Implantable Cardioverter-Defibrillator (ICD)

Details

An ICD is a surgically implanted device that continuously monitors cardiac rhythm and delivers high-energy shocks (defibrillation) or antitachycardia pacing (ATP) to terminate life-threatening ventricular arrhythmias (VF, sustained VT). Modern devices are also capable of bradycardia pacing and can incorporate CRT (CRT-D). ICD use spans two broad indications: (1) secondary prevention — after a survived cardiac arrest or sustained VT; (2) primary prevention — before a first life-threatening arrhythmia event, based on structural, functional, or genetic risk criteria.

Device types: Transvenous ICD (TV-ICD; standard), Subcutaneous ICD (S-ICD; no intracardiac leads), CRT-D (biventricular pacing + ICD), Wearable defibrillator (LifeVest; temporary bridge).

Key evidence trials: MADIT-I/II (CAD primary prevention); SCD-HeFT (NICM/CAD primary prevention); DANISH (NICM; no significant overall benefit in modern GDMT era); AVID/CASH/CIDS (secondary prevention); PARADIGM-HF/DAPA-HF era now shifts GDMT impact on LVEF recovery.

Key Facts

Secondary Prevention — AUC 2025 Ratings

• Documented VF or sustained VT (not due to reversible transient cause): universally Appropriate across all structural heart disease types (sources/icd-crt-auc-2025, rating: high)
• Genetic channelopathies post-sustained VT/VF (LQTS, SQTS, CPVT, Brugada, ARVC, HCM): A(9) (sources/icd-crt-auc-2025, rating: high)
• CAD + documented VF without recent MI (>40 days): A(8–9) regardless of LVEF (sources/icd-crt-auc-2025, rating: high)
• Acute MI (<40 days) VT/VF within 48h: R(2–3) — likely reversible; VT/VF later in admission + LVEF ≤35%: M(5–6) (sources/icd-crt-auc-2025, rating: high)
• Reversible cause (electrolytes, drug toxicity, acute ischaemia with revascularisation): R(1–3) (sources/icd-crt-auc-2025, rating: high)

Primary Prevention — CAD

• 40-day post-MI waiting period mandatory before primary prevention ICD assessment (sources/icd-crt-auc-2025, rating: high)
• CAD + LVEF ≤30% + >40 days post-MI + ≥3 months post-revascularisation: A(8–9). MADIT-II (n=1,232; 3:2 ICD vs conventional; no EPS required; mean LVEF 23%; mean 20-month follow-up): all-cause mortality HR 0.69 (95% CI 0.51–0.93; P=0.016); 14.2% vs 19.8%; consistent across all subgroups; benefit emerged at ~9 months. Trial stopped early at efficacy boundary. HF hospitalisation trend higher in ICD arm (19.9% vs 14.9%; P=0.09). SCD-HeFT (Bardy et al., NEJM 2005; n=2,521; 52% ischaemic + 48% non-ischaemic; LVEF ≤35%; NYHA II [70%]/III [30%]; median 45.5 months): ICD HR 0.77 (97.5% CI 0.62–0.96; P=0.007); no aetiology interaction (P=0.68) — aetiology-agnostic benefit; extends the evidence base to LVEF ≤35%. The current Class I primary prevention ICD threshold of LVEF ≤35% derives directly from SCD-HeFT (MADIT-II used ≤30%). (sources/icd-hfref-scdheft-nejm-2005, sources/icd-icm-maditii-nejm-2002, sources/icd-crt-auc-2025, rating: very high)
• CAD + LVEF 31–35% + NSVT + EPS-inducible sustained VT: A(7–8) (sources/icd-crt-auc-2025, rating: high)
• CAD + LVEF 31–35% + no NSVT: M(4–5) (sources/icd-crt-auc-2025, rating: high)
• 90-day post-revascularisation waiting period applies before reassessing LVEF and ICD candidacy (sources/icd-crt-auc-2025, rating: high)

Primary Prevention — NICM

• <3 months GDMT: R(1–3) — GDMT may recover LVEF; ICD not appropriate before trial (sources/icd-crt-auc-2025, rating: high)
• ≥3 months GDMT + LVEF ≤35% + NYHA II–III, age <75: A(7–9) (sources/icd-crt-auc-2025, rating: high)
• ≥3 months GDMT + LVEF ≤35% + NYHA II–III, age ≥85: M(5–6) — competing mortality reduces net benefit (sources/icd-crt-auc-2025, rating: high)
• LGE-positive NICM + LVEF 36–49%: M(5–6) — emerging evidence, not yet A (sources/icd-crt-auc-2025, rating: high)
• NYHA IV + on transplant list/LVAD candidate: A(7); Refractory, no device option: R(2) (sources/icd-crt-auc-2025, rating: high)
• SCD-HeFT — foundational NICM ICD evidence: SCD-HeFT (Bardy et al., NEJM 2005; n=2,521; 48% non-ischaemic; LVEF ≤35%; NYHA II [70%]/III [30%]; ACEi + BB background; pre-SGLT2i/ARNi era; median 45.5 months): ICD HR 0.77 (P=0.007); no aetiology interaction (P=0.68) — first large RCT establishing that ICD primary prevention benefit applies equally to non-ischaemic HF. Establishes the ≤35% guideline threshold. (sources/icd-hfref-scdheft-nejm-2005, rating: very high)
• DEFINITE — first large NICM-specific primary prevention RCT: n=458; NICM; LVEF <36%; ambient arrhythmias (NSVT or ≥10 PVCs/hour); symptomatic HF; ACEi 85.6% + beta-blocker 84.9%; 37 US centres; mean 29.0 months; St. Jude Medical-funded. Single-chamber ICD (shock-only, VF zone 180 bpm) vs standard therapy. Primary (all-cause mortality): HR 0.65 (95% CI 0.40–1.06; P=0.08) — NOT significant. 2-year mortality 7.9% vs 14.1%. Secondary (sudden arrhythmic death): HR 0.20 (95% CI 0.06–0.71; P=0.006) — significant; 3 vs 14 sudden deaths. Subgroups (prespecified): men HR 0.49 (P=0.018); NYHA III HR 0.37 (P=0.02). Underpowered because only ~1/3 of deaths were arrhythmic (vs >50% assumed); high ACEi/beta-blocker compliance eliminated a larger portion of background SCD than anticipated — the same competing-mortality pattern subsequently replicated in DANISH (2016). (sources/icd-nicm-definite-nejm-2004, rating: high)
• COMPANION — CRT-D mortality in NYHA III/IV with QRS ≥120ms: COMPANION (Bristow et al., NEJM 2004; n=1,520; NYHA III/IV; LVEF ≤35%; QRS ≥120ms; ischaemic + non-ischaemic; Guidant-funded): CRT-D vs OPT all-cause mortality HR 0.64 (95% CI 0.48–0.86; P=0.003) — 36% reduction; first large RCT demonstrating statistically significant mortality benefit with CRT + ICD in a population including non-ischaemic CM. Non-ischaemic CM subgroup: HR 0.50 (P=0.015). CRT-P mortality borderline non-significant (HR 0.76; P=0.059 — subsequently confirmed by CARE-HF 2005 with longer follow-up). Published in same NEJM issue as DEFINITE; together they established the evidence base for ICD therapy in NICM prior to DANISH. (sources/crt-companion-nejm-2004, rating: very high)
• DANISH caveat: DANISH trial (n=1,116; LVEF ≤35%; NYHA II–IV; NT-proBNP >200; 58% CRT; median 67.6 months; enrolled 2008–2014): all-cause mortality HR 0.87 (95% CI 0.68–1.12; P=0.28) — NOT significant; SCD HR 0.50 (P=0.005) — halved; CV death HR 0.77 (NS); 31% non-CV deaths. Age <68 subgroup: HR 0.64 (P=0.01) — significant. Device infections (non-CRT ICD) HR 6.35 (P=0.006); inappropriate shocks 5.9%. DANISH did not include SGLT2i or sacubitril–valsartan; modern quadruple GDMT likely reduces background SCD further, potentially narrowing the NNT further. (sources/icd-nicm-danish-nejm-2016, sources/icd-crt-auc-2025, rating: high)

Primary Prevention — Specific Etiologies

(Ratings apply regardless of GDMT duration given disease-specific arrhythmic substrate)

• Giant-cell myocarditis + LVEF ≤35%: A(7–8) (sources/icd-crt-auc-2025, rating: high)
• Cardiac sarcoidosis + LVEF ≤35%: A(7); + LGE or prior AVB but LVEF 36–49%: M(6) (sources/icd-crt-auc-2025, rating: high)
• Myotonic dystrophy type 1 + PR >240ms or QRS >120ms: A(8); no conduction disease: M(5) (sources/icd-crt-auc-2025, rating: high)
• Chagas cardiomyopathy + LVEF ≤35%: A(8); LVEF 36–49% + NSVT: M(6) (sources/icd-crt-auc-2025, rating: high)
• ATTR amyloidosis + LVEF ≤35% + NYHA II–III: M(6); NYHA IV: M(5) (sources/icd-crt-auc-2025, rating: high)
• Acute lymphocytic myocarditis, recovering: M(4); once LVEF normalised: R(3) (sources/icd-crt-auc-2025, rating: high)

Primary Prevention — Genetic Conditions

• HCM + ≥1 major SCD risk factor (LVEF <50%, LV wall ≥30mm, unexplained syncope, FH of SCD, NSVT, abnormal BP response on exercise): A(8) (sources/icd-crt-auc-2025, rating: high)
• ARVC + ≥1 risk factor (syncope, extensive RV dysfunction, NSVT) without prior VA: A(7) (sources/icd-crt-auc-2025, rating: high)
• LMNA + LVEF 35–45% + ≥1 risk factor: A(7); LVEF ≤35%: A(8) (sources/icd-crt-auc-2025, rating: high)
• Gene carrier + normal echo/ECG (no phenotypic expression): M(4) (sources/icd-crt-auc-2025, rating: high)
• CPVT + NSVT despite beta-blocker + flecainide: M(5–6) (sources/icd-crt-auc-2025, rating: high)
• LQTS + asymptomatic + QTc >500ms + beta-blocker intolerant: M(6) (sources/icd-crt-auc-2025, rating: high)

Comorbidity Modifiers

• Age ≥85 + NYHA II + LVEF ≤35%: M(5–6) (sources/icd-crt-auc-2025, rating: high)
• Active IV/illicit substance use disorder: R(2) — device infection risk (sources/icd-crt-auc-2025, rating: high)
• Dialysis-dependent ESRD + LVEF ≤35%: M(4–5) — high competing mortality (sources/icd-crt-auc-2025, rating: high)
• Severe frailty + LVEF ≤35%: M(4–5) (sources/icd-crt-auc-2025, rating: high)

Generator Replacement

• Prior indication valid + LVEF still ≤35%: A(8) (sources/icd-crt-auc-2025, rating: high)
• LVEF improved 36–49%: M(6) (sources/icd-crt-auc-2025, rating: high)
• LVEF normalised ≥50%, no prior VA: M(4) (sources/icd-crt-auc-2025, rating: high)
• Prior sustained VA regardless of current LVEF: A(8–9) (sources/icd-crt-auc-2025, rating: high)

ICD Programming — Key Recommendations (HRS/EHRA 2015)

The following summarises the 2015 international consensus (32 recommendations; 96% mean ballot consensus; HRS/EHRA/APHRS/SOLAECE). See sources/icd-programming-hrs-2015 for full detail. (sources/icd-programming-hrs-2015, rating: high)

Tachycardia Detection:

• Prolonged detection (30/40 intervals or equivalent time delay) as default — supported by PREPARE, RELEVANT, MADIT-RIT, ADVANCE III, PROVIDE (~7,000 patients): 50% reduction in inappropriate shocks; 30% reduction in all-cause mortality meta-analysis; no increase in syncope
• VF zone rate cutoff ≥200 bpm for primary prevention — MADIT-RIT Arm B (200 bpm): 80% ↓ inappropriate therapy; mortality 3.2% vs 6.6% (conventional 170 bpm); overlap of SVT/VT rates concentrated 181–213 bpm
• Enable SVT-VT discrimination algorithms — morphology (primary, single-chamber), stability (AF rejection), onset (sinus tachycardia rejection), dual-chamber AV rate comparison; SVT limit ≤230 bpm in adults
• Multi-zone programming preferred; ATP before or during charging in all VT zones
• S-ICD: dual-zone (conditional shock zone + shock zone) preferred; rate and morphology-based discrimination; 18/24 interval detection (nonprogrammable); 80 J shock (nonprogrammable)

Tachycardia Therapy:

• ATP first-line for all VT, including fast VT (CL 240–320 ms / 188–250 bpm) — PainFREE Rx II: 71% relative shock reduction; second ATP burst increases first-burst efficacy from 64% to 83%
• Up to 2 ATP bursts before escalation; 8-pulse burst as effective as 15-pulse (ADVANCE-D RCT); burst > ramp for fast VT (PITAGORA ICD)
• ATP during capacitor charging: safe and effective
• ICD shocks: 5-fold mortality increase in SCD-HeFT (appropriate shocks); 2-fold (inappropriate shocks) — shock minimisation is a primary programming goal

Bradycardia Mode:

• Minimise RV pacing in ICD patients without bradycardia indication — DAVID trial: DDDR-70 associated with worse HF/mortality vs VVI-40 backup pacing due to unnecessary RV pacing
• MVP or extended AV delay algorithms to reduce RV pacing burden; avoid AV intervals >250–300 ms
• CRT-D: biventricular pacing percentage ≥98% target — MADIT-CRT: <90% biventricular pacing had no CRT benefit vs ICD alone; every 1% ↑ → 6% ↓ HF/death, 10% ↓ death alone
• AF + AVJ ablation: minimum pacing rate 80–90 bpm for first 1–2 months post-ablation (QT prolongation/SCD risk)

Defibrillation Threshold Testing (DT):

• Routine DT can be safely omitted for standard left-sided transvenous ICD with confirmed lead position (R-wave ≥5–7 mV) — SIMPLE trial (n=2,500 RCT; HR 0.86; P noninferiority <0.001); NORDIC-ICD (n=1,077; 3.0% noninferior first shock efficacy)
• DT still required: S-ICD (no safety data without DT); when lead position or system integrity is uncertain; HCM/channelopathies/right-sided implant at discretion
• DT contraindicated: intracardiac thrombus; AF without ≥3 weeks therapeutic anticoagulation; haemodynamic instability
• Anodal shock polarity preferred (cathodal polarity OR 2.37 for shock failure in MODALITY registry)

Potassium Level Optimization in ICD Patients — POTCAST 2025

• POTCAST trial (Jøns/Bundgaard, NEJM 2025; n=1,200; ICD patients; baseline K ≤4.3 mmol/L; 3 Danish centres; median 39.6 months; open-label; event-driven superiority RCT): Targeting high-normal plasma potassium (4.5–5.0 mmol/L) via KCl supplementation, MRA (spironolactone/eplerenone), dietary guidance, and diuretic reduction vs standard care. Primary composite (sustained VT, appropriate ICD therapy, unplanned arrhythmia/HF hospitalization, all-cause death): 22.7% vs 29.2%; HR 0.76 (95% CI 0.61–0.95; P=0.01; NNT=12.3). Appropriate ICD therapy/documented VT: HR 0.75 (0.57–0.98). Hospitalization for cardiac arrhythmia: HR 0.63 (0.42–0.93). All-cause death NS (HR 0.85). (sources/potcast-nejm-2025, rating: very high)
• Benefit consistent across ischaemic/non-ischaemic, HF/no HF, primary/secondary prevention, and MRA vs non-MRA subgroups — the potassium-raising strategy is broadly applicable to the ICD population regardless of aetiology
• Mean potassium rise was only ~0.3 mmol/L (4.01 → 4.36 mmol/L); only 41.5% reached the 4.5–5.0 mmol/L target; yet benefit was present even in those not reaching the target — the mechanism is primarily avoidance of low-normal potassium rather than achievement of high-normal levels
• Intervention is low-cost and widely available; 75% adherence through trial end; eGFR ≥30 required to mitigate hyperkalemia risk (sources/potcast-nejm-2025, rating: very high)

Contradictions / Open Questions

• Defibrillation testing non-mandatory but not harmful — when to still test: SIMPLE (n=2,500; noninferior without DT; HR 0.86; P noninferiority <0.001) and NORDIC-ICD established that omission is safe for standard left-sided transvenous implants with confirmed lead integrity. However, both trials were predominantly ischaemic/NICM populations with left-sided implants using a single manufacturer. Data for HCM, channelopathies (LQTS, CPVT, Brugada), right-sided implants, and generator replacements with uncertain lead function are limited — DT remains reasonable in these scenarios. The SIMPLE second safety composite trended toward harm with DT (3.2% vs 4.5%; P=0.08) — not significant but biologically plausible (VF induction → myocardial injury, haemodynamic compromise). (sources/icd-programming-hrs-2015, rating: high)
• ICD programming paradigm shift: prolonged detection vs rapid therapy: The 2015 consensus reversed the longstanding culture of fastest-possible detection, demonstrating that 30/40-interval prolonged detection and VF cutoff ≥200 bpm reduce inappropriate therapies by 50–76% and may improve survival (PROVIDE HR 0.70; MADIT-RIT mortality 3.2% vs 6.6%). However, secondary prevention data are limited (25% of ADVANCE III); patients with haemodynamically unstable slow VT may not tolerate prolonged detection; and the optimal strategy has not been tested in channelopathy-specific populations where arrhythmias may behave differently. The 2015 document preceded widespread remote monitoring that now facilitates real-time programming adjustment. (sources/icd-programming-hrs-2015, rating: high)
• MADIT-II: trial stopped early and pre-GDMT era limit modern applicability. MADIT-II was stopped after a mean 20-month follow-up when the efficacy boundary was reached (P=0.027) — early termination typically overestimates effect size. The trial enrolled 1997–2001: no SGLT2i, no ARNi, no CRT-D; modern quadruple GDMT independently reduces arrhythmic substrate and improves LVEF, meaning the background SCD risk and residual NNT in contemporary practice may differ from the trial population. The entry LVEF threshold was ≤30% — current Class I guidelines use ≤35%, extrapolating beyond direct MADIT-II evidence (partially supported by SCD-HeFT). The higher HF hospitalisation trend in the ICD arm (19.9% vs 14.9%; P=0.09) raises the question of whether preventing SCD shifts the mode of death to progressive HF without changing overall healthcare burden. (sources/icd-icm-maditii-nejm-2002, sources/icd-hfref-scdheft-nejm-2005, rating: very high)
• DEFINITE: primary endpoint not significant; competing non-arrhythmic mortality attenuates all-cause benefit: DEFINITE (n=458; NICM; LVEF <36%; mean 29 months) achieved a statistically significant reduction in sudden arrhythmic death (HR 0.20; P=0.006) but not all-cause mortality (HR 0.65; P=0.08) — the trial's primary endpoint. The reason is mechanistic: the trial was designed assuming >50% of deaths would be arrhythmic, but only ~1/3 were, because high background ACEi + beta-blocker compliance (85% each) substantially reduced background SCD. The remaining deaths were HF-progressive and non-cardiac (cancer, pneumonia, stroke), against which ICD provides no protection. DEFINITE was also powered for only 68 events — too few once the arrhythmic fraction was lower than expected. Investigators noted ICD "should be considered on a case-by-case basis" rather than routinely recommended — a conclusion that mirrors the later DANISH trial (HR 0.87; P=0.28). The persistent SCD reduction (80% relative) confirms that ICD eliminates arrhythmic deaths even when all-cause mortality benefit is underpowered to reach significance. (sources/icd-nicm-definite-nejm-2004, rating: high)
• DANISH trial — does benefit still apply with modern GDMT? DANISH (2016) showed no significant reduction in all-cause mortality with ICD in NICM (HR 0.87; 95% CI 0.68–1.12; P=0.28) despite a significant SCD reduction (HR 0.50; P=0.005). The neutral primary endpoint reflects a 31% competing non-CV mortality burden and a pre-SGLT2i/ARNi GDMT era (enrolled 2008–2014). Modern quadruple GDMT (including SGLT2i and ARNi) independently improves LVEF and may reduce arrhythmic substrate, potentially narrowing the residual ICD benefit further. Benefit was significant only in patients aged <68 years (HR 0.64; P=0.01); the ≥68 subgroup showed no benefit (HR 1.05). (sources/icd-nicm-danish-nejm-2016, sources/icd-crt-auc-2025, rating: high)
• LGE-guided ICD in NICM (LVEF 36–49%): Growing observational evidence that LGE predicts SCD risk above and beyond LVEF, but no RCT has tested LGE-guided ICD implantation as a primary endpoint. AUC rates as M(5–6) only. The PARTITA trial (LGE + EPS-guided ICD in LVEF >35%) and ESTIMATE registry data support this but have methodological limitations.
• Age ≥85 ICD: The AUC rates this May Be Appropriate, but NNT analysis and patient preference data are limited. Deactivation wishes not adequately captured in AUC scenarios.
• Wearable ICD (LifeVest): Not covered in the AUC; indicated as bridge during waiting periods (40-day post-MI, post-revascularisation) but RCT evidence (VEST trial) showed no significant benefit on arrhythmic death.
• Diastolic dysfunction grading as a complementary primary prevention criterion (LVEF >35%): A prospective pilot study (Pezawas et al., Circ EP 2020; n=210; 7-year follow-up) found grade III diastolic dysfunction (restrictive pattern) independently predicted arrhythmic death/RCA (HR 3.52; 95% CI 2.00–6.22; P<0.001) regardless of LVEF. In the LVEF >35% subgroup, grade III patients had 41% cumulative AD/RCA risk at 8 years — comparable to patients who do qualify for primary prevention ICD under current LVEF ≤35% criteria. This suggests a high-risk subgroup currently ineligible for ICD. Data are pilot-level only (single centre, pre-2016 ASE diastolic criteria); not yet incorporated into guidelines. Prospective randomised validation is required. (sources/arrhythmia-diastolic-circep-2020, rating: medium)

Connections

• Related to entities/CRT — combined as CRT-D for HFrEF with broad QRS

• Related to entities/S-ICD — alternative when transvenous leads contraindicated

• Related to entities/Heart-Failure — primary prevention in HFrEF; CCM fills narrow-QRS gap

• Related to entities/HCM — A(8) with ≥1 SCD risk factor

• Related to entities/ARVC — A(9) secondary prevention; A(7) primary prevention with risk factors

• Related to entities/LMNA — A(7–8) at LVEF 35–45%

• Related to concepts/VA-Risk-Stratification-DCM — LGE + LVEF 36–49% in NICM

• Related to concepts/Sudden-Cardiac-Death — disease-specific thresholds

• Related to concepts/HCM-Risk-SCD — risk factor model underpinning A(8)

• Related to concepts/Arrhythmogenic-Cardiomyopathy — ARVC ICD ratings

• Related to concepts/LV-Diastolic-Function — grade III diastolic dysfunction as candidate additional arrhythmic risk criterion

• Related to concepts/Potassium-and-Ventricular-Arrhythmias — POTCAST: targeting high-normal K reduces appropriate ICD therapy and arrhythmia hospitalization (HR 0.76 composite; NNT=12.3)

Sources

• sources/icd-crt-auc-2025 — ACC/AHA/HRS 2025 AUC (primary source)
• sources/crt-companion-nejm-2004 — COMPANION RCT: CRT-D HR 0.64 (P=0.003) all-cause mortality in NYHA III/IV + QRS ≥120ms; non-ischaemic CRT-D HR 0.50 (P=0.015); first significant CRT-D mortality RCT
• sources/icd-icm-maditii-nejm-2002 — MADIT-II RCT: foundational primary prevention ICD in ICM; HR 0.69; basis for AUC A(8–9)
• sources/icd-nicm-danish-nejm-2016 — DANISH RCT: primary prevention ICD in NICM; neutral primary endpoint; SCD halved
• sources/icd-nicm-definite-nejm-2004 — DEFINITE RCT: first NICM-specific primary prevention ICD trial; all-cause mortality NS (HR 0.65; P=0.08); SCD HR 0.20 (P=0.006); underpowered; competing mortality pattern
• sources/arrhythmia-diastolic-circep-2020 — diastolic function and arrhythmic death risk (pilot data)
• sources/icd-hfref-scdheft-nejm-2005 — SCD-HeFT RCT: ICD HR 0.77 P=0.007; LVEF ≤35% threshold origin; aetiology-agnostic (no interaction P=0.68); NYHA II HR 0.54 vs NYHA III HR 1.16 NS
• sources/icd-programming-hrs-2015 — HRS/EHRA/APHRS/SOLAECE 2015 consensus; 32 programming recommendations; prolonged detection; ATP first-line; DT omission safe for standard transvenous implants (SIMPLE/NORDIC-ICD)
• sources/potcast-nejm-2025 — POTCAST RCT: potassium targeting reduces composite arrhythmia/death endpoint in ICD patients (HR 0.76; P=0.01; NNT=12.3)