Association of metformin with lower atrial fibrillation risk among patients with type 2 diabetes mellitus: a population-based dynamic cohort and in vitro studies

This study used the “Longitudinal Cohort of Diabetes Patients Database (LHDB) 1999-2010,” which is released by the Taiwan National Health Research Institutes for research purposes. The National Health Insurance program, the universal health care system in Taiwan, was launched in 1995 and currently covers 23.72 million enrollees, representing about 99% of the national population. In the LHDB, a first diagnosis of DM between 1999 and 2010 is defined as: (1) the patient had at least one hospitalization with DM as one of the diagnoses (ICD-9-CM code 250 after 2000, or Acode A181 before 2000) or the patient used DM medication (Additional file 1: Supplement 1); or (2) the patient had at least two outpatient visits due to DM in the same year; and (3) the patient had not been hospitalized or treated as an outpatient due to DM before 1999. Approximately 120,000 eligible patients were randomly selected for each year from 1999 to 2010, and their original claims data including admission, outpatient, and medication use were stored in LHDB. For the eligible DM patients, the LHDB includes: (1) data after the DM diagnosis, (2) data before the diagnosis, and (3) data on the treatments for diseases other than DM. Rigid criteria were ensured diagnostic validity and minimized potential misclassification. The LHDB has been tested and confirmed by many other studies to be representative of the Taiwanese DM population [16]–[19]. The identification number of each patient was encrypted as part of an established privacy policy, and therefore informed consent was not required. This study was approved by the Institutional Review Board of the Chang Gung Medical Foundation, Taiwan.

The patients who were diagnosed with AF (ICD-9-CM code 427.31) prior to DM, who had two prescriptions for insulin or any anti-diabetic medication other than metformin, or who were younger than 18 years were excluded. The status of metformin use could be changed over time to reflect the actual prescription (Figure 1). The primary outcome was defined as new onset AF during the follow-up period (ICD-9-CM code 427.31, coded two consecutive times). The medications are listed in Additional file 1: Supplement 1, and the ICD-9-CM codes of co-morbidities are listed in Additional file 1: Supplement 2.

Most chemicals were purchased from Sigma (St Louis, MO, USA). HL-1 atrial myocytes were maintained in Claycomb medium with or without 1 mM of metformin (Sigma) for 2 hours. The treated atrial myocytes were then subjected to field stimulation as described previously [15],[20]. To induce tachycardia, HL-1 cells (≥1 × 10^6 cells) were cultured on 4-well rectangular dishes (Nuclon, Netherlands) and placed into C-Dish 100TM-Culture Dishes (IonOptix, Milton, MA, USA). The cells were paced with 10-ms stimuli of 40-V intensity at 4 Hz. Capture efficiency > 90% was confirmed by microscopic examination and by shortening of the action potential duration as described previously [15].

Intracellular oxidative stress from ROS was measured with fluorescent dye DCF-DA (2,7-dichlorofluorescein diacetate) and detected by confocal microscopy (Leica TCS SP2, Carl Zeiss, Jena, Germany) after pacing. DCF-DA was excited at 488 nm with an argon laser, and emission at 525 nm was recorded. Two-dimensional images (512 × 512 pixels) were acquired. Immunohistochemical and cytochemical analyses were performed by confocal microscopy using myosin heavy chain (MHC, Abcam) primary antibodies followed by fluorescein isothiocyanate (FITC) conjugated secondary antibodies (Chemicon, Temecula, CA, USA). Nuclei were visualized by diamidino-2-phenylindole (DAPI)-staining. Myosin degradation was quantified as the cytoplasmic MHC-area divided by the nuclear area. The relative expression levels of MHC were normalized to the control level. For each analysis, at least five random fields were chosen to observe more than 30 myocytes.

Equal amounts of protein in sodium dodecyl sulfate-polyacrylamide gel (SDS-PAGE) sample buffer was sonicated and subjected to electrophoresis on 12.5% SDS-PAGE. After the sample had been transferred to nitrocellulose membranes (Stratagene, Netherlands), the membranes were incubated with primary antibodies against cTroponin I (Fitzgerald Industries) or tubulin (Santa Cruz, Delaware Avenue, CA, USA). Signals were detected by an enhanced chemiluminescence substrate-detection method (Amersham, Netherlands) and quantified by densitometry. Signal bands were in the linear immunoreactive range, and they were expressed relative to tubulin.

From 1999 to 2010, 645,710 patients who were newly diagnosed type 2 DM, were not using insulin, were older than 18 years, and had not used other oral anti-diabetics were enrolled in this study. The mean follow-up duration was 5.4 years. Within the first year of DM diagnosis, 85,198 (13.2%) patients used metformin and 560,512 patients (86.8%) did not. The mean age of the study population was 58.6 ± 17.1 years, and males accounted for 49.8% of the patients. The demographic characteristics, relevant co-morbidities, and medications used in the first and ≥10th year after the diagnosis of DM are shown in Table 1. During the follow-up period after the diagnosis of DM, the metformin group had higher incidences of hypertension and sleep apnea syndrome, while the non-user group had higher incidences of congestive heart failure, chronic kidney disease, asthma, myocardial infarction, stroke, and peripheral arterial disease. In general, anti-hypertensive drugs were commonly used among the patients with DM, and the proportion increased from 62.8% to 70.7% during the follow-up period. Statins were the second most commonly used drug, with an increase in prevalence from 20.5% to 27.8%. The metformin group took more anti-hypertensives and statins than the non-user group (Additional file 1: Supplement 3).

The overall AF rate was 287 per 100,000 person-years, and the AF rate for the metformin user group was significantly lower than that of the nonuser group (245 for metformin users and 293 for nonusers, p < 0.0001). From the first to the second year after the diagnosis of DM, the metformin group had significantly lower incidence rates of AF than the non-user group (Figure 2). Figure 3 demonstrates Kaplan-Meier curves of the AF-free survival rates among the diabetic patients who did and did not use metformin. The metformin group had a significantly higher cumulative AF-free rate during the 13 years of follow-up than the non-user group (p < 0.001).

Unadjusted and adjusted HRs of AF calculated by Cox multivariate regression analyses are presented in Table 2. After adjusting for age, sex, hypertension, congestive heart failure, chronic kidney disease, asthma, myocardial infarction, ischemic stroke, peripheral arterial disease, the use of anti-hypertensives, and statins, the metformin group had a significantly lower AF occurrence rate with a HR of 0.81 (95% CI 0.76-0.86, p < 0.001).

A large body of evidence suggests that there is a link between oxidative stress and AF [21]. During tachyarrhythmia, substantial oxidative damage and cellular (ionic and structural) remodeling occurs in atrial myocardium that can induce AF [22]. We used MHC and troponin I degradation in cultured HL-1 atrial myocytes to evaluate the effects of metformin on preventing tachycardia-induced oxidative stress and on reducing the extent of cellular remodeling. HL-1 cells were paced at 4 Hz for 24 hours. Figure 4 shows that tachypacing induced intracellular oxidative stress from reactive oxygen species (ROS) in atrial myocytes, and that metformin reversed this effect. Tachycardia-induced myofibril degradation quantified by MHC and troponin-I was also rescued by 1 mM of metformin.