Risk Prediction Model Based on Biomarkers of Remodeling in Patients with Acute Anterior ST-Segment Elevation Myocardial Infarction

Abstract

BACKGROUND The aim of the present study was to develop a risk prediction model in patients with acute anterior ST-segment elevation myocardial infarction (STEMI). MATERIAL AND METHODS Clinical data from 333 patients with acute anterior STEMI were retrospectively analyzed. Clinical echocardiographic and angiographic data from patients with left ventricular remodeling (LVR) and those without LVR were compared. Factors that influenced risk were identified using multivariate logistic regression analysis. The area under the curve (AUC) of the receiver operating characteristic curve was used to assess the diagnostic performance of the model. RESULTS After 6-month follow-up, 135 of the patients experienced LVR (LVR group), whereas 198 did not (non-LVR group). Results of multivariate analysis showed that the number of stenosed coronary vessels, left ventricular end-diastolic volume (LVEDV), left ventricular ejection fraction (LVEF), transforming growth factor-beta (TGF-ß) at admission, and cardiac troponin I 3 days after admission (3-d cTnI) were all factors predictive of LVR in patients with acute anterior STEMI (all P<0.05). The established prediction model was Y=-20.639+0.711×number of stenosed coronary vessels + 0.137×LVEDV-0.129×LVEF+0.026×TGF-ß at admission + 0.162×3-d cTnI. The estimated AUC of this model was 0.978 (95% confidence interval [CI] 0.955-0.991), significantly superior to the single-factor numbers for stenosed coronary vessel of 0.650 (95% CI 0.597-0.702), LVEDV of 0.876 (95% CI 0.836-0.910), LVEF of 0.684 (95% CI 0.631-0.734), TGF-ß at admission of 0.696 (95% CI 0.644-0.745), cTnI at admission of 0.913 (95% CI 0.877-0.941), and 3-d cTnI of 0.945 (95% CI 0.914-0.967). CONCLUSIONS The established model had excellent diagnostic accuracy for predicting LVR in patients with acute anterior STEMI.

Figure 1

Forest plots for models 1 and 2.

Figure 2

Receiver operating characteristic curve for the prediction model.