Inter-relationships of different electrocardiographic indicators of left ventricular hypertrophy in 25,000 Chinese adults

Abstract Background Several electrocardiographic (ECG) indices have previously been proposed to diagnose left ventricular hypertrophy (LVH). Such indices utilize different criteria, including increased QRS voltage, prolonged QRS duration, left axis deviation, and LAFB and LBBB-like patterns in varying combinations. However, the inter-relationships of these different electrocardiographic indices of LVH with each otherare not fully understood. The purpose of this study is to estimate the relationships between the different LVH indices. Methods Electronic tracings of 12-lead ECGs recorded in 24,786 adult participants in the China Kadoorie Biobank recorded in 2014 were accessed to detect presence of LVH. ECG parameters of LVH, including individual and combinations of such parameters (voltage-only criteria such as Sokolov-Lyon, Cornell, Gubner as well as scores such as Romhilt-Estes point score) were estimated using customized computer software (Cardiolyse Oy, Finland) and compared proprietary LVH point score algorithms (Mortara, USA). Relationships between the different indices were assessed using a correlation matrix to explore associations between individual parameters and combined indices of such parameters. The matrices were used to construct heat maps and identify clustering of individual parameters. Results Overall, approximately 10% of the population had LVH assessed by Mortara algorithms. This graph shows that along with the free standing peaks, the rest are divided into two clusters of interconnected individual parameters. In the center of the largest cluster is the parameter, reflecting R-peak magnitude in aVL lead. In the center of another cluster was the left ventricular strain pattern, as described in Romhilt-Estes point score system. Overall, the Sokolov-Lyon index and LVH point score in Veritas algorithm had the strongest relationships with each other, but the Gubner and Cornell indices had relatively weak correlations with the other indices (Table 1). The mean correlation coefficient between five combined LVH indices was as follows: Veritas (Mortara) – 0.58, Sokolov-Lyon – 0.56, Romhilt-Estes – 0.45, Gubner – 0,32, Cornell – 0,3. Conclusions Parameters such as R-peak magnitude in aVL and left ventricular strain pattern were the most strongly related with each other. Aggregate indices, such as the Sokolov-Lyon index and LVH point score in Veritas algorithm were the most strongly related with all other indices, but the Gubner and Cornell indices