COMPARING THE DISCRIMINATIVE ABILITY OF DIFFERENT ELECTROGRAPHIC CRITERIA FOR LEFT VENTRICULAR HYPERTROPHY IN PREDICTING CARDIOVASCULAR EVENTS IN HYPERTENSIVE PATIENTS. (April 2021)

• Record Type: Journal Article
• Authors: Harvey, George; Godec, Thomas; Kapil, Vikas; Mackay, Judith; Whitehouse, Andrew; Poulter, Neil; Sever, Peter S.; Gupta, Ajay
• Is Part Of: Journal of hypertension. Volume 39(2021)e-Supplement 1
• Publication Date: 2021-04
• Subjects: Hypertension -- Periodicals; Hypertension -- Periodicals; 616.132005
• DOI: 10.1097/01.hjh.0000746232.05690.50
• Languages: English
• ISSNs: 1473-5598

Abstract

Abstract: Objective: Several ECG criteria are used clinically to define Left Ventricular Hypertrophy (LVH), but there is no consensus on which of these criteria is better at predicting cardiovascular (CV) events in hypertensive patients. In this post-hoc analysis, our objective was to compare the discriminative ability and strength of association of different LVH criteria to predict the risk of CV-morbidity and mortality. Figure. No caption available. Design and method: We used the database of the Anglo-Scandinavian Cardiovascular Outcome Trial Our primary outcome was composite of fatal and non-fatal CV event including: coronary heart disease, stroke and coronary re-vascularisation. We developed several cox regression models, separately for each LVH criterion (Modified Sokolov, Cornell, Framingham, Strain Pattern, Cornell-Strain, Cornell-Product), after adjusting for a-priori confounders (age, sex, smoking status, diabetes, pulse pressure, chronic kidney disease stage 3) and other confounders found using stepwise selection (number of anti-hypertensives, previous vascular disease, metabolic syndrome, atrial fibrillation). We compared discriminative ability using c-statistic, goodness of fit of the model using Bayesian information criterion (BIC), hazard ratios (HR) and z-statistics for strength of association of each one of these LVH criteria in respective models. Results: 9.7% to 17.2% of 17959 hypertensive patients had LVH at baseline as defined by one or more LVH criterion (see Fig. 1). During a median follow up of 5.6 Years (IQR 5.2–6.1) there were 2736 CV events. Figure 1 shows that, whilst c-statistics and accuracy of most criteria are similar, strain pattern and the Cornell strain are better in improving the prediction models and with a higher strength of association with the CV-outcomes (HR 1.52 95% CI: 1.35 - 1.71, Z-statistic 6.88 and HR 1.37 95% CI: 1.226 - 1.519, Z-statistic 5.70, respectively). Both criteria had excellent BIC too. Conclusions: In hypertensive patients, compared to other ECG criteria of LVH, strain pattern and Cornell-strain index are stronger and better predictor of the risk of CV morbidity and mortality.