Prediction of atherosclerotic cardiovascular disease mortality in a nationally representative cohort using a set of risk factors from pooled cohort risk equations

The American College of Cardiology/American Heart Association developed Pooled Cohort equations to estimate atherosclerotic cardiovascular disease (ASCVD) risk. It is unclear how well the equations predict ASCVD mortality in a nationally representative cohort. We used the National Health and Nutrition Examination Survey (NHANES) 1988–1994 and Linked Mortality through 2006 (n = 6,644). Among participants aged 40–79 years without ASCVD at baseline, we used Cox proportional hazard models to estimate the 10-year probability of ASCVD death by sex and race-ethnicity (non-Hispanic white (NHW), non-Hispanic black (NHB) and Mexican American (MA)). We estimated the discrimination and calibration for each sex-race-ethnicity model. We documented 288 ASCVD deaths during 62,335 person years. The Pooled Cohort equations demonstrated moderate to good discrimination for ASCVD mortality, with modified C-statistics of 0.716 (95% CI 0.663–0.770), 0.794 (0.734–0.854), and 0.733 (0.654–0.811) for NHW, NHB and MA men, respectively. The corresponding C-statistics for women were 0.781 (0.718–0.844), 0.702 (0.633–0.771), and 0.789 (CI 0.721–0.857). Modified Hosmer-Lemeshow χ(2) suggested adequate calibration for NHW, NHB and MA men, and MA women (p-values: 0.128, 0.295, 0.104 and 0.163 respectively). The calibration was inadequate for NHW and NHB women (p<0.05). In this nationally representative cohort, the Pooled Cohort equations performed adequately to predict 10-year ASCVD mortality for NHW and NHB men, and MA population, but not for NHW and NHB women.