ICD implantation remains the cornerstone of therapy for the secondary prevention of sudden cardiac death (SCD). However, the value of ICDs in the primary prevention of SCD among patients with heart failure (HF) remains a matter of ongoing debate. Several factors contribute to this uncertainty.
First, the main discriminator guiding current clinical decision-making is the left ventricular ejection fraction (EF), a functional parameter with limited predictive precision. Moreover, the assumption that EF behaves as a dichotomous (<35% versus >35%) rather than a continuous variable for predicting SCD risk has proven to be erroneous. Current European and US guidelines recommend ICD implantation for the primary prevention of SCD in patients with ischaemic cardiomyopathy or nonischaemic cardiomyopathy and an EF of <35%, if they remain symptomatic in New York Heart Association (NYHA) class II or III despite optimal therapy (class I recommendation), or even in NYHA class I (class I in the US and class II in Europe for ischaemic cardiomyopathy; class II for nonischaemic cardiomyopathy).1
The risk of SCD is not homogeneous among patients with EF <35%. Those who have suffered an acute MI (AMI) are at increased risk of ventricular arrhythmias more than 40 days after the acute event, when ventricular remodelling has already created an arrhythmogenic substrate.2 Even within this subgroup, however, primary prevention requires individualisation, as risk is highest in patients with recurrent infarction, severely reduced EF, significant ventricular dilation, and extensive myocardial fibrosis documented by cardiac MRI.3
A second major challenge in risk stratification for the primary prevention of SCD lies in the fact that accurate prediction of individual post-MI SCD risk using EF or other clinical predictors remains unfeasible after careful evaluation of individual patients from pooled cohorts.4
A third challenge concerns the changing epidemiology of SCD: the incidence of SCD has declined substantially over recent decades. The population enrolled in the original landmark trials no longer reflects today’s clinical reality. Accordingly, the number needed to treat to save one life has increased from five in the 1990s to 38 in the present era.5 This decline primarily reflects advances in coronary artery disease therapy and the widespread adoption of the four pillars of contemporary HF management: renin-angiotensin-aldosterone system inhibitors/angiotensin receptor neprilysin inhibitors, β-blockers, mineralocorticoid receptor antagonists, and sodium–glucose cotransporter 2 inhibitors. Nevertheless, patients with HF with reduced EF still carry a residual annual SCD risk exceeding 3%.1
The Risk Estimation Following Infarction Noninvasive Evaluation (REFINE) study, presented at the European Society of Cardiology Congress 2025 (not yet published), addressed the hypothesis that ICD therapy might benefit a specific subgroup of AMI patients characterised by impaired autonomic tone, a known risk factor for SCD. Previous small studies had failed to show a benefit for ICD in this context, but those findings were attributed to early post-event instability.
REFINE enrolled patients with left ventricular EF <40% in the first 48 hours after MI, or with persistently reduced EF <50% beyond 48 hours. The study evaluated T wave alternans, as well as parameters of heart rate variability and turbulence.6 Initiated in 2011 and completed in 2025, it involved 76 centres across Europe, Canada, the US, the Middle East and Africa, screening 1,943 patients and randomising 597 (mean age 65 years; women 21%). The ineligible patients were included in a Canadian registry as a control group. Inclusion criteria included left ventricular EF 36–50%, impaired heart rate turbulence, and abnormal modified moving average T wave alternans measured at least 2 months post-MI. Most participants were in NYHA class I or II with EF >40%. This group exhibited a twofold higher SCD risk compared with the control cohort (noneligible patients included in the Canadian registry).
However, during a mean follow-up of 5.7 years, prophylactic ICD therapy did not reduce all-cause mortality – the primary endpoint – compared with usual care (24.5% versus 21.3%; HR 1.07; 95% CI [0.77–1.50]). There were no significant differences in cardiac, sudden cardiac, or noncardiac mortality between groups. Notably, nearly half of all deaths (47.4%) were adjudicated as noncardiac. Cardiac mortality was 8.8% in the ICD group and 7.6% in the control group. SCD occurred in 2.6% of ICD patients versus 3.8% in controls – a low incidence consistent with recent reports.4
This study is significant for several reasons:
- It identifies a population at risk that does not warrant primary prevention with ICD.
- It underscores the need for novel research to identify additional SCD risk markers, including genetic markers and the use of artificial intelligence, to enable refined risk stratification.
- It highlights the importance of investigating other prognostic determinants of post-AMI mortality in HF, beyond arrhythmic causes, which might not be preventable through ICD implantation.
- It reinforces the limitations of relying on crude markers such as EF for device implantation decisions, particularly in light of the evolving risk profile of contemporary HF populations compared with those in earlier trials.