VA Risk Stratification in Dilated Cardiomyopathy
Definition
Risk stratification for ventricular arrhythmias (VA) and sudden cardiac death (SCD) in non-ischaemic dilated cardiomyopathy (DCM) is the process of identifying which patients are at high enough arrhythmic risk to benefit from implantable cardioverter-defibrillator (ICD) implantation. LVEF ≤35% has been the traditional threshold, but this criterion — derived from ischaemic disease — is increasingly recognized as insufficient for non-ischaemic DCM.
Key Concepts
Annual Event Rate and Baseline Risk
- The crude annual rate of sustained VA (sustained VT, VF, resuscitated SCD, or appropriate ICD discharge) in DCM is approximately 4.5% (meta-analysis, 55 studies, 11,451 patients, mean follow-up 3.7 years). (sources/VA-DCM-Sammani-2020, high)
- This event rate underscores the importance of ICD consideration in this population. (sources/VA-DCM-Sammani-2020, high)
Pooled Risk Predictors (Meta-Analysis Data)
Key pooled hazard ratios from the largest VA-specific DCM meta-analysis (Sammani 2020):
- LGE presence on CMR: HR 5.55 [4.02–7.67] — strongest single predictor (sources/VA-DCM-Sammani-2020, high)
- Prior sustained VA: HR 4.15 [1.32–13.02] — high heterogeneity (I²=93%) (sources/VA-DCM-Sammani-2020, high)
- Hypertension: HR 1.95 [1.26–3.00] (sources/VA-DCM-Sammani-2020, high)
- LVEF per 10% decrease (echocardiography only): HR 1.45 [1.19–1.78] — significant only when measured by echo, not when CMR and echo are pooled (sources/VA-DCM-Sammani-2020, high)
- LV dilatation (LVEDV/LVESV per 10 mL/m²): HR 1.10 for both parameters — small but significant (sources/VA-DCM-Sammani-2020, high)
- Younger age (per 10-year increase): HR 0.82 [0.74–1.00] — younger age confers higher risk (sources/VA-DCM-Sammani-2020, high)
Non-Pooled Predictors (Systematic Review Signal)
- Non-sustained VT: 9 of 14 studies reported significantly increased arrhythmic risk; could not be pooled due to missing HRs. Supported by AMIOVIRT and DEFINITE trial populations. (sources/VA-DCM-Sammani-2020, high)
- T-wave alternans (TWA): HR 6.5 [2.46–17.14]; clinical utility limited by poor standardization. (sources/VA-DCM-Sammani-2020, high)
- Genetic mutations — PLN, LMNA, FLNC: Each independently associated with VA risk in DCM; FLNC truncating variants associated with VA/SCD in 82% of patients in cohort studies. (sources/VA-DCM-Sammani-2020, high)
Non-Significant Predictors
- Male sex (trend only), NYHA class, family history of DCM, AF, LBBB, QRS duration, LVEF <30% or <35% as categorical thresholds, signal-averaged ECG, standard biomarkers (BNP, ANP, eGFR). (sources/VA-DCM-Sammani-2020, high)
- Left ventricular hypertrabeculation — added to this list at 1,160 patients. In the largest genotyped DCM cohort, hypertrabeculation was present in 30.5% and carried no arrhythmic signal after adjustment (HR 0.93, 95% CI 0.59–1.48); unadjusted it appeared protective (HR 0.64) because hypertrabeculated patients had higher LVEF and less LGE. It should not enter any risk model as a variable. (sources/trabeculation-dcm-circ-2026, rating: high) See concepts/LV-Hypertrabeculation.
What Actually Predicts Major Ventricular Arrhythmia in Genotyped DCM
The Circulation 2026 cohort provides a clean, centrally adjudicated hierarchy (median follow-up 5.1 years, 136 MVA events, 11.7%):
| Predictor | Adjusted HR (95% CI) |
|---|---|
| Nuclear envelope genotype (LMNA, FLNC group) | 4.4 (2.0–9.8) |
| LGE present | 3.1 (2.1–4.8) |
| LVEF, per 1% decrease | 1.03 (1.01–1.05) |
| Hypertrabeculation | 0.93 (0.59–1.48) |
- Genotype now outranks every imaging variable, and by a wide margin. This is the strongest single argument for including genotype in DCM risk stratification, and it is consistent with the genotype-specific calculators already endorsed by ESC 2023.
- For advanced heart failure in the same cohort, the predictors were genotype, LGE, NYHA class and LVEF — the same four variables, without hypertrabeculation. (sources/trabeculation-dcm-circ-2026, rating: high)
Diastolic Dysfunction as LVEF-Independent Arrhythmic Risk Predictor
- Grade III diastolic dysfunction (restrictive pattern) confers HR 3.52 (95% CI 2.00–6.22; P<0.001) for arrhythmic death or RCA in ischaemic and dilated cardiomyopathy, independent of LVEF. (sources/arrhythmia-diastolic-circep-2020, rating: medium)
- This association persists in both LVEF ≤35% (P=0.001) and LVEF >35% (P=0.014) subgroups, with 41% cumulative AD/RCA risk at 8 years in grade III patients with LVEF >35% — identifying a high-risk group missed by standard primary prevention ICD criteria. (sources/arrhythmia-diastolic-circep-2020, rating: medium)
- Cardiomyopathy aetiology (ischaemic vs. dilated) was not a significant confounder — the grade III effect was aetiology-independent. (sources/arrhythmia-diastolic-circep-2020, rating: medium)
- Grade I dysfunction was independently protective (HR 0.41; P=0.002); grade II was neutral. No patient with normal diastolic function died of arrhythmic death over 8 years. (sources/arrhythmia-diastolic-circep-2020, rating: medium)
- Caveat: Pilot study (n=210, single centre); diastolic grading used pre-2016 ASE criteria. Not included in current ESC/AHA VA risk stratification frameworks. See concepts/LV-Diastolic-Function for grading methodology.
LVEF Limitations
- The standard ICD threshold of LVEF ≤35% was established by SCD-HeFT (Bardy et al., NEJM 2005; n=2,521; 48% non-ischaemic; LVEF ≤35%; median 45.5 months): ICD HR 0.77 (P=0.007) with no aetiology interaction (P=0.68), confirming primary prevention benefit applies equally to non-ischaemic HF. However, the subsequent DANISH trial (n=1,116; LVEF ≤35%; median 67.6 months; enrolled 2008–2014, after ARNi availability) showed no all-cause mortality benefit in NICM (HR 0.87; P=0.28), despite halving SCD (HR 0.50; P=0.005); benefit was age-dependent (HR 0.64 in age <68 subgroup, P=0.01). The discordance between SCD-HeFT (ACEi + BB background only) and DANISH (partial ARNi era) may reflect progressive GDMT improvements independently reducing arrhythmic substrate. (sources/icd-hfref-scdheft-nejm-2005, sources/icd-nicm-danish-nejm-2016, sources/VA-SCD-ESC-2022, sources/VA-DCM-Sammani-2020, high)
- LVEF as a continuous echocardiographic variable predicts VA (HR 1.45 per 10%), but categorical cutoffs (LVEF <30% or <35%) are non-significant in pooled analysis, suggesting LVEF should not be the sole criterion. (sources/VA-DCM-Sammani-2020, high)
- Genotype-specific ICD criteria now extended to LVEF 35–50% range for high-risk genotypes (PLN, LMNA, FLNC, RBM20) with additional risk factors. (sources/VA-SCD-ESC-2022)
Multi-Parameter Risk Stratification (Current Guidelines)
- ESC 2022 multi-risk-factor ICD threshold (Class IIa): ICD in DCM/HNDCM with LVEF <50% AND ≥2 of: syncope, LGE on CMR, inducible SMVT at PES, pathogenic mutation in LMNA/PLN/FLNC/RBM20. (sources/VA-SCD-ESC-2022)
- ESC CMP 2023: Genotype-guided ICD considered in PLN, LMNA, FLNC, RBM20, DES, TMEM43 carriers with LVEF >35% and additional risk factors (NSVT, syncope, LGE). (sources/esc-cmp-2023)
- Genotype-specific calculators: LMNA risk-VTA calculator (lmna-risk-vta.fr); PLN p.Arg14del risk calculator endorsed for ICD decisions in respective mutation carriers. (sources/esc-cmp-2023)
Review Synthesis — ECG, Programmed Stimulation, Imaging and Genetics (Ebert 2019)
- ECG-based markers: in a meta-analysis of risk tools, T-wave alternans (TWA) and fragmented QRS had the highest odds ratios for arrhythmic events, with TWA the most sensitive; all non-autonomic parameters were significantly associated, but sensitivity (28.8–74.2%) and specificity (36.2–78.4%) are too low for individual decisions. (sources/vt-dcm-hrs-2018, rating: high)
- Programmed ventricular stimulation (PVS): pooled sensitivity 28.8%, specificity 87.1%, PPV 29.2% and NPV 86.9% — useful mainly to exclude risk. QRS prolongation during PVS, reflecting progressive activation delay, predicts inducible polymorphic VT/VF and correlates with long interstitial fibrosis strands rather than focal LGE. (sources/vt-dcm-hrs-2018, rating: high)
- Imaging: LGE-CMR detects fibrosis in 40–50% of DCM; midwall LGE identifies elevated SCD risk at LVEF ≥40%, while absence of LGE may identify low risk even at LVEF ≤35%. Patients without LGE can still have polymorphic VT/VF, especially after out-of-hospital cardiac arrest. (sources/vt-dcm-hrs-2018, rating: high)
- Syncope: event rates in DCM with syncope are similar to those with sustained VT, supporting syncope as a high-risk marker. (sources/vt-dcm-hrs-2018, rating: high)
- Genetics: LMNA carriers with LVEF <45%, non-sustained VT, male sex or non-missense variants are at independent risk; DES and FLNC truncating variants are also associated with arrhythmic risk, though the cohorts are small. (sources/vt-dcm-hrs-2018, rating: high)
Future Directions
- A dedicated DCM-VA risk calculator analogous to HCM Risk-SCD and the ARVC risk model is needed.
- Such a model should incorporate: younger age, hypertension, prior (non-)sustained VA, LVEF (echo), LV dilatation, LGE, genetic mutations (PLN, LMNA, FLNC), and sex.
- Competing risk analysis (VA vs. HF-related outcomes) is essential for accurate real-world risk estimation. (sources/VA-DCM-Sammani-2020, high)
Contradictions / Open Questions
- LVEF echo vs. CMR discordance: LVEF predicts VA only when measured by echocardiography (HR 1.45 significant), but not when CMR and echo are pooled (HR 1.30 non-significant). The modality-specific difference may reflect population-selection differences or systematic underestimation of LVEF by echo in DCM. This has practical implications for how LVEF thresholds are applied in ICD guidelines. (sources/VA-DCM-Sammani-2020, high)
- ESC 2022 dual ICD thresholds: Two Class IIa criteria (standard LVEF ≤35% and multi-factor LVEF <50%) coexist without guidance on which applies when a patient qualifies under both. (sources/VA-SCD-ESC-2022)
- Quality of evidence is moderate: Crude (unadjusted) HRs are pooled; confounder adjustment was limited in most primary studies. The true independent contribution of each risk factor remains uncertain. (sources/VA-DCM-Sammani-2020, high)
- Competing risks not modelled: DCM patients face both arrhythmic death and HF-related death/transplantation. Competing risks modelling may change individual risk estimates substantially. (sources/VA-DCM-Sammani-2020, high)
- Diastolic grade III as an additional ICD selection criterion (LVEF >35%): Current multi-parameter frameworks (LGE, genotype, LVEF, NSVT, syncope) do not incorporate diastolic grading. The Pezawas pilot study (n=210) found 41% cumulative AD/RCA risk in LVEF >35% patients with grade III dysfunction — comparable to conventional high-risk cohorts. The data are insufficient for guideline incorporation, and the pre-2016 diastolic grading criteria limit direct translation. Prospective validation in large cohorts using contemporary ASE diastolic grading is needed. (sources/arrhythmia-diastolic-circep-2020, rating: medium)
- Every individual non-invasive marker is too weak to decide alone. T-wave alternans and fragmented QRS top the odds ratios but have sensitivity of 28.8–74.2% and specificity of 36.2–78.4%; programmed stimulation has a positive predictive value of only 29.2%. How to combine them — and whether any combination beats the current multi-factor rule — remains untested in a prospective DCM cohort. (sources/vt-dcm-hrs-2018, rating: high)
- The best-powered DCM cohort to date still cannot tell us how to combine its own predictors. Genotype and LGE are each strong, adjusted hazard ratios are multiplicative in form but not in biology, and no validated combined calculator exists — the gap identified under Future Directions remains open, now with one more variable (nuclear envelope genotype) that clearly belongs in it. (sources/trabeculation-dcm-circ-2026, rating: high)
- A negative LGE-CMR does not exclude an arrhythmogenic substrate. Whole-heart histology shows that diffuse and patchy fibrosis can be invisible to LGE regardless of the sequence used, and patients without LGE still develop polymorphic VT/VF, particularly after cardiac arrest. Whether T1 mapping closes this gap is unproven. (sources/vt-dcm-hrs-2018, rating: high)
Connections
- Related to entities/DCM
- Related to concepts/LV-Hypertrabeculation — the trait that adds nothing to this model
- Related to entities/LMNA
- Related to entities/PLN
- Related to entities/FLNC
- Related to entities/NDLVC
- Related to concepts/Late-Gadolinium-Enhancement
- Related to concepts/Sudden-Cardiac-Death
- Related to concepts/HCM-Risk-SCD
- Related to concepts/Cascade-Family-Screening
- Related to concepts/LV-Diastolic-Function — grade III diastolic dysfunction as LVEF-independent arrhythmic predictor