Age, sex, hypertension and the sensitivity and specificity of traditional, new and a machine learning ECG criteria for prediction of left ventricular hypertrophy

Abstract

Objective

To evaluate the impact of age, sex and hypertension to improve ECG diagnosis of Left Ventricular Hypertrophy (LVH).

Methods

The study evaluated 14 different QRS voltage criteria as well as our recently proposed criteria of S in V3 plus S in V4 in a population of 159 patients, of whom 14.5 % had echocardiographic evidence of LVH. Statistical analyses assessed the influence of age, sex, and hypertension on the sensitivity and specificity of each criterion. In addition, a machine learning model was used for enhanced diagnostic accuracy.

Results

The new SV3 + SV4 criterion had the highest F1 and AUC scores. Among traditional ECG criteria, the Peguero criterion showed the highest sensitivity (0.438), while Wilson and Mazeloni criteria demonstrated the highest specificity (0.9412). The new SV3 + SV4 criterion with sex-specific cut offs achieves a sensitivity of 0.500 and specificity of 0.809 in females, while in males, sensitivity reached 0.556 with specificity at 0.910. Multiple regression analysis indicated that age, sex, and hypertension significantly improved the diagnostic performance of specific criteria, including Sokolow-Lyon, Romhilt voltage, Murphy, and Grant criteria. However, other criteria were not impacted by considering age, sex, or hypertension. ML analysis improved diagnostic accuracy with clinical variables, with the highest performance in males with the addition of age (accuracy 0.959, sensitivity 0.556, and specificity 1.00).

Conclusion

Considering age, sex, and hypertension can enhance the diagnostic performance of certain ECG criteria and especially in a ML model for LVH. Findings support a more individualized approach for LVH diagnosis in diverse patient populations.