Multi-marker prediction of coronary heart disease risk: the Women’s Health Initiative [abstract] Abstract uri icon
  • Context: The utility of newer biomarkers remains uncertain when added to predictive models using only traditional risk factors for coronary heart disease (CHD) risk assessment.
    Objective: To investigate whether multiple biomarkers contribute to improved CHD risk prediction in postmenopausal women compared to assessment with traditional risk factors only.
    Design, Setting, Participants: The Women’s Health Initiative Hormone Trials (WHI-HT) enrolled 27,347 postmenopausal women aged 50 –79. Associations of traditional risk factors and 18 newer biomarkers were assessed in a nested case-control study including 321 CHD cases and 743 controls. We compared four prediction equations for 5-year CHD risk: Framingham Risk Score (FRS) models with original and newly-estimated coefficients; traditional risk factor (TRF) model which included statin treatment, hormone treatment, and previous cardiovascular disease history as well as the FRS covariates; additional biomarker (ABM) model which additionally included interleukin-6, D-dimer, coagulation Factor VIII, von Willebrand factor, and homocysteine (significant associations after adjusting for other variables).
    Main Outcome Measures: Nonfatal myocardial infarction, CHD death, and incident silent myocardial infarction. Statistical evaluation included measures of accuracy, discrimination, and reclassification.
    Results: The TRF model showed improved C-statistic (0.729 vs. 0.699, p= 0.001), significant integrated discrimination improvement (IDI, 0.011, p<0.001) and net reclassification improvement (NRI, 15.6%) when compared to the model with FRS covariates only. The ABM model showed improved discrimination (C-statistic=0.751, p=0.001; IDI=0.016, p<0.001) and reclassification (NRI=15.6%) compared to the TRF model. Predicted CHD risks in a continuous scale showed a high agreement between TRF and ABM models (Cronbach’s = 0. 899). Among newer factors, C-reactive protein level did not significantly improve CHD prediction either alone or in combination with other biomarkers.
    Conclusion: We found very modest improvement in CHD risk prediction when a group of 18 newer biomarkers were added to predictive models using traditional risk factors in postmenopausal women.

  • publication date
  • 2009
  • published in
  • Circulation  Journal
  • Research
  • Biomarkers
  • Cardiovascular Diseases
  • Forecasting
  • Heart Diseases
  • Models
  • Randomized Controlled Trials
  • Risk Factors
  • Additional Document Info
  • 120
  • issue
  • 18 Suppl