BACKGROUND:Twelve-lead ECG represents the most common diagnostic tool in clinical cardiology and allows an immediate screening of left ventricular hypertrophy (LVH), but current criteria might have poor clinical usefulness in everyday clinical practice due to lack of sensitivity. METHODS:The current study aims to review and compare the clinical performance of known ECG criteria of LVH in a real-life setting; 2134 patients had ECG and echocardiographic exams performed during the same hospitalization. All traces were retrospectively analysed, and the amplitudes of the waves were manually measured. Transthoracic echocardiography was considered as the gold standard to assess LVH. RESULTS:LVH had a prevalence of 58%. Considering the diagnostic performance of ECG criteria for LVH, the Cornell voltage carried the best area under the receiver operating characteristic curve (0.678), while RaVF (R wave in aVF lead) had the poorer result (0.440). The R5/R6 criterion had the best sensitivity (60%), but with the worst specificity (37.4%). The 'Q or S aVR' had the best specificity (99.9%) but lacks sensitivity (0.80%). The Peguero Lo Presti criterion had a sensitivity of 42.3% and a specificity of 75.8%. The Cornell voltage and the Cornell product had similar area under the receiver operating characteristic curve values which were found to be significantly greater compared with other criteria. CONCLUSION:Current ECG criteria of LVH have low sensitivity despite an acceptable specificity. Among these, Cornell voltage and Cornell product criteria were equally found to have a more accurate diagnostic performance compared with other criteria. To overcome the intrinsic limitations of the current ECG LVH criteria, a new paradigm in the analysis of electrocardiographic data might be necessary.