Left ventricular hypertrophy by voltage criteria in hypertension: more than meets the eye?

Left ventricular hypertrophy is an important manifestation of end-organ damage in hypertension and has potential implications on the progression of cardiovascular disease including myocardial ischaemia/infarction, heart failure, cardiac arrhythmias and death [1–4]. Development of left ventricular hypertrophy is related to a variety of factors, including mechanical stress, humoral factors including insulin like growth factor-1 (IGF-1), transforming growth factor beta (TGF-β), catecholamines, angiotensin II, endothelin-1 and other inflammatory mediators, with different mechanisms of physiological and pathological hypertrophy [5–7]. Traditionally, ECG has been used to detect left ventricular hypertrophy despite a low sensitivity (11–13%), due to its universal availability and insignificant costs [1,8]. Transthoracic echocardiography is a useful imaging tool to assess for left ventricular hypertrophy and hypertensive heart disease (HHD). Left ventricular mass index (LVMI), left ventricular diastolic dysfunction (LVDD) and global longitudinal strain (GLS) are important markers of damage and have prognostic significance [9,10]. Four geometric patterns of left ventricular remodelling have been described by echo and include normal geometry, concentric remodelling, concentric hypertrophy and eccentric hypertrophy, all with prognostic significance [11,12]. Cardiac magnetic resonance (CMR) is the gold standard for ventricular function and mass assessment, but with limited availability. CMR allows tissue characterisation, measurement of volumes, myocardial mass, ejection fraction and regional kinetic disorders [13,14]. Further, it can also differentiate true cardiac hypertrophy from cardiac masses or pseudo masses and can analyse cardiac function and characterised tissues in the presence of confirmed cardiac hypertrophy [15–17]. CMR can also detect the four distinct LVH phenotypes as echocardiography. In the current issue, Naderi et al.[18] have explored the prognostic and diagnostic value of ECG predicted hypertension related LVH phenotypes using machine learning and have sought external validation of their findings from a population-based study from West Pomerania, NE Germany (n = 1474) [19,20]. They used data from the UK biobank (UKB) to identify over 23 000 hypertensive individuals (blood pressures ≥ 130/85 mmHg) with both ECG and CMR data and analysed their blood pressure readings, resting 12-lead ECG data and CMR parameters. The researchers considered clinical variables, including hypercholesterolaemia and diabetes, smoking status and alcohol consumption in the study. Interestingly, they also recorded association with cardiovascular outcomes in the UKB participants and defined primary endpoints. Appropriate methods, including logistic regression, support vector machine and random forest, were used to build classification models integrating ECG and clinical variables. Most participants were noted to have normal left ventricles or left ventricular remodelling (over 98%) and a small number with eccentric or concentric left ventricular hypertrophy. In the Study of Health in Pomerania (SHIP), just over 95% of the participants had either normal left ventricles or different remodelling, 3% had concentric LVH and the rest eccentric LVH. LVH by voltage criteria using the Sokolow-Lyon and Cornell criteria correlated better with the latter between the two groups (2 and 6% versus 8 and 9%, respectively), although the significance of this is unclear. Machine learning classifiers proved to be robust with satisfactory correlations between the UKB and the SHIP cohort. They identified an association between ECG-predicted eccentric LVH and heart failure, suggesting a potential marker for progression to heart failure in these patients. Those with eccentric LVH had the lowest ventricular rate, likely related to dilated left ventricle with resultant impaired contractility and increased filling pressures. ECG features, including ventricular rate, QRS amplitude and other measures related to the QRS complex, were analysed with changes heralded as a potential marker of remodelling. It would however have been useful to assess these findings and the machine learning models in relation to antihypertensive therapies and whether optimal blood pressure management influenced the observed findings. As acknowledged in the manuscript, both groups consisted of individuals belonging to white European background; similar studies in other ethnicities particularly the African Caribbeans and African–Americans would be useful considering that hypertension-related LVH is higher amongst them compared to white individuals and have a greater impact on cardiovascular prognosis. Exploration of the ECG-predicted phenotypes with clinical outcomes suggested a significant association between eccentric left ventricular hypertrophy and heart failure with an over three-fold increased hazard rate compared to those with normal left ventricular geometry because of myocardial thickening, resultant failure of muscle relaxation and diastolic dysfunction, with eventual deterioration in systolic dysfunction with chronic volume and pressure overload [21,22]. Artificial intelligence and its adaptation in medical science, with the ever developing algorithms and machine learning, is likely to play a significant role in the development of future models to accurately predict the journey and progression of health conditions such as hypertension, development of end-organ damage and their influence on clinical outcomes and perhaps pave the way for developing timely interventions that would potentially alter the course of such progression, hence preventing adverse clinical outcomes [13,14]. ACKNOWLEDGEMENTS Conflicts of interest There are no conflicts of interest.