The ARIC Study is a community-based prospective cohort study of 15792 participants aged 45–64 years at baseline and recruited from four communities in the United States.[7
] The first examination occurred 1987–89, with three follow-up visits taking place, each three years apart. Participants (or proxy) are contacted annually by telephone to ascertain information on hospitalizations and deaths. Active community-wide surveillance of local hospitals was performed to identify additional hospitalizations. Visit 2 (1990–92) was the only visit for which stored whole blood samples were available for the measurement of HbA1c; this was the baseline for the present study unless otherwise stated. There were 14348 participants who attended visit 2 of whom 1323 individuals were excluded for the following reasons: non-white and non-African-American (n=91); prevalent AF or atrial flutter (n=114); no electrocardiogram (ECG) or unreadable at baseline (n=273); non-fasting (<8 hours) blood sample (n=471); missing information on exposure or covariates of interest (n=374). The final sample size was 13025 individuals. The ARIC Study protocol was approved by institutional review boards at each site and informed consent was obtained from all study participants.
Blood collection and processing techniques from visit 2 have been previously described.[8
] Briefly, glucose was measured in serum using a hexokinase/glucose-6-phosphate dehydrogenase method. Insulin was measured by radioimmunoassay (125Insulin kit; Cambridge Medical Diagnosis, Bilerica, MA) in Visit 1 samples only.[8
] HbA1c was measured using high-performance liquid chromatography (Tosoh 2.2 Plus Glycohemoglobin Analyzer in 2003–2004 and the Tosoh G7 in 2007–2008, Tosoh Corporation) on all participants with available stored whole blood at Visit 2.[9
] The reliability of measurements from these stored samples has been previously reported.[10
] Attained educational level, income, cigarette smoking and use of antihypertensive and diabetic medications in the past two weeks were obtained from questionnaires and from documentation of medications.[8
] Sitting systolic BP was measured three times using a random-zero sphygmomanometer after 5 minutes of rest. The mean of the last two measurements were used for the analysis. BMI (kg/m2
) was computed from weight in a scrub suit and standing height.
Study participants with a FSG <100 mg/dL, a HbA1c<5.7%, no use of diabetic medication, and no history of physician-diagnosed diabetes were considered to have an optimal level of blood glucose and were categorized as having no diabetes.[11
] Individuals with a FSG 100–125 mg/dL or HbA1c 5.7–6.4%, no use of diabetic medication, and no history of physician-diagnosed diabetes were considered to have a sub-optimal glucose profile and classified as having pre-diabetes. Those with FSG≥126 mg/dL or HbA1c≥6.5% or use of diabetic medication or history of physician-diagnosed diabetes were categorized as having diabetes. Individuals with FSG≥126 mg/dL or HbA1c≥6.5% but no history of diabetic medication usage or physician-diagnosed diabetes were categorized as having undiagnosed diabetes.
Individuals with evidence of AF or atrial flutter on an ECG at visits 1 or 2, or a hospitalization for AF between both visits, were excluded from this analysis. Diagnoses of incident AF and atrial flutter were obtained through the end of 2007 from three sources: ECGs done at study visits 3 and 4, presence of an ICD9 code for AF/atrial flutter (427.31 or 427.32) listed on the hospital discharge record, or AF listed as any cause of death on the death certificate. Hospitalizations with AF associated with open cardiac surgery were not considered events. Date of AF incidence was the earliest of any AF diagnosis. All ARIC examination ECGs were recorded using MAC PC Personal Cardiographs (Marquette Electronics, Inc, Milwaukee, WI). A standard supine 12-lead resting ECG was recorded at each clinic visit and was transmitted by modem to the ARIC ECG Reading Center for automatic reading and coding. All AF events that were automatically detected from the study ECGs were visually rechecked by a cardiologist.[12
Means (or percentages) and standard deviations (SD) were calculated separately in those with diabetes, pre-diabetes and those without diabetes. The age-standardized incidence of AF for both diabetes and pre-diabetes was calculated. Associations between diabetes and pre-diabetes with AF were estimated using time-dependent Cox proportional hazards models with time to AF as the dependent variable. Separate race and gender analyses were conducted; models were adjusted for age, study site, education, income, prevalent CHD, BMI, systolic BP, antihypertensive medications, and smoking. Additional analyses included FSG in the model as a means to control for differences in control of diabetes. We explored the assumption of proportional hazards by computation of Schoenfeld residuals, and inspection of log(−log[survival function]) curves.
Initially, we explored the association between measures of diabetes markers and AF risk modeling FSG, fasting insulin, and HbA1c using restricted cubic splines. After confirming linearity, the association between FSG, fasting insulin and HbA1c levels with incident AF was examined separately in diabetics and non-diabetics using Cox models with adjustment for the same covariates as previously described. Chi-square tests for trend were conducted across categorical levels and tests for interaction with gender and race were performed. P-values <0.05 were considered significant.
We conducted sensitivity analyses excluding individuals with a history of CHD prior to visit 2. Prevalent CHD included individuals with a history of myocardial infarction (MI), MI adjudicated from the baseline ECG, or history of coronary bypass or angioplasty. We also examined the associations between tertiles of FSG and HbA1c with incident AF after reclassifying individuals as diabetic or not based on prior history of treatment for diabetes and self-reported history of diabetes (Y/N). To examine the relationship between duration of diabetes and risk of AF, self-reported age of diabetes diagnosis was obtained during an annual phone follow-up between between 1994–96 among 10371 individuals free of AF at visit 4 (1996–98).