The study was approved by the Johns Hopkins Institutional Review Board. Caucasian and African American probands with premature (< age 60) CAD events (sudden death or myocardial infarction [26.4%], coronary artery bypass surgery [35.0%], coronary angioplasty [28.6%], or anginal symptoms with a ≥ 50% lesion in one or more vessels by coronary angiography [9.1%]) were identified at the time of event from 10 Baltimore area hospitals from 1983 to 1996. CAD events associated with aortic stenosis, cardiac transplantation, chronic glucocorticosteroid therapy, chest irradiation, or cocaine intoxication were excluded. Their siblings < 60 years of age and apparently free of CAD were recruited for screening. Siblings with known autoimmune disease, taking chronic glucorticosteroids, or who had any life-threatening diseases (e.g., AIDS, cancer) with a life expectancy of <5 years were also excluded.
Participants gave informed consent prior to screening. During screening, information regarding demographics was obtained from interviewer-verified self-administered questionnaires. Medical history, including current medication use, was obtained by a physician or nurse, and a physical examination was performed by a cardiologist. All measures of risk factors were identical or nearly identical to those measured in the Framingham Heart Study.14
Any reported cigarette smoking within one month or expired carbon monoxide ≥ 8 ppm on 2 readings were considered current smoking. Venous blood was obtained after a 12-hour overnight fast. Total and high density lipoprotein (HDL) cholesterol, triglycerides, and glucose were measured in the Johns Hopkins Analytical Chemistry Laboratory using standardized methods (coefficient of variation for total cholesterol <3%). Low density lipoprotein (LDL) cholesterol was calculated if triglyceride levels were < 400 mg/dL.19
Diabetes mellitus was defined as a self-reported physician’s diagnosis, current insulin or hypoglycemic medication use, and/or a measured fasting glucose ≥ 126 mg/dl. Blood pressure was measured using methods described by the Joint National Committee on the Detection, Evaluation, and Treatment of High Blood Pressure,20
and the average of three resting measures was used. Hypertension was defined as resting blood pressure ≥ 140/90 mmHg and/or current antihypertensive medication use.
The Wilson et al 1998 FRE scoring system provides estimates of 10-year absolute risks for total CAD.14
For calculating the risk using this model, we followed the same sex-specific FRE14
scoring system based on categories of risk for baseline levels of total cholesterol and blood pressure, smoking, diabetes mellitus and age as for the Framingham Heart Study.14
Total cholesterol categories (in mg/dL) were: (1) <160, (2) 160–199, (3) 200–229, (4) 230–239, (5) 240–279, (6) ≥280. HDL categories (in mg/dL) were: (1) < 35, (2) 35–44, (3) 45–49, (4) 50–59, (5) ≥60. Blood pressure (BP) categories (in mmHg) were: (1) diastolic BP < 80 and systolic BP < 120, (2) diastolic BP = 80–84 or systolic BP = 120–129, (3) diastolic BP = 85–89 or systolic BP = 130–139, (4) diastolic BP = 90–99 or systolic BP = 140–159, (5) diastolic BP > 100 or systolic BP > 160. Classification was by the highest value for either the systolic or diastolic blood pressure.
Siblings were contacted periodically by trained telephone interviewers for follow-up from 1993–2005 and completed a standardized health status and cardiovascular disease event questionnaire. Medical records were obtained for all siblings reporting a CAD event, any possibly related diagnosis, related diagnostic procedure (exercise test, coronary calcium scan, thallium imaging, stress echocardiography, or coronary angiography) or related therapeutic procedure (including percutaneous coronary intervention or coronary bypass surgery). For this study, total events included all CAD events as defined in the Framingham Heart Study, using the same classifications and definitions.14,21
Total CAD events thus included angina pectoris, myocardial infarction, or coronary heart disease death. Event adjudication was done using a manner similar to the Framingham Heart Study.14,21
Medical records were independently reviewed by two cardiologists and one cardiovascular epidemiologist. Whenever there was disagreement in adjudication, an External Adjudication Committee consisting of at least one non-study cardiologist from Johns Hopkins and a cardiologist from another institution reviewed the event and determined the final event classification using the standardized coding schema.
All analyses were sex-specific. Demographic and risk factor characteristics of the cohort were tabulated according to methods similar to those used in the Framingham Heart Study.14
The number of total CAD events expected within 10-years in siblings based on the FRE was calculated as a sum of proportions with its standard error. This expectation was compared against the actual events observed in the sibling cohort with binomial standard error. In preliminary survival analysis, we examined the necessity for including a family-specific frailty to account for different baseline hazard functions in family clusters. There was no statistically significant within-family frailty, thus models without frailty are presented.
We compared 10-year cumulative incidence in siblings with their baseline Framingham 10-year predicted risk using the techniques recommended by the Framingham Heart Study by D’Agostino et al.18
A Cox regression analysis was fitted to the data using the 1998 FRE risk factors as independent variables.14
The coefficient estimates and their standard errors were compared against the coefficients from the standard FRE.14
For these comparisons, we used p < 0.10 as the critical value for significance testing, as recommended by D’Agostino et al.18
Calibration analysis was done using the Hosmer Lemeshow χ2
statistic. The observed and expected numbers of events predicted by the FRE were calculated for every decile of risk. The χ2
statistic was considered significant at χ2
> 20 (p < 0.01) as suggested.18
Also, as recommended by the Framingham Heart Study whenever the FRE is applied to a different population, the FRE was recalibrated specifically to the sibling population.18
Recalibration involved the estimation of two parameters: (1) the mean survival, and (2) the G-statistic. The mean survival was calculated as (1- the observed cumulative 10-year incidence). The G-statistic is a population-specific constant used for the calculation of cumulative survival from the relative hazard function. This method has been used to validate the predictive capacity of the FRE in different ethnic groups.18