Christopher B. Granger, MD
In patients with suspected coronary artery disease (CAD) and normal electrocardiography (ECG), does a nomogram model based on clinical and exercise variables predict mortality?
Prospective cohort study.
Cleveland Clinic, Cleveland, Ohio (derivation group), and Kaiser Permanente Colorado, Denver, Colorado (validation group), United States.
Patients ≥ 30 years of age (derivation group, n = 33 268, median age 52 y, 62% men; validation group, n = 5821, median age 55 y, 55% men) with normal ECGs and suspected CAD referred for symptom-limited exercise treadmill testing. Exclusion criteria were coronary disease, heart failure, left ventricular systolic dysfunction, cardiomyopathy, valvular or congenital heart disease, past renal or cardiac transplantation, atrial fibrillation, digitalis use, pacemaker or defibrillator placement, or end-stage renal disease.
Multivariable associations between clinical (age, sex, smoking history, hypertension, diabetes, and typical angina) and exercise (exercise capacity, ST-segment deviation, test angina, abnormal heart rate recovery, and ventricular ectopy during recovery) variables and mortality were assessed using Cox modeling and used to develop a nomogram. The nomogram was assessed for accuracy using the concordance index (c-index) (1 = perfect discrimination for mortality, 0.5 = random discrimination) and compared with the Duke treadmill score, a scale used for risk stratification.
Mortality was 5% in the derivation group (median follow-up 6.2 y) and 1% in the validation group (maximum follow-up 4.3 y). The nomogram combined predictors of mortality (Table) and predicted death (P < 0.001) with a c-index in the derivation group of 0.83 vs 0.73 with the Duke score (P < 0.001); the c-index in the validation group was 0.77. 3-year mortality rates were similar in both groups below (1.7% vs 2.5%) and above (25% vs 29%) a nomogram cut-point intended to identify low-risk patients (negative predictive value, 0.97).
A nomogram model based on clinical and exercise variables predicted mortality in patients with suspected CAD and normal ECG.
Significant predictors of mortality in patients with suspected CAD and normal ECG at a median follow-up of 6.2 years*
*CAD = coronary artery disease; ECG = electrocardiography; other abbreviations defined in Glossary. Other variables in the nomogram (hypertension, typical or test angina, and ST-segment deviation) did not significantly predict mortality.
Granger CB. In patients with suspected CAD and normal ECG, a nomogram model based on clinical and exercise variables predicted mortality. Ann Intern Med. ;148:JC4–14. doi: 10.7326/0003-4819-148-12-200806170-02014
Download citation file:
Published: Ann Intern Med. 2008;148(12):JC4-14.
Cardiac Diagnosis and Imaging, Cardiology.
Results provided by:
Copyright © 2018 American College of Physicians. All Rights Reserved.
Print ISSN: 0003-4819 | Online ISSN: 1539-3704
Conditions of Use