This study sought to establish the best definition of left ventricular adverse remodeling (LVAR) to predict outcomes and determine whether its assessment adds prognostic information to that obtained by early cardiac magnetic resonance (CMR). LVAR, usually defined as an increase in left ventricular end-diastolic volume (LVEDV) is the main cause of heart failure after an ST-segment elevated myocardial infarction; however, the role of assessment of LVAR in predicting cardiovascular events remains controversial. Patients with ST-segment elevated myocardial infarction who received percutaneous coronary intervention within 6 h of symptom onset were included (n = 498). CMR was performed during hospitalization (6.2 ± 2.6 days) and after 6 months (6.1 ± 1.8 months). The optimal threshold values of the LVEDV increase and the LV ejection fraction decrease associated with the primary endpoint were ascertained. Primary outcome was a composite of cardiovascular mortality, hospitalization for heart failure, or ventricular arrhythmia. The study was completed by 374 patients. Forty-nine patients presented the primary endpoint during follow-up (72.9 ± 42.8 months). Values that maximized the ability to identify patients with and without outcomes were a relative rise in LVEDV of 15% (hazard ratio [HR]: 2.1; p = 0.007) and a relative fall in LV ejection fraction of 3% (HR: 2.5; p = 0.001). However, the predictive model (using C-statistic analysis) failed to demonstrate that direct observation of LVAR at 6 months adds information to data from early CMR in predicting outcomes (C-statistic: 0.723 vs. 0.795). The definition of LVAR that best predicts adverse cardiovascular events should consider both the increase in LVEDV and the reduction in LV ejection fraction. However, assessment of LVAR does not improve information provided by the early CMR.