c levels and outcomes among elderly Japanese patients with NVAF and to compare the risks of outcomes in patients treated with DOAC vs. warfarin for each HbA1c level.

Methods

Study design

The ANAFIE Registry was a large-scale multicenter, observational study conducted between October 2016 and January 2020 in Japan with the participation of 1273 medical facilities that enrolled elderly patients (≥ 75 years old) diagnosed with NVAF. The rationale and details of the study design, including disease classifications, diagnostic criteria, and assessments based on the AF guidelines of the Japanese Circulation Society and Japanese Heart Rhythm Society have been published previously [14, 18]. Data collection took place at baseline. Collected data included patient demographics, medical history including information on comorbidities (e.g., severe hepatic disease, hyperuricemia, heart failure and/or reduced left ventricular ejection fraction), occurrence of AF (i.e., type of AF, date and method of diagnosis, symptoms, and treatment decisions); history of catheter ablation, type of anticoagulant used; use of drugs other than anticoagulants (e.g., use of antiarrhythmic drugs, anti-platelet agents, proton pump inhibitors, and P-glycoprotein inhibitors); polypharmacy; and blood coagulation test under warfarin [14]. The patient follow-up was planned for a 2-year duration. As this was not an interventional study, the decision to prescribe OACs (e.g., warfarin or direct OACs [DOACs]) or not (no OACs), as well as oral or injectable antidiabetic drugs, depended entirely on the treating physician’s judgement.

Patients

The overall Registry included adult men and women diagnosed with NVAF based on electrocardiogram results who could attend hospital visits. This sub-analysis included patients with available data on HbA1c measurements at baseline. National Glycohemoglobin Standardization Program HbA1c values were used. Clinical diagnosis of DM by treating physicians was not an inclusion criterion for this sub-analysis.

The main exclusion criteria for the overall Registry were patients with NVAF who were participating in other studies; had undergone artificial valve replacement; had a recent history of a cardiovascular event such as stroke, myocardial infarction, cardiac intervention, heart failure, any bleeding leading to hospitalization within 1 month prior to enrollment, and life expectancy < 1 year; and deemed inappropriate for study participation by the investigator.

Study endpoints

The main endpoints evaluated were stroke/SEE, major bleeding, intracranial hemorrhage, cardiovascular death, all-cause death, and net clinical outcome (a composite endpoint of stroke/SEE, major bleeding, and all-cause death). Endpoints were stratified by HbA1c levels (HbA1c < 6.0%, 6.0% to < 7.0%, 7.0% to < 8.0%, and ≥ 8.0%). Subgroup analyses were also performed by type of OAC (warfarin and DOACs) for all patients and by type of antidiabetic drug (oral and injectable antidiabetic agents) for patients with a clinical diagnosis of DM.

Statistical analysis

Categorical variables were analyzed using frequency tables. Summary statistics were calculated for continuous variables and included n, mean, and standard deviation. Analysis of variance was used for comparison of continuous variables, and the chi-squared test, for categorical variables.

The probability of event occurrence during the 2-year follow-up period for each HbA1c level was estimated using the Kaplan–Meier method, and the corresponding P-values were calculated using the log-rank test. Incidence rates per 100 person-years with 95% confidence intervals (CIs) were also estimated. Hazard ratios (HR) were estimated using the Cox proportional hazards model adjusted by prognostic factors among the four HbA1c groups. The variables selected for this analysis were those possibly associated with the selection of the anticoagulant therapy or the incidence of outcomes [17]. A P-value < 0.05 was considered statistically significant. The statistical software used for analysis was SAS version 9.4 (SAS Institute, Tokyo, Japan).

Results

Patient disposition and characteristics

Of the 32,275 patients enrolled in the ANAFIE Registry, 17,526 patients with known HbA1c values at baseline were included in this analysis. The median (Q1–Q3) follow-up duration of this study was 2.0 (1.92–2.00) years. By HbA1c group at baseline, 8725 (49.8%) patients had HbA1c < 6.0%, 6700 (38.2%) had 6.0% to < 7.0%, 1548 (8.8%) had 7.0% to < 8.0%, and 553 (3.2%) had ≥ 8.0%.

In the overall ANAFIE population [16, 17], patients had a mean age of 81.5 years; 57.3% were male; mean CHA2DS2-VASc score, 4.5; mean HAS-BLED score,1.9; proportion of patients with paroxysmal AF, 42.1%; mean creatinine clearance (CrCl), 48.4 mL/min; and OAC use, 92.4% (DOACs, 66.9%; warfarin, 25.5%). Patients with HbA1c data at baseline (n = 17,526) had similar clinical characteristics, and 40.4% had a clinical diagnosis of diabetes mellitus (Table 1). Regarding the use of antidiabetic drugs, 24.6% were taking oral antidiabetic drugs, and 3.6% were taking injectable antidiabetic drugs (insulin [3.2%] and glucagon-like peptide-1 [GLP-1] receptor agonists [0.4%]). The most common oral antidiabetic drugs were dipeptidyl peptidase-4 (DPP-4) inhibitors (19.2%), sulfonylureas (6.1%), alpha-glucosidase inhibitors (4.9%), sodium glucose co-transporter 2 (SGLT2) inhibitors (2.1%), thiazolidinediones (1.9%), and others (5.0%) (Table 1).

Table 1 Characteristics of 17,526 patients at baseline by HbA1c levels
Full size table

Patients with HbA1c ≥ 8.0% were significantly more likely to have lower CrCl, higher CHA2DS2-VASc and HAS-BLED scores, and higher prevalence of non-paroxysmal AF, hypertension, dyslipidemia, chronic kidney disease, cardiovascular disease, thromboembolism-related disease, cerebrovascular disease, malignant tumor, dementia, falls within 1 year, and lower administration of DOACs, but higher administration of warfarin compared with the other subgroups. Regarding oral antidiabetic drug use, patients with HbA1c 7.0% to < 8.0% and those with HbA1c ≥ 8.0% had higher rates for sulfonylureas, alpha-glucosidase inhibitors, DPP-4 inhibitors, SGLT2 inhibitors, and others, as well as higher rates for injectable antidiabetic drugs, including insulin and GLP-1 receptor agonists, compared with the other subgroups (Table 1).

Study endpoints

Figure 1 shows the Kaplan–Meier estimates of the probability of occurrences of each event. The probabilities of event occurrences were significantly different among HbA1c categories for all-cause death (log-rank P = 0.003) and net clinical outcome (log-rank P = 0.007), and both event occurrences were visually higher in patients with HbA1c ≥ 8.0%. A similar trend was observed for stroke/SEE in the HbA1c ≥ 8.0% subgroup, but the difference did not reach statistical significance.

Fig. 1
figure 1

Kaplan–Meier curves for clinical outcomes by HbA1c levels

HbA1c glycated hemoglobin, SEE systemic embolic events

Full size image

Table 2 summarizes the incidence rates of events by HbA1c level subgroup. Although the incidence of events was comparable among groups with HbA1c < 8.0, a remarkable increase was observed in the HbA1c ≥ 8.0% subgroup, particularly for stroke/SEE (2.50/100 person-years [95% CI: 1.52–3.49]), all-cause death (5.69/100 person-years [95% CI: 4.23–7.16]), and net clinical outcome (7.83/100 person-years [95% CI: 6.09–9.57]).

Table 2 Incidence rates of stroke/SEE, major bleeding, ICH, cardiovascular death, all-cause death, and net clinical outcome of 17,526 patients
Full size table

Figure 2 shows the results of multivariate analysis using the Cox proportional hazards model by HbA1c, with HbA1c < 6.0% as reference. The risk (aHR [95% CI]) of stroke/SEE (1.48 [0.97–2.25]) was numerically high at HbA1c ≥ 8.0%. The risks of all-cause death (1.46 [1.11–1.93]) and net clinical outcome (1.33 [1.05–1.68]) were significantly higher in the HbA1c ≥ 8.0% subgroup, while there were no significant differences in the risks of major bleeding (0.94 [0.52–1.70]) or other outcomes. No significant difference was observed in the risk of any event, with HbA1c levels ranging between 6.0% to < 7.0% and 7.0% to < 8.0%. Supplementary Table 1 shows the multivariate analysis using a Cox proportional hazards model in patients with AF and diagnosed with diabetes by each participating physician’s judgement by HbA1c, with HbA1c 6.0% to < 7.0% as reference. In patients with HbA1c ≥ 8.0%, significantly higher risk was observed for all-cause death (1.47 [1.10–1.97]).

Fig. 2
figure 2

Multivariate analysis using Cox proportional hazards models by HbA1c levels

HbA1c < 6.0% was the reference; bars represent 95% confidence interval

HbA1c glycated hemoglobin, HR hazard ratio, SEE systemic embolic events

Full size image

Table 3 shows the multivariate analysis using the Cox proportional hazards model by HbA1c, no OAC, and OAC treatment. Among patients with HbA1c 6.0% to < 7.0%, DOAC use was associated with significantly lower risk (aHR [95% CI]) of stroke/SEE (0.71 [0.53–0.95]), major bleeding (0.56 [0.39–0.82]), intracranial hemorrhage (0.55 [0.36–0.85]), and net clinical outcome (0.80 [0.68–0.95]) compared with warfarin. For patients with HbA1c ≥ 8.0%, there was a higher risk of cardiovascular death (5.52 [1.03–29.62]) and net clinical outcome (1.69 [0.96–2.97]) with DOAC use. However, no interaction was observed between HbA1c level and anticoagulants. In Supplementary Tables 2, we show the number (%) of events by HbA1c level and anticoagulant treatment (i.e., warfarin, no OAC, or DOAC).

Table 3 Multivariate analysis using Cox proportional hazards models with DOACs (n = 11,921) and No OAC (n = 1262) versus warfarin (n = 4336) by HbA1c levels
Full size table

Table 4 shows the multivariate analysis using Cox proportional hazards models by HbA1c and oral and injectable DM medication in patients with DM. Overall, using the subgroup with no oral antidiabetic drug use as a reference, oral antidiabetic drug use was not associated with risk reduction. Similar results were observed for patients receiving injectable antidiabetic drugs. No significant difference was observed in event risk with versus without injectable antidiabetic drugs.

Table 4 Multivariate analysis of 8733 diabetes patients using Cox proportional hazards models with oral and injectable antidiabetic medication
Full size table

Discussion

This is the first study to evaluate the incidence of the primary outcomes by HbA1c level and the efficacy and safety of OAC in elderly NVAF patients aged ≥ 75 years. The main findings of this sub-analysis were as follows. First, patients in the HbA1c ≥ 8.0% subgroup had significantly higher probabilities of event occurrence for all-cause death and net clinical outcome. A similar trend was observed for stroke/SEE, but the differences did not reach statistical significance. Multivariate analysis of HbA1c (< 6.0% as reference) did not show significant differences in the risk of any event with HbA1c levels ranging between 6.0% to < 7.0% and those between 7.0% and < 8.0%; however, HbA1c ≥ 8.0% was significantly associated with increased risk of all-cause death and net clinical outcome. Although recent data from Japan [19, 20] indicate that diabetes was not a risk factor for stroke in AF patients, our results suggest that event risk is indeed influenced by blood glucose levels. Second, multivariate analysis using the Cox proportional hazards model showed a lower event risk of cardiovascular death and net clinical outcome for DOAC use (compared with warfarin) among patients in the HbA1c 6.0% to < 7.0% subcategory but a higher risk of these events among patients with HbA1c ≥ 8.0%. Nevertheless, there was no significant interaction between HbA1c and OAC medication. The lack of significant interaction may be attributable to the extremely low number of events among patients with HbA1c ≥ 8.0%, and there is a high possibility that the observed data were due to chance. Further, this result is comparable with that shown in the analysis of each event by OAC treatment for the overall population, in which the risks of stroke/SEE, all-cause death, and net clinical outcomes were lower in the DOAC group vs. warfarin [17]. Third, regarding the risk of each event with oral or injectable antidiabetic drug use, neither oral nor injectable antidiabetic drugs were associated with risk reduction. Finally, the present findings show that the HbA1c threshold, which is the target of diabetes treatment, is appropriate even for elderly patients with NVAF, which has important implications for clinical practice.

Previous studies have reported that the risk of ischemic stroke in AF patients with DM is more closely related to the disease duration of DM than to glycemic control based on HbA1c [21]. In contrast, in this sub-analysis, HbA1c ≥ 8.0% compared with HbA1c < 6.0% had a numerically increased risk of stroke/SEE, which aligns with findings in a previous study [22]. Of note, we did not assess the effect of disease duration on evaluated clinical outcomes. The present results are also consistent with a previous report in which the incidence of all-cause death increased with high HbA1c levels [23]. According to the current Japanese Clinical Practice Guidelines for Diabetes, the HbA1c target for glycemic control among elderly patients with DM aged ≥ 65 years is < 8.0% if intensification of treatment is difficult [10]. The results of this sub-analysis align with the recommended HbA1c target and suggest that this cutoff is also applicable for elderly patients with DM who have concomitant NVAF. Furthermore, many patients with HbA1c ≥ 8.0% were being treated with warfarin rather than DOACs. The reason why these patients were treated preferentially with warfarin may be attributable to the fact that many patients had decreased renal function. Per current treatment guidelines, vitamin K antagonists remain the first-line treatment for preventing thromboembolic events in patients with chronic kidney disease who require anticoagulation [24].

The main limitations of the ANAFIE Registry are primarily associated with the observational design and have been previously reported [16, 17]. Although the overall ANAFIE Registry population comprised over 32,000 elderly patients with NVAF, approximately 15,000 patients did not have data on HbA1c measurements at baseline; however, no significant difference was observed in patient characteristics and probability of event occurrence between those with and those without HbA1c measurement. As this sub-analysis evaluated HbA1c at baseline, we could not assess changes in HbA1c levels over time. Patients with high HbA1c at baseline were included in this sub-analysis, but in some cases, HbA1c may have been measured even though the patient had not been clinically diagnosed with DM. The disease duration of DM was not considered in the present analysis. Rates of use of DPP-4 and SGLT2 inhibitors were low in Japan when the study was conducted. However, these antidiabetic drugs are now widely used, which limits extrapolation of these results to current clinical practice. Finally, HbA1c ≥ 8.0% was evaluated in 553 cases, and this limited number of cases, along with the duration of the observation period, may have precluded an accurate evaluation of the incidence of events.

Conclusions

Among elderly Japanese patients with NVAF, HbA1c ≥ 8.0% was associated with an increased risk of all-cause death, but the risk of the events did not increase with HbA1c levels of 6.0% to < 8.0% as compared with < 6.0%. Furthermore, there was no significant interaction between HbA1c and OAC medication. Neither oral nor injectable antidiabetic drugs were associated with clinical outcome risk reduction. Individualized treatment goals and strategies based on comorbidities and existing treatments—considering polypharmacy and drug–drug interactions—along with close monitoring and strict control of DM, are crucial for preventing such outcomes and for improving prognosis in this population.