BACKGROUND:

Over the last three decades a sustained and marked decline in death rates from coronary heart disease (CHD) has occurred for all major demographic groups of the US population. Recently, however, the decline has proceeded less rapidly in blacks than in whites, and in women than men. These trends have focused further attention on possible heterogeneity in the risk factor patterns among the demographic sub-groups. As is well recognized, the knowledge on CHD risk factors and statistical models used to predict personal risk of CHD have been based on studies among white populations. In some epidemiological studies with samples from black populations, large variation in the effect of specific factors has been noted. Given the small sample of blacks under investigation, however, low statistical power exists for many of the black-white comparisons. The fundamental obstacle to progress in this area remains the absence of the cohorts of sufficient size that is representative of all four major groups. Hence, a strong rationale exists to undertake a "pooling project" to clarify a set of important unanswered questions in cardiovascular disease epidemiology.

DESIGN NARRATIVE:

Person-level meta-analysis will be conducted by pooling 9 large United States epidemiologic studies with both black and white samples: the First National Health and Nutrition Examination Survey (NHANES I) Epidemiological Follow-up Study, the NHANES II Mortality Follow-up Study, the Charleston Heart Study, the Evans County Heart Study, the Chicago Heart Association Detection Project in Industry, the Atherosclerosis Risk in Communities Study (ARIC), the Follow-up Study of the screenees for the Multiple Risk Factors Intervention Trial (MRFIT), the Follow-up Study of the participants from the MRFIT, and the Follow-up Study of the participants from the Hypertension Detection and Follow-up Program (HDFP). Black-white comparison will be performed on the CHD incidence and mortality, exposure-outcome relationship, patterns of co-morbidity (or coexistence of risk factors) and population attributable risk. Multivariate risk functions will also be examined in blacks and whites in the three contexts: ordering risk, magnitude of relative risks, and estimation of absolute risk.