Logistic regression to enhance risk assessment by left ventricular ejection fraction and f99

Abstract

Sudden cardiac death remains one of the leading causes of death in developed countries. Left ventricular ejection fraction (LVEF) and f99 are two noninvasive indexes of cardiovascular risk (traditional the former and innovative the latter) which, taken singularly, have not shown sufficiently high SCD predictive power to justify preventive actions. Thus, the aim of the present study was to investigate if their combination improves predictability of ventricular arrhythmias. To this aim, ECG recordings from 266 ICD patients, of which 76 developed ventricular tachycardia or fibrillation during the 4-year follow-up (ICD_Cases), and 190 did not (ICD_Controls). The ECGs of each patient was used to compute the f99, a repolarization index defined as the frequency at which the cumulative power energy reaches 99%. Eventually, a logistic regression between LVEF and f99 was performed in order to derive a combined predictor (CP) of ventricular arrhythmia. Goodness of each index was evaluated in terms of the area under the receiver operator curve (AUC). When used singularly, LVEF and f99 respectively provided an AUC of 0.67 and 0.64. When combined to get CP=-0.15-0.05•LVEF+0.03•f99, this provided an AUC of 0.71. In conclusion, use of logistic regression improves LVEF and f99 predictability of ventricular arrhythmias.