Abstract 20587: A Prediction Rule Derived from Clinical Variables and Stress Cardiac Magnetic Resonance Imaging Provided Strong Cardiac Prognostication in Patient with Suspected Myocardial Ischemia
Objectives: We derived and validated a parsimonious predictive rule incorporating clinical and stress CMR data in characterizing cardiac events in patients with suspected ischemia.
Methods/Results: Out of 625 patients referred for CMR assessment of suspected ischemia, 19 were excluded for insufficient imaging quality. The remaining 606 patients were followed for cardiac events (100% follow-up; median of 2.9 years, IQR 2.4–6.9) including cardiac death, acute MI, cardiac hospitalization, and late coronary revascularization (>90 days). We randomized patients 1:1 into a training group (TRAIN, n=303) and a testing group (TEST, n=303). A prediction rule was built by stepwise Cox regression in the TRAIN group, considering clinical, ECG, and CMR data using P<0.05 for both entry and stay, adjusting to age and gender. The prediction rule was then validated on the TEST group. Age, gender, hx of PCI, hx of hypertension, late gadolinium enhancement, and reversible perfusion score by CMR (PerfSCORE) formed the prediction rule (Table 1). Eighty-nine patients (15%) experienced cardiac events, 49 (16%) and 40 (13%) in the TRAIN and TEST groups, respectively. The predictive rule demonstrated strong association with cardiac events in the TRAIN group (model LRχ2=79, p<0.0001) and maintained its robust prognostic value when tested on the TEST group (LRχ2=47, p<0.0001). Adjusted to variables in the prediction rule, PerfSCORE was the strongest multivariable predictor in both the TRAIN and TEST groups (adjusted HR 1.23 and 1.26, LRχ2 23 and 31, both P<0.0001, respectively). Fig 1 shows the ROC curves characterizing the association of the prediction rule with cardiac events, in the TRAIN and the TEST groups, respectively.
Conclusion: We derived and validated a prediction rule for cardiac events using clinical and CMR data for patients with suspected myocardial ischemia. PerfSCORE by CMR demonstrated strong prognostication to cardiac events in the prediction rule.
- © 2010 by American Heart Association, Inc.