Limited Relationship of Voltage Criteria for Electrocardiogram Left Ventricular Hypertrophy to Cardiovascular Mortality

Background Numerous methods have been proposed for diagnosing left ventricular hypertrophy using the electrocardiogram. They have limited sensitivity for recognizing pathological hypertrophy, at least in part due to their inability to distinguish pathological from physiological hypertrophy. Our objective is to compare the major electrocardiogram–left ventricular hypertrophy criteria using cardiovascular mortality as a surrogate for pathological hypertrophy. Methods This study was a retrospective analysis of 16,253 veterans < 56 years of age seen at a large Veterans Affairs Medical Center from 1987 to 1999 and followed a median of 17.8 years for cardiovascular mortality. Receiver operating characteristics and Cox hazard survival techniques were applied. Results Of the 16,253 veterans included in our target population, the mean age was 43 years, 8.6% were female, 33.5% met criteria for electrocardiogram–left ventricular hypertrophy, and there were 744 cardiovascular deaths (annual cardiovascular mortality 0.25%). Receiver operating characteristic analysis demonstrated that the greatest area under the curve (AUC) for classification of cardiovascular death was obtained using the Romhilt-Estes score (0.63; 95% confidence interval, 0.61-0.65). Most of the voltage-only criteria had nondiagnostic area under the curves, with the Cornell being the best at 0.59 (95% confidence interval, 0.57-0.62). When the components of the Romhilt-Estes score were examined using step-wise Wald analysis, the voltage criteria dropped from the model. The Romhilt-Estes score ≥ 4, the Cornell, and the Peguero had the highest association with cardiovascular mortality (adjusted hazard ratios 2.2, 2.0, and 2.1, consecutively). Conclusion None of the electrocardiogram leads with voltage criteria exhibited sufficient classification power for clinical use.