Background
Diabetes increases the risk of atrial fibrillation (AF) and is associated with increased symptom burden, lower quality of life, and increased hospitalization and mortality rates [
1]. Among AF patients with diabetes, both thromboembolic and haemorrhagic events were significantly increased [
2], indicating that diabetes is an important factor in CHA2DS
2-VASc bleeding risk scores that are commonly used in patients with AF [
3]. Four large phase III clinical trials (Dabigatran etexilate RE-LY [
4], Rivaroxaban ROCKET AF [
5], Apixaban ARISTOTLE [
6], and Edoxaban ENGAGE AF- TIMI 48 [
7]) have compared the effectiveness and safety of new direct oral anticoagulants (NOACs) with warfarin, and the results showed noninferiority for safety and efficacy [
8]. A previous meta-analysis of the four NOAC trials also found no significant interaction between treatment and diabetes status for stroke/systemic embolism (SSE) or major bleeding [
9,
10]. Therefore, current international guidelines recommend the use of NOACs as effective, safer and more convenient alternatives to warfarin among patients with NVAF with diabetes [
11,
12].
However, in post hoc analyses of the RE-LY study [
15] and ARISTOTLE study [
18], diabetes and treatment had a significant interaction for the risk of major bleeding, and the risk of major bleeding for dabigatran 110 mg or apixaban over warfarin was diminished in AF patients comorbid with diabetes. This difference may be attributed to differences in data at baseline (patient age, NOAC dose, patient's underlying cardiovascular disease, the specific definition of major bleeding in the trial, the mean CHADS2 score and the varying degrees of renal metabolism) in diabetic and nondiabetic patients.
Data from real-world studies have shown that NOACs are more effective than warfarin in terms of treatment. However, results regarding the risk of major bleeding have varied widely [
13,
16,
17,
19,
20]: some studies have shown that NOACs are better than warfarin in terms of the risk of major bleeding [
16,
19,
20], while other studies have reported no difference [
13,
17]. To further explore the effectiveness and safety of NOACs in the treatment of patients with NVAF and diabetes, we conducted this systematic review and meta-analysis based on the results of subgroup analyses of randomized controlled trials (RCTs) and real-world data.
Methods
We conducted a systematic review and meta-analysis in accordance with the Meta-analyses of Observational Studies in Epidemiology guidelines [
22]. Our study is registered with PROSPERO (URL:
https://www.crd.york.ac.uk/prospero/. Unique identifier: CRD42020192098). Note: When registering for the study, we originally planned to use odds ratio (OR) as an outcome indicator. However, most of the studies we included used hazard ratio (HR) as an outcome indicator. Therefore, we used HR as an outcome indicator to conduct meta-analysis and report the results.
Literature search strategy
We searched the PubMed, Embase, Cochrane Library, and Web of Science databases from inception through June 2020. The ClinicalTrials.gov databases was also searched for ongoing and unpublished studies. No language restriction was applied. The reference lists of the related studies, reviews and meta-analyses were also examined. The search terms were [“apixaban” or “dabigatran” or “rivaroxaban” or “edoxaban” or “new oral anticoagulants” or “direct oral anticoagulants” or “DOACs” “non-vitamin K oral anticoagulants” or “NOACs”] AND [“Atrial Fibrillation” or “Auricular Fibrillation”] AND [“diabetes mellitus” or “diabetes” or “hyperglycemia”].
Inclusion and exclusion criteria
The inclusion criteria for the present meta-analysis were as follows: (1) prospective or retrospective cohort studies that started with the recruitment of patients with NVAF in the setting of diabetes who received either NOACs or dose-adjusted warfarin; (2) subgroup analyses of RCTs that compared the risk of efficacy and safety of any NOACs with dose-adjusted warfarin by diabetes status; and (3) investigated NOACs include apixaban, dabigatran, edoxaban, and rivaroxaban. The exclusion criteria were as follows: (1) studies that lack corresponding outcome indicators; (2) duplicate results from the same population; and (3) studies without relevant data after contacting the original author.
Outcomes
The primary efficacy outcome was the SSE composite measure, and the safety outcome was major bleeding. The secondary efficacy outcomes included ischaemic stroke and haemorrhagic stroke, and the secondary safety outcomes included intracranial bleeding, gastrointestinal bleeding, myocardial infarction and all-cause mortality.
Data relevant to this study were independently extracted from the screened literature by two reviewers (JXD and YZ) using a data collection sheet in accordance with the recommendations from the Cochrane handbook for systematic reviews of interventions. The following data were extracted: study design, publication year, number of test and control groups, age of test subjects and dose of test drugs, CHADS2 score, baseline characteristics of participants, methods used to identify and verify the diagnosis of NVAF and diabetes. Disagreements were resolved by referring back to the original articles and consensus with a third member of our team (ZW).
Risk of bias
The inclusion and data extraction of all studies were performed independently by two researchers (JXD and YZ) according to the corresponding criteria, and the original literature or data were checked by a third researcher (ZW) for inconsistencies. The study of subgroup analyses of RCTs was the same as a cross-sectional study, and we assessed the quality of these analyses using the Agency for Health care Research and Quality (AHRQ) quality indicators [
23]. The quality of cohort studies was assessed using the Newcastle‒Ottawa Scale (NOS) [
24]. Both tools were applied independently by two review authors (JXD and YZ). If the raters disagreed, a third review author (ZW) was consulted.
Data synthesis and statistical analysis
Meta-analysis was performed using Stata 15.1 SE software (StataCorp, 2017). The generic inverse variance method was used, and hazard ratios (HRs) and 95% confidence intervals (CIs) were used to describe the outcomes. Cochran’s Q test was used for statistical heterogeneity. A value of I2 < 50% and P > 0.10 represents low heterogeneity, and in such cases, a fixed effects model was used for meta-analysis; in cases of high heterogeneity, the random effects model was used for meta-analysis. Additionally, subgroup analyses were conducted for both the primary efficacy and safety outcome to explore the heterogeneity among treatment effects. This was carried out based on the possible sources of heterogeneity, including drug type, drug dose and complicating disease of the patient.
Discussion
Diabetes is an independent risk factor for stroke in patients with NVAF. Studies have shown that the risk of stroke in patients with NVAF combined with diabetes is increased by approximately 70% [
2]. Therefore, prevention of stroke is the key to patients with NVAF and diabetes, anticoagulation therapy is the core measure for the prevention of stroke, and it can significantly reduce the risk of stroke and the mortality of patients with NVAF [
25]. Warfarin was the cornerstone of oral anticoagulant therapy before the launch of NOACs. Since rivaroxaban was approved for market use in 2010, NOACs have been developed rapidly. Compared with warfarin, these drugs have many advantages, such as more predictable pharmacodynamics, fewer drug and food interactions and the lack of need for routine laboratory monitoring. Therefore, the 2018 European Heart Rhythm Association Room Fibrillation anticoagulation guidelines refer to the recommendation of NOACs as the first choice for stroke prevention in patients with NVAF [
26]. Previous studies have shown that when warfarin is used for patients with NVAF with diabetes, it may be more difficult to achieve anticoagulation standards, and the compliance rate is low, which further increases the risk of anticoagulation failure [
27]; thus, NOACs have a superior application advantage in patients with NVAF and diabetes. The results of our study further prove the effectiveness and safety of NOACs and can provide important evidence-based guidance for clinical applications.
In this study, a meta-analysis of the data of 26,7272 patients showed that for patients with NVAF and diabetes, NOACs can significantly reduce the incidence of SSE, ischaemic stroke, and haemorrhagic stroke, intracranial bleeding, gastrointestinal bleeding, myocardial infarction, and vascular death compared to warfarin. However, only apixaban had a lower risk of major bleeding than warfarin. Dabigatran, rivaroxaban and edoxaban had a similar risk of major bleeding to warfarin. The all-cause mortality of NOACs also did not show an advantage compared with warfarin.
A combined analysis of the results of the RE-LY [
4], ROCKET AF [
5], ARISTOTLE [
6], and ENGAGE AF- TIMI 48[
7] trials showed that the comprehensive risk of SSE in diabetic patients treated with NOACs was 3.16% (9096 patients received NOCAs) and 3.96% among patients treated with warfarin (8990 patients treated with warfarin) (RR = 0.80; 95% CI 0.69, 0.93) [
28], indicating that NOACs have a slight advantage over warfarin. Several previous studies [
10,
29,
30] also showed that the use of NOACs and vitamin K antagonists (VKA) in patients with NVAF has a similar risk of SSE and major bleeding in diabetes (RR = 0.97 95% CI 0.79, 1.18)) and nondiabetes (RR = 0.76 95% CI 0.65, 0.88) patients. Our study is consistent with the above results. For patients with NVAF and diabetes, the use of NOACs can reduce the risk of SSE without increasing the incidence of major bleeding.
The meta-analysis results of Ruff et al. [
29] showed that the use of NOACs in patients with NVAF can significantly reduce all-cause mortality (RR = 0.90, 95% CI 0.85, 0.95; p = 0.0003) but increased gastrointestinal bleeding (RR = 1.25, 95% CI 1.01, 1.55; p = 0.04). However, the results of our study showed that NOACs did not increase the incidence of gastrointestinal bleeding and had no difference in all-cause mortality compared with warfarin. The cause might be that the previous studies were all RCTs, the inclusion criteria of patients were relatively strict, and in our study, the population included was less restrictive and more representative of the real-world population. A study by Patti et al. [
10] showed that NOACs did not have advantages for the occurrence of ischaemic stroke and intracranial bleeding compared to warfarin, while the results of our study showed that the use of NOACs was superior to warfarin in terms of the incidence of ischaemic stroke and intracranial bleeding. This difference may be related to the small sample size of the previous study and the large sample size of our study.
The results of our study show that for patients with CKD, the safety of NOACs is higher than that of warfarin, and the risk of major bleeding is lower. However, the results of the study should be interpreted carefully because NOACs have strict limitations on the renal function of patients. For patients with severe renal insufficiency, the safety and effectiveness of NOACs are relatively lacking, so in real-world studies, patients who take NOACs may have better renal function status than those who take warfarin. The warfarin group may include more patients with end-stage renal disease or renal failure, so the effectiveness and safety of the two types of drugs used by patients with CKD is insufficient. A previous meta-analysis showed that NOACs did not differ from warfarin in reducing SSE (RR = 0.81, 95% CI 0.65, 1.00) or major bleeding (RR = 0.79, 95% CI 0.59, 1.04) [
31]. The newly published retrospective study based on a database system [
32] included 21,733 patients with NVAF with different CKD levels. The results showed that compared with warfarin, NOAC use in patients with impaired renal function was associated with a lower risk of mortality and major bleeding that required hospitalization in patients with all kidney function levels (eGFR > 60%, eGFR > 30–60% and eGFR ≤ 30% or on dialysis). NOACs seem to show clinical advantages in people with renal insufficiency. However, eGFR ≤ 30% or on dialysis patients accounted for only 7.0% in this study, and there was no analysis of CKD stage 5 patients with eGFR ≤ 15%. There is currently no evidence of the use of NOACs in CKD stage 5 patients with an eGFR ≤ 15%, so it is necessary to strengthen the monitoring of CKD stage 5 patients, and more research evidence is needed to support the effectiveness and safety of NOACs in CKD patients.
Medication compliance is an important factor affecting drug efficacy and safety. The results of the meta-analysis on NOAC medication compliance showed that the overall compliance with NOACs was significantly higher than that with vitamin K antagonists (OR = 1.44; 95% CI 1.12–0.86]. Additionally, NOAC nonadherence was associated with an increased risk of stroke (HR = 1.39; 95% CI 1.06–1.81) [
33]. Although NOACs improved compliance compared with warfarin and have certain advantages in clinical application, the current study showed that the overall population was more likely to have high medication compliance, so it is still necessary for medical staff to strengthen the education of medication patients in the future, improve patients’ awareness of compliance, and ensure the effective and safe application of drugs in clinical practice.
Although we performed a systematic search and detailed analysis, this study has some limitations. (1) All the included studies were not RCTs, and there may be an imbalance in the inclusion of subjects. The cohort studies included in the analysis were not prospectively designed, and the research methods were not uniform. Some studies used propensity score matching (PSM) to group subjects, and some used natural grouping, thereby introducing potential bias in the analysis. (2) The study did not analyse the patient’s diabetes type, blood sugar control status, or the impact of the current hypoglycaemic program on the results. These factors may have a great impact on the results. (3) Finally, a study from China [
34] was not included in the analysis even though it met our inclusion criteria because the outcome of the study was not reported using HRs; thus, we could not combined the data from that study with the data from the included studies. The exclusion of the study from China may have introduced some study bias. Therefore, we still need to be cautious when interpreting the evidence of this study. More large-scale, multicentre, random, double-blind experiments are needed to provide additional evidence.
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.