Bigger et al. (CAPS)^{96} | N = 331 (30 deaths) | 24-hour, 1 year after enrolling in CAPS and 1 week after stopping meds | 3 years | ULF, VLF, LF, HF all significant, univariate predictors of all-cause mortality. After adjustment for covariates, VLF was the strongest predictor |

Copie et al.^{97} | N = 579, (54 deaths, 42 cardiac, 26 sudden) | 24-hour, before discharge (median 7 days after MI) | > 2 years | HRV index better predictor than mean RR interval for sensitivity <40%. For sensitivity ≥40% mean R-R interval and HRV index equal. Mean R-R interval <700 ms predicted cardiac death (45% sensitivity, 85% specificity, 20% PPA) and predicted all-cause, cardiac and sudden death better than LVEF |

Fei et al.^{98} | N = 700 (45 cardiac deaths, 24 sudden) | 24-hour, 5-minute period, 5–8 days before discharge | 1 year | SDNN for 5-minutes had lower PPA than HRV index, but could preselect those who require 24-hour Holter ECG for risk stratification |

Huikuri et al. (DIAMOND-MI)^{74} | N = 446 with LVEF ≤ 0.35, 114 deaths, 75 arrhythmic, 28 nonarrhythmic | 24-hour, predischarge, traditional and nonlinear HRV | 685 ± 360 days | α_{1} <0.75 RR 3.0, 95% CI 2.5–4.2 for all cause mortality, independent predictor after adjustment. Predicted by arrhythmic and nonarrhythmic death |

La Rovere et al. (ATRAMI)^{99} | N = 1284 (44 cardiac deaths, 5 nonfatal sudden) | 24-hour, <28 days after MI | 21 ± 8 months | SDNN <70 ms vs SDNN ≥70 ms |

Mäkikallio et al. (TRACE)^{100} | N = 159 with LVEF ≤ 35, 72 deaths | 24-hour, traditional and nonlinear | 4 years | α_{1} <0.85 best univariate predictor of mortality (RR 3.17, 95% CI 1.96–5.15), PPA 65% and NPA 86%. Remained significant after adjustment |

Odemuyiwa et al.^{101} | N = 433 (first MI), (46 total deaths, 15 sudden deaths) | 24-hour, before discharge | 4 weeks to 5 years | HRV index <20 univariate predictor of mortality for whole follow-up but independent predictor of total cardiac mortality for first 6 months only |

Odemuyiwa et al.^{37} | N = 385 (44 deaths, 14 sudden) | 24-hour, before discharge | 151–1618 days | HRV index <39 sensitivity 75%, specificity 52% compared with LVEF ≤40% which had specificity of 40% for all-cause mortality. HRV + LVEF better specificity for sensitivity <60% |

Quintana et al.^{102} | N = 74 (18 deaths 9 nonfatal MI), 24 normal controls | 24-hour, mean 4 days after MI | 36 ± 15 months | LnVLF <5.99 independent predictor of all-cause mortality (RR = 1.9) or mortality/nonfatal infarction (RR = 2.2) |

Tapanainen et al.^{103} | N = 697, 49 deaths | 24-hour, 2–7 days after MI | 18.4 ± 6.5 months | α_{1} <0.65 most powerful predictor of mortality RR 5.05, 95% CI 2.87–8.89). After adjustment, α(1) remained independently associated with mortality (RR = 3.90, 95% CI 2.03–7.49) |

Touboul et al. (GREPI)^{104} | N = 471 (26 deaths for 1 year FU, 39 for long-term FU, 9 sudden) 45% had thrombolysis | 24-hour HRV, 10 days after MI | 1 year and long term (median 31.4 months) | Nighttime AVGNN <750 ms (RR = 3.2), daytime SDNN <100 ms (RR = 2.6) |

Viashnav et al.^{105} | N = 226 (19 cardiac deaths) | 24-hour, mean 83 hours after MI | Mean 8 months | Cox regression not performed Decreased SDNN, SDANN, ASDNN, LF, HF, LF/HF among nonsurvivors, but rMSSD and pNN50 not different |

Voss et al.^{106} | N = 572 (43 all-cause, 14 sudden arrhythmic, 22 sudden, 34 cardiac, 13 nonfatal VT/VF) | 24-hour, 5–8 days after MI, standard, nonlinear HRV | 2 years | For best combination of predictors maximum specificity at 70% sensitivity where PPA for endpoints was 16–29% compared with 6–17% for HRV alone |

Zabel et al.^{107} | N = 250 (30 endpoints) | 24-hour HRV, stable, before discharge | Mean 32 months | SDNN significantly higher in event-free (no VT, resuscitated VF, or death) |

Zuanetti et al. (GISSI)^{52} | N = 567 males treated with thrombolysis (52 total deaths, 44 cardiac deaths) | 24-hours at discharge (median 13 days) | 1000 days | Independent predictors of all-cause mortality: NN50 + (RR = 3.5), SDNN (RR = 3.0), rMSSD (RR = 2.8) |