Abstract
Prospective studies harnessing late gadolinium enhancement (LGE) cardiovascular magnetic resonance (CMR) afford the potential to non-invasively characterise the phenotypic substrate for sudden cardiac death (SCD) and simultaneously interrogate its mechanistic drivers.
To assess the utility of infarct characterisation by CMR, including scar microstructure analysis, to predict SCD in prospectively investigated patients with coronary heart disease (CHD).
Patients with stable CHD were prospectively recruited into a registry between August 2009 and January 2016. The primary outcome for this study was SCD or aborted SCD. Patients with a secondary prevention implantable cardioverter defibrillator (ICD) indication were excluded. All patients had CMR with LGE imaging. Infarct quantification (core scar and peri-infarct zone [PIZ]) was performed by an independent level 3 CMR reader. Outcome events were adjudicated by a panel of cardiologists blinded to the CMR data. To investigate fibrosis microstructure, bespoke computational image processing algorithms were applied to the LGE images in order to extract specific morphological and texture related features.
Of 437 patients (mean age 64, mean left ventricular ejection fraction [LVEF] 47%, 91% with LGE) followed for a median of 6.3 years, 49 patients (11.2%) experienced the primary outcome. Patients with higher PIZ mass had an increased risk of the primary outcome (10-year risk 0.7%, 24.0% and 37.8% for patients with PIZ mass <5.66g, 5.66–12.28g and ≥12.29g respectively, P<0.001; figure 1a). On univariable analysis, an increase in PIZ mass and core infarct mass was associated with an increased risk of the primary outcome (per gram: HR 1.12, 95% CI 1.09–1.15, P<0.001 and HR 1.05, 95% CI 1.04–1.06, P<0.001 respectively). PIZ mass and core infarct mass remained independently associated with the primary outcome after adjustment for baseline predictors (per gram: HR 1.10, 95% CI 1.06–1.14, P<0.001 and HR 1.04, 95% CI 1.02–1.06, P<0.001 respectively) and together provided incremental value compared to conventional variables in predicting the endpoint (Harrell's C-statistic 0.76 to 0.82, figure 1b-c). Bespoke analysis of imaging data identified several shape-based scar metrics that associated with the primary outcome (figure 2). These included core infarct transmurality, radiality and interface length (the latter defining the core scar-PIZ boundary length), and the number of PIZ islets.
In this large prospective study of patients with stable CHD, both PIZ mass and core infarct mass independently predicted long-term SCD risk after adjusting for conventional predictors including LVEF. Reassuringly, minimal or absent LGE portended a comparatively low risk of SCD. Analysis of the scar microstructure identified several shape-based features that associated with SCD. These results highlight a potential avenue towards a more personalised approach to ICD implantation decisions.
