Left ventricular wall thickness heterogeneity improves cardiovascular disease diagnosis and prognosis: a UK Biobank cardiovascular magnetic resonance cohort study

Abstract

Aims: Left ventricular hypertrophy (LVH) regionality carries diagnostic and prognostic importance. Mean absolute deviation of maximum segmental wall thickness (MadWT) is a novel left ventricular wall thickness (LVWT) heterogeneity biomarker from cardiovascular magnetic resonance imaging (CMR).

Objectives: To compare MadWT to indexed LV mass (LVMi), maximum (MaxWT) and mean (MeanWT) wall thickness to predict incident cardiovascular disease (CVD) and differentiate physiological from pathological LVH in highly physically active individuals.

Methods and results: Deep learning-assisted analysis of 44 930 UK Biobank CMR scans produced WT indices. Cox regression modelled major adverse cardiovascular events (MACE), heart failure (HF), arrhythmia, and all-cause death against LVWT indices. In the top 1% most physically active biomarker differences between propensity score matched hypertensive and non-hypertensive groups were compared. Over median (Q1, Q3) follow-up of 5.7 (4.9, 7.1) years, MadWT, MaxWT, MeanWT, and LVMi were associated with greater risk of MACE, HF, arrhythmia (P < 0.05), but not all-cause death (P > 0.05). After adjusting for CMR biomarkers, including LVMi, MadWT remained independently prognostic of the greatest number of endpoints, including MACE, HF, and arrhythmia [HR 1.13 (1.04-1.23); HR 1.15 (1.01-1.32); and HR 1.26 (1.18-1.35) respectively]. In the top 1% most physically active by three metrics, MadWT was the only significantly different biomarker between hypertensive and non-hypertensive participants (P < 0.05).

Conclusion: MadWT is important prognostically beyond LV mass and may be useful when differentiating physiological from hypertensive LVH. Although findings require confirmation in athletic and diseased cohorts, MadWT is readily translatable to deep learning-assisted clinical CMR reporting, especially in early unexplained LVH.

Keywords: biomarker; cardiovascular magnetic resonance imaging; heterogeneity; left ventricular hypertrophy; wall thickness.

Graphical Abstract

MadWT assesses LV wall thickness heterogeneity. MadWT quantifies LV wall thickness heterogeneity, a characteristic of LVH not captured by conventional indices, including LVMi and LVMVR. The marker (i) prognosticates independently and incrementally ischaemic and HF risk above conventional CV risk factors and CMR biomarkers; and (ii) facilitates the differentiation between physiological and pathological LVH in intermediate phenotypes. CI, confidence interval; CMR, cardiovascular magnetic resonance imaging; CV, cardiovascular; GLS, global longitudinal strain; HF, heart failure; HR, hazard ratio; LVH, left ventricular hypertrophy; LVMi, indexed left ventricular mass; LVMVR, left ventricular mass to volume ratio; MACE, major adverse cardiovascular events; MadWT, mean absolute deviation of maximum segmental wall thickness; MaxWT, maximum end-diastolic wall thickness; MeanWT, mean wall thickness; MET, metabolic equivalent of task; PA, physical activity; SD, standard deviation; WT, wall thickness.

Figure 1

Segmental wall thickness heterogeneity and MadWT. Heterogeneity in maximum segmental wall thickness is visibly greater in UK Biobank participants with (Panel B) vs. without (Panel A) established cardiac disease. On the historical American Heart Association 16-segment left ventricular model, mean absolute deviation of MadWT quantifies WT heterogeneity by first, calculating differences between maximum segmental and mean WT and second, deriving the mean of the 16 differences (Panels C and D).

Figure 2

Forest plot of effect sizes (A) and bar charts of C indices (B), comparing LVWT indices to predict CV endpoints. After adjusting for CV risk factors and established CMR biomarkers in Model 4, all three LVWT indices were strongly associated with arrhythmia events, but only MadWT remained predictive of MACE, HF, stroke and CV death (P < 0.05). Accordingly, the overall discriminative ability of Model 4 for arrhythmia was improved by MadWT, MaxWT, and MeanWT, for MACE by MadWT and MaxWT and for stroke and CV death only MadWT (△C-index P < 0.05). ^†Filled vs. unfilled marker indicates P < 0.05 vs. ≥0.05. CI, confidence interval; CMR, cardiovascular magnetic resonance imaging; CV, cardiovascular; HF, heart failure; LVWT, left ventricular wall thickness; MACE, major adverse cardiovascular event; MadWT, mean absolute deviation of maximum segmental wall thickness; MaxWT, maximum end-diastolic wall thickness; MeanWT, mean end-diastolic wall thickness.

Figure 3

Grouped bar charts, comparing CMR-derived WT, volume and functional parameters by hypertension status and PA level. In PSM cohorts, WT indices MadWT, MaxWT, MeanWT and volumetric parameters LVEDVi, LVMi, LVMVR, LVEF, but not GLS and native T1 were significantly different between hypertensive and non-hypertensive participants. In the top 1% most physically active by vigorous MET-min/week (Panel A) as well as mean acceleration vector (Panel B), only MadWT was consistently at least nominally significantly different between hypertensive and non-hypertensive groups with no overlap of their 95% CIs. *Bonferroni-corrected P < 0.006. **Bonferroni-corrected P < 0.0001. CI, confidence interval; CMR, cardiovascular magnetic resonance imaging; GLS, global longitudinal strain; LVEDVi, indexed left ventricular end-diastolic volume; LVEF, left ventricular ejection fraction; LVMi, indexed left ventricular mass; LVMVR, left ventricular mass to volume ratio; MadWT, mean absolute deviation of maximum segmental wall thickness; MaxWT, maximum end-diastolic wall thickness; MeanWT, mean end-diastolic wall thickness; MET, metabolic equivalent of task; PA, physical activity; WT, wall thickness.