Epidemiology and treatment of atrial fibrillation in patients with type 2 diabetes in the UK, 2001–2016

Patients with Type 2 diabetes mellitus (T2DM) have an increased risk of atrial fibrillation (AF). The current study aimed to investigate the prevalence and treatment of AF in patients with T2DM, assess the impact of direct oral anticoagulants (DOACs) introduction on oral anticoagulant (OACs) prescribing rates, and factors associated with OAC initiations in patients with T2DM and AF. The Health Improvement Network (THIN) database (2001–2016), was used to examine the annual prevalence and treatment of AF in T2DM. The impact of DOACs introduction on OAC prescribing rates were investigated using interrupted time series analysis (ITS). Factors associated with OAC initiations were also identified using multivariate logistic regression. The prevalence of AF increased from 2.7 [95% confidence intervals (CI) 2.5–2.8] in 2001 to 5.0 (4.9–5.1) in 2016 per 100 persons. OACs prescribing within 30-days of AF diagnosis increased from 21.5% in 2001 to 56.8% in 2016. ITS analysis showed that OAC prescribing increased after DOAC introduction (P < 0.001), however, no immediate change was observed (P = 0.29). T2DM patients with AF, aged 60–79, male gender and BMI ≥ 25 were more likely to receive OAC, adjusted OR 1.3 (1.2–1.5) for aged 60–79, 1.3 (1.2–1.4) for male gender and 2.0 (1.9–2.2) for BMI ≥ 25, respectively. This study highlighted an increase in prevalence of AF in patients with T2DM during the study period. Further studies are warranted to investigate factors contributing to the underuse of OAC in patients with T2DM and AF.

Scientific RepoRtS | (2020) 10:12468 | https://doi.org/10.1038/s41598-020-69492-z www.nature.com/scientificreports/ Previous studies that examined the prevalence and treatment of AF among patients with T2DM are limited. Estimating the burden of AF in patients with T2DM across the UK population will help to develop a better understanding of the co-existence of both conditions, their treatment and explore population levels trend in order to plan health policy.
The objectives of this study were (i) to examine the trends of the prevalence of AF in patients with T2DM from 2001 to 2016; (ii) to investigate the proportions of patients with T2DM who were initiated oral anticoagulants (OAC) on/or after AF diagnosis, and to assess the impact on OAC prescribing rates after the introduction of direct oral anticoagulants (DOACs), and (iii) to investigate factors associated with the initiation of OAC in patients with T2DM and AF.

Methods
Data sources. This retrospective population-based longitudinal study used data in the Health Improvement Network (THIN). THIN is a UK primary care database containing anonymized, clinical and prescribing data with more than 15 million cumulative patients, covering approximately 6.0% of the UK population 9 . THIN database is widely used healthcare database for the population-based medical research and has previously been used to study prescribing of OAC medications 10-14 . Ethical considerations. The present study is based on anonymised and unidentifiable THIN data, thus the need for informed consent was waived by the THIN scientific review committee (SRC). This study was reviewed and scientific approved by THIN SRC in 2018 (18THIN009). The research was reported in accordance with strengthening the reporting of observational studies in epidemiology (STROBE) Statement (Supplement).

Study population.
Patients with T2DM aged ≥ 18 years old and registered within THIN database between 2001 and 2016 were included in the study. Only patients who were registered with the general practice for at least 12 months prior to the first T2DM diagnosis being recorded were included. They were identified based on the Read Codes of (1) a diagnostic code for T2DM or (2) a non-specific code of diabetes but had a record of any prescribed oral hypoglycaemic agent. Patients who had a diagnostic code for T2DM accounted for 92.7% of the entire cohort, while the remaining were of criteria two. Patients who had a non-specific code of diabetes but had only records of insulin prescriptions were excluded because they may have type 1 diabetes mellitus (T1DM), although their age at first event was taken into account.
Prevalence of AF in T2DM. T2DM patients who had a record of AF were identified on/or after their diagnosis of T2DM using the AF read codes, and the first record of AF was defined as the start date. Patients diagnosed with AF before the diagnosis of T2DM were excluded. Patients with valvular heart disease were excluded. Patients were censored if they left the practices, transferred out or died during the study period.
OAC use in patients with T2DM and AF. Patients with T2DM and AF who received at least one prescription of an OAC were identified using drug codes. Patients with T2DM who received an OAC prescription on/ or after the diagnosis of AF were included in the treatment analysis. Patients who received an OAC prescription prior to the diagnosis of AF were excluded from the treatment analysis. Patients were divided into two groups; one group received an OAC prescription and a second group who did not receive an OAC prescription. Further stratification by type of OAC into warfarin and DOACs (dabigatran, apixaban, rivaroxaban, and edoxaban) were also undertaken. Selection of study cohort is presented in Fig. 1.
Factors associated with OAC use. Patients characteristics included in the study were age, gender, smoking status, alcohol consumption, body mass index (BMI), coronary heart disease (CHD), heart failure, hypertension, hyperlipidaemia, chronic kidney disease (CKD), bleeding and liver diseases. These comorbidities were identified using Read codes 15,16 . Medications; antiplatelet drugs, antihypertensive drugs, and lipid-lowering drugs. In addition, we included polypharmacy as a covariate in the study and it was defined as the use of ≥ 4 chronic cardiac medications 17 . CHA 2 DS 2 -VASc score for stoke risk and HASBLED score for risk of bleeding were also calculated 6 .
Outcomes. The primary outcome was the annual prevalence of AF in patients with T2DM from 2001 to 2016. Secondary outcomes were: the proportions of patients with T2DM who were initiated an OAC on/or after AF diagnosis from 2001 to 2016; the effect of the introduction of DOACs on the rate of OACs initiations in patients with T2DM and AF; and factors associated with OAC initiation in patients with T2DM and AF. Data analysis. Descriptive statistics were used to describe patients' demographics, medications use and comorbidities. Continuous data were reported as mean ± standard deviation (SD), and categorical data were reported as number (percentage). The prevalence of AF in patients with T2DM was presented per 100 person with 95% confidence intervals (CIs). This was calculated annually by dividing the number of all T2DM patients diagnosed with AF during the particular year over the mid-year population of patients with T2DM in the same calendar year during the study period. Trends in the prevalence of AF were further stratified by age and gender. Temporal trends in the distribution of the prevalence were assessed using a Poisson method. The annual proportions of patients with T2DM who initiated an OAC (PPIOAC) on/or after AF diagnosis from 2001 to 2016 was calculated.
The impact of the introduction of DOACs on the rate of OAC initiation was plotted graphically over time. In addition, we fitted a segmented regression analysis using a Poisson regression 18 . Durbin-Watson test was used to examine any first order autocorrelation that may lead to an overestimations of the significance of an intervention. Residual analyses were conducted, and showed no evidence of autocorrelations. Overall, we included 44 data points (monthly quarters); representing repeated OAC prescriptions from July-October 2005 up to April-July 2016. DOACs were first authorized for the treatment of non-valvular AF in 2011 19 , therefore, we accounted for the intervention in this model from the first quarter of the next calendar year (January-April 2012). We used multivariable logistic regression to identify factors associated with the initiation of OACs prescribing in patients with T2DM and AF compared with no OAC prescribing, and stratified by OAC type (warfarin Vs. DOACs). Unadjusted and adjusted odds ratios (OR) with 95% confidence intervals (CIs) were estimated for all the aforementioned baseline covariates. All analyses were performed using SAS version 9.4 (SAS Institute, Cary, NC, USA).

Results
Demographics and characteristics. During the study period of 2001 and 2016, a total of 405,718 patients with T2DM were identified of whom only 23,124 patients with T2DM and AF were included. Around 12,124 (52.4%) received an OAC prescription at some point on/or after the diagnosis of AF (Fig. 1). The characteristics of patients are summarised in (Table 1).

PPIOAC =
Number of patients with T2DM who received OAC on/or after the diagnosis of AF in a particular year Number of patients with T2DM and were diagnosed with AF in that particualr year   (Fig. 2). Similarly increased trends for both men and women were observed for the first two years and then men started to have a higher increase rate over the study period than women (Fig. 2). The prevalence of AF varied among the different age groups. The prevalence of AF among patients aged 75 years and above increased from 5.5% (95% CI 5.1-5.8) in 2001 to 9.9% (95% CI 9.7-10.0) in 2016 per 100 persons with T2DM, p < 0.001. There was about 43-55% increase in AF prevalence among younger patients from 3.0% (95% CI 2.7-3.

OAC treatment at various time points after AF diagnosis. The proportions of patients with T2DM
who received an OAC prescription within 30-days of AF diagnosis increased from 21.5% in 2001 to 56.8% in 2016, p < 0.001. In sensitivity analysis, the proportions of patients with T2DM who received an OAC prescription within 90-days of AF diagnosis was higher, 29.8% in 2001 to 69.9% in 2016, p < 0.001. In addition, the proportions of patients with T2DM who received an OAC prescription within 1-year after the diagnosis of AF was markedly higher in comparison to 30-days and 90-days from diagnosis, ranging from 39.4% in 2001 to 78.0% 2016, p < 0.001 (Fig. 4).    patients with T2DM who received OAC prescription on/or after 30-days of AF is presented in (Fig. 5). There was no immediate change in the rate of OAC prescribing after the introduction of DOACs (p = 0.29). However, the rate of OAC initiation then increased gradually (p < 0.001) (Table S1). 2), respectively). In addition, the use of ACEI/ARB, BB, CCBs and statins was a strong predictor to initiate OAC, while use of aspirin and polypharmacy were protective factors against the initiation of OAC. Table 2 presents details of the results from a logistic regression model.

Discussion
In this population-based study, we investigated trend in the prevalence and treatment of AF in patients with T2DM over a 16-year period. The key findings were: 1) the prevalence of AF in patients with T2DM has increased from 2001 to 2016, 2) the proportion of patients with T2DM who were initiated on an OAC after AF diagnosis increased between 2001 and 2016, 3) the rate of OAC initiation after the introduction of DOACs into the market increased, and 4) our study demonstrated that age ≥ 75 years, previous bleeding or stroke/TIA and history of CKD, were strong predictors for DOACs initiation. Previous studies reporting the prevalence of AF in patients with T2DM are lacking, A study by Adderley et al., using a national UK database, reported that the prevalence of AF in the UK general population increased from 2.0% in 2000 to 3.2% in 2016 20 . The authors reported that the prevalence of AF was higher among those aged 65 years and above and was higher among male patients which was supported by our study findings in the T2DM patients. Our results showed a higher prevalence trend among male patients and among those aged 65 years and above which was similar to their results. In addition, the increase of AF prevalence in patients with T2DM over the years could also be related to an evolved physician's sensibility and consequent more aggressive search for AF 21 .
Ageing is an important risk factor for AF and the prevalence of AF increases with age, in the Framingham study it was reported that the prevalence of AF increased by 0.5% for those aged 50-59 years compared to 8.8% for those aged 80-89 years 22 . Furthermore, in a European community based studies it was reported that cumulative incidence of AF increased markedly after the age of 50 for men and 60 for women 23 . Our study showed a www.nature.com/scientificreports/ higher trend in the prevalence of AF in males and patients aged 75 years and above compared to females and patients younger than 75 years, which was similar to published data for the general population 22,23 . Previous studies reported the association of T2DM and AF; however, the mechanism of the development of AF in patients with T2DM is not fully understood. It has been suggested that the metabolic process in patients with T2DM, including the inflammatory response and the atrial remodelling, might play a major role in the association between both diseases 24,25 . In addition, patients with T2DM have a high cardiac risk-profile and a higher body mass index. These are known risk factors for AF, which was highlighted in our study 26,27 .
T2DM and AF are both highly prevalent in the general population with about 6.0%-7.0% of the population having diabetes 1 and about 1.5%-2.9% of the population have AF 2,20 . T2DM and AF have also been linked to several comorbidities and increased risk of stroke and mortality. It is therefore important to recognise the coexistence of both conditions to increase the awareness and to closely monitor this population. Several guidelines rely on the CHA2DS2-VASc score, in which T2DM is a criterion for score calculation and have recommended the use of OACs in patients with atrial fibrillation in order to prevent future stoke events 6,7 .
Although, our study demonstrated that the rate of OAC initiation has increased over time, our study also highlighted the possible underuse of OAC in this population. Particularly if we take into consideration that the majority of the patients were eligible for anticoagulation, based on the CHA 2 DS 2 -VASc score, as shown in Table 1 6 . There were 44.0% of our study patients in 2016 who still did not receive OAC within 30-days of AF diagnosis. It is important to mention there are other factors that physicians might consider before prescribing an OAC to their patients including risk of bleeding 6,7,28 .
In this study, we found that the rate of OAC initiation after the introduction of DOACs increased significantly, however, this change was not immediate. This could be explained because new drugs are prescribed with a greater caution due to uncertainties in regards to their effectiveness and safety 29 . In addition, prescribing patterns are likely to be influenced by other factors including, updates in guidelines recommendation. This was highlighted by Komen et al., who reported that the update in the European Guidelines was associated with an increase DOACs initiations 30 .  www.nature.com/scientificreports/ Our analysis also identified some of the individual-level characteristics that may influence the overall and the type of OAC prescribing. BMI ≥ 25 and male gender were strong predictors for the initiating of OAC. These results were also in line with a previous large observational study where the authors reported that both BMI ≥ 25 and male gender are likely to influence the OAC prescribing 31,32 . Other predictors including; the use of ACEI/ ARB, BB, CCBs and statins were also associated with the initiation of OAC prescribing. This could be explained by the fact that these medications are commonly indicated for the management of cardiac diseases, where hypertension, peripheral vascular diseases, stroke and congestive heart failures are all criteria in CHA2DS2-VASc score calculations 33 . However, our results demonstrated that the use of aspirin was negatively associated with OAC prescribing. Aspirin is one of the criteria in HASBLED score 34 , in which it is given a total of 1 point in the total score which predicts the risk of bleeding and therefore, it reasonable to assume that patients who use aspirin are less likely to receive OAC. In addition, we found that both age ≥ 75 years and having a history of previous bleeding were significant predictors of DOAC prescribing. Several randomized trials studies have shown safer and non-inferiority of DOACs use in patients with AF [35][36][37][38] . In addition, recent observational studies have demonstrated a safer profile of DOACs compared to warfarin 39 , and less bleeding events among patients with AF ≥ 90 years of age 40 . Furthermore, having a history of CKD was associated with more likelihood of receiving DOACs. This finding was in line with some evidence-based literature, as DOACs showed favourable safety and efficacy profile in patients with CKD 41 .

Strengths and limitations.
To the best of our knowledge, this was the first study that examined the prevalence and treatment of AF in T2DM over a 16-years period. This study used a primary care database, which is representative of the UK general population, however, there are some limitations in our study. Firstly, THIN only provides information of primary care setting, and therefore, underestimation of the prevalence and treatment of AF in T2DM would be possible as THIN was not able to include patients from other health care settings. Secondly, patients were identified using relevant Read code lists and algorithms. In addition, we were not able to do data stratification based on AF type (i.e. paroxysmal, persistent, permanent) and management strategy (i.e. rhythm vs. rate control), which may influence OAC prescription rates.

Conclusions
This study found that there was an increase in prevalence of AF in patients with T2DM between 2001 and 2016, and that both older and male patients were at higher risk of developing AF. The proportions of patients with T2DM who received OACs after AF diagnosis has increased during the study period. Further studies at individual and clinical practice level are warranted to investigate the factors associated with the underuse of OAC in patients with T2DM and AF in order to help in providing better responses and interventions in the management of this high-risk population.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons license, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this license, visit http://creat iveco mmons .org/licen ses/by/4.0/.