Medical history for prognostic risk assessment and diagnosis of stable patients with suspected coronary artery disease
Authors
James K. Min ; Allison Dunning ; Heidi Gransar ; Stephan Achenbach ; Fay Y. Lin ; Mouaz Al-Mallah ; Matthew J. Budoff ; Tracy Q. Callister ; Hyuk-Jae Chang ; Filippo Cademartiri ; Kavitha Chinnaiyan ; Benjamin J. W. Chow ; Ralph D’Agostino ; Augustin DeLago ; John Friedman ; Martin Hadamitzky ; Joerg Hausleiter ; Sean Hayes ; Philipp Kaufmann ; Gilbert L. Raff ; Leslee J. Shaw ; Louise Thomson ; Todd Villines ; Ricardo C. Cury ; Gudrun Feuchtner ; Yong-Jin Kim ; Jonathon Leipsic ; Daniel S. Berman ; Michael Pencina
Citation
AMERICAN JOURNAL OF MEDICINE, Vol.128(8) : 871-878, 2015
OBJECTIVE: To develop a clinical cardiac risk algorithm for stable patients with suspected coronary artery disease based upon angina typicality and coronary artery disease risk factors.
METHODS: Between 2004 and 2011, 14,004 adults with suspected coronary artery disease referred for cardiac imaging were followed: 1) 9093 patients for coronary computed tomography angiography (CCTA) followed for 2.0 years (CCTA-1); 2) 2132 patients for CCTA followed for 1.6 years (CCTA-2); and 3) 2779 patients for exercise myocardial perfusion scintigraphy (MPS) followed for 5.0 years. A best-fit model from CCTA-1 for prediction of death or myocardial infarction was developed, with integer values proportional to regression coefficients. Discrimination was assessed using C-statistic. The validated model was tested for estimation of the likelihood of obstructive coronary artery disease, defined as ≥50% stenosis, as compared with the method of Diamond and Forrester. Primary outcomes included all-cause mortality and nonfatal myocardial infarction. Secondary outcomes included prevalent angiographically obstructive coronary artery disease.
RESULTS: In CCTA-1, best-fit model discriminated individuals at risk of death or myocardial infarction (C-statistic 0.76). The integer model ranged from 3 to 13, corresponding to 3-year death risk or myocardial infarction of 0.25% to 53.8%. When applied to CCTA-2 and MPS cohorts, the model demonstrated C-statistics of 0.71 and 0.77, respectively. Both best-fit (C = 0.76; 95% confidence interval [CI], 0.746-0.771) and integer models (C = 0.71; 95% CI, 0.693-0.719) performed better than Diamond and Forrester (C = 0.64; 95% CI, 0.628-0.659) for estimating obstructive coronary artery disease.
CONCLUSIONS: For stable symptomatic patients with suspected coronary artery disease, we developed a history-based method for prediction of death and obstructive coronary artery disease.