We evaluated Framingham Heart Study Offspring who attended either the fifth (1991–1994) or seventh (1998–2001) examination cycles (total possible person examinations=7338). Enrollment details have been described extensively elsewhere.7
We excluded 1,580 participants under the age of 50 years (low risk of incident AF), 214 with prevalent AF, 588 with prevalent diabetes (in which the insulin levels are not informative), and 373 participants with missing HOMA-IR at both examinations. The Institutional Review Board at Boston University Medical Center approved the study protocols for all examination cycles, and participants signed informed consents at each study visit.
Participants attended a routine Framingham Heart Study clinical visit every four to eight years. Blood pressure was measured at rest twice using a mercury column sphygmomanometer. Body mass index was calculated as the weight in kilograms divided by the square of the height in meters (kg/m2
). Clinically significant heart murmur was diagnosed by the presence of a grade three or more out of six systolic murmur, or any diastolic murmur, as determined during standardized examination at the Heart Study. Hypertension treatment was ascertained by self-report of medications. Diabetes was defined as fasting serum glucose ≥126 mg/dl, history of physician-diagnosed diabetes or use of medications for diabetes. Heart failure was adjudicated incorporating Framingham clinic and outside medical records on the basis of major and minor clinical criteria as described previously.8
Fasting blood samples were collected at each examination. Standardized insulin levels were measured in plasma as immunoreactive insulin. Insulin resistance was defined as the top quartile of the previously validated homeostasis model assessment of insulin resistance (HOMA-IR), assessed by the formula = (fasting plasma insulin [microunits per milliliter]) X (fasting plasma glucose [millimoles per liter])/22.5.9
Fasting plasma insulin was measured with different assays at the 2 examinations. At examination 5, we used EDTA plasma as total immunoreactive insulin and calibrated to serum levels for reporting purposes. Cross-reactivity of this assay with proinsulin at mid curve is approximately 40%; the intra-assay and interassay CVs ranged from 5.0% to 10.0%. At exam 7, the insulin level was specific, having essentially no cross-reactivity to insulin split-products (Linco Research, Inc., St. Charles, MO). Assay coefficient of variation was <10% on examination cycle 5 and <6.8% on examination cycle 7.10,11
Identification of AF was made based on records collected from participants’ Framingham examinations (interim cardiovascular events were routinely ascertained by Heart Study physicians), outside office visits and hospitalizations.12
For Framingham Offspring participants, biennial health history updates included a routine question on AF. Participants were determined as having AF if an electrocardiogram showed either atrial fibrillation or atrial flutter. Incident AF cases underwent review by one of two Framingham cardiologists (DL or EJB).
Descriptive statistics were examined including percentages for discrete variables and mean and standard deviations for continuous variables. Insulin resistance was analyzed as a dichotomous variable defined as the upper quartile of HOMA-IR (yes/no). We assessed HOMA-IR and clinical risk factors for AF at examination 5 and followed participants for development of incident AF for up to 10 years or their 7th
examination. We redefined the baseline characteristics at examination cycle 7, excluded individuals with interim AF and followed forward for the development of incident AF for up to 10 years. Multivariable-adjusted hazard ratios (HR) were estimated using Cox modeling to examine the relation between insulin resistance and AF up to 10 years of follow-up after confirming proportionality of hazards. Hazard ratios in the first model were adjusted for age and sex. In addition to age and sex, the second model was further adjusted for systolic blood pressure, hypertension treatment, PR interval, significant heart murmur, and heart failure – all established clinical risk factors for AF, as described previously.13
The third model had the same covariates as the second model plus body mass index. We determined the primary analysis had 80% power to identify an association between insulin resistance and new-onset AF at a HR of 1.41 with an alpha level of 0.05. All analyses were generated using SAS software, Version 9.1 (SAS Institute Inc. Cary, NC, USA). A two-sided p<0.05 was considered statistically significant.