Traditional and Nonlinear Heart Rate Variability
Traditional and Nonlinear Heart Rate Variability
Introduction: Decreased heart rate variability (HRV) and abnormal nonlinear HRV shortly after myocardial infarction (MI) are risk factors for mortality. Traditional HRV predicts mortality in patients with a range of times post-MI, but the association of nonlinear HRV and outcome in this population is unknown.
Methods and Results: HRV was determined from 740 tapes recorded before antiarrhythmic therapy in Cardiac Arrhythmia Suppression Trial patients with ventricular premature contractions (VPCs) suppressed on the first randomized treatment. Patients were 70 ± 121 days post-MI. Follow up was 362 ± 241 days (70 deaths). The association between traditional time and frequency-domain HRV and mortality and nonlinear HRV and mortality were compared for the entire population (ALL), those without coronary artery bypass graft post-MI (no CABG), and those without CABG or diabetes (no CABG, no DIAB) using univariate and multivariate Cox regression analysis. Strength of association was compared by P values and Wald Chi-square values. Nonlinear HRV included short-term fractal scaling exponent, power law slope, and SD12 (Poincaré dimension). For ALL and for no CABG, increased daytime SD12 had the strongest association with mortality (P = 0.002 ALL and P < 0.001 no CABG). For no CABG, no DIAB increased 24-hour SD12 hours had the strongest association (P < 0.001) with mortality. Upon multivariate analysis, increased SD12, decreased ln ULF (ultra low frequency), and history of prior MI and history of congestive heart failure each remained in the model.
Conclusion: Nonlinear HRV is associated with mortality post-MI. However, as with traditional HRV, this is diluted by CABG surgery post-MI and by diabetes. Results suggest that decreased long-term HRV and increased randomness of heart rate are each independent risk factors for mortality post-MI.
Beginning in 1987 with the results of the Multi-Center Post-Infarction Project (MPIP) and numerous subsequent trials, it has been clearly demonstrated that traditional time-domain and frequency-domain indices of heart rate variability (HRV) measured in the peri-infarction period identify postmyocardial infarction (post-MI) patients at increased risk for mortality. Recently, techniques from nonlinear dynamics have been applied to peri-infarction patients. The index power law slope was applied to the MPIP data and found to be a better predictor of mortality than any time- or frequency-domain HRV index. In other studies, the short-term fractal scaling exponent was a better predictor of mortality than traditional HRV measures. Thus, limited data suggest that nonlinear HRV may provide superior risk stratification to traditional HRV post-MI.
In risk stratification studies, HRV almost always is measured in the peri-infarction period. Time- and frequency-domain HRV are known to increase in many patients during recovery from MI. The predictive value of HRV measured at a later time is less clear. The Cardiac Arrhythmia Pilot Study (CAPS), the pilot study for the Cardiac Arrhythmia Suppression Trial (CAST), reported that HRV measured in the frequency domain 1 year post-MI continues to predict mortality during an approximately 2-year follow-up.
CAST was an historic post-MI study of patients with impaired left ventricular ejection fraction and high-grade ventricular arrhythmias that were suppressed by class IC drugs. Holter recordings were obtained at a large range of times post-MI. We previously reported that time-domain indices of HRV, such as SDNN (standard deviation of all normal-to-normal interbeat intervals), which are strong predictors of mortality in the peri-infarction period, were not associated with mortality when the entire CAST group was taken together. However, after patients with diabetes and/or coronary artery bypass graft (CABG) surgery post-MI were excluded from the analysis, decreased SDNN was strongly associated with mortality (P = 0.006). Similarly, in the frequency domain, when patients with diabetes or CABG post-MI were excluded, as was the case with peri-infarction post-MI studies, decreased ln total power (P < 0.001) and decreased ln ultra low frequency power (P < 0.001) had the strongest associations with mortality.
In the current investigation, we compared the association with all-cause mortality in the CAST of 24-hour, daytime, and nighttime nonlinear HRV indices: power law slope, SD12 (sometimes called the Poincaré dimension), and short-term fractal scaling exponent, with previously computed associations for standard time- and frequency-domain HRV. Because of the confounding effect of including patients with CABG post-MI and diabetes on the association of traditional HRV and mortality, these associations were again determined for the entire dataset, excluding CABG post-MI and excluding both CABG and diabetes. Finally, the independent association of traditional and nonlinear HRV with mortality was determined.
Introduction: Decreased heart rate variability (HRV) and abnormal nonlinear HRV shortly after myocardial infarction (MI) are risk factors for mortality. Traditional HRV predicts mortality in patients with a range of times post-MI, but the association of nonlinear HRV and outcome in this population is unknown.
Methods and Results: HRV was determined from 740 tapes recorded before antiarrhythmic therapy in Cardiac Arrhythmia Suppression Trial patients with ventricular premature contractions (VPCs) suppressed on the first randomized treatment. Patients were 70 ± 121 days post-MI. Follow up was 362 ± 241 days (70 deaths). The association between traditional time and frequency-domain HRV and mortality and nonlinear HRV and mortality were compared for the entire population (ALL), those without coronary artery bypass graft post-MI (no CABG), and those without CABG or diabetes (no CABG, no DIAB) using univariate and multivariate Cox regression analysis. Strength of association was compared by P values and Wald Chi-square values. Nonlinear HRV included short-term fractal scaling exponent, power law slope, and SD12 (Poincaré dimension). For ALL and for no CABG, increased daytime SD12 had the strongest association with mortality (P = 0.002 ALL and P < 0.001 no CABG). For no CABG, no DIAB increased 24-hour SD12 hours had the strongest association (P < 0.001) with mortality. Upon multivariate analysis, increased SD12, decreased ln ULF (ultra low frequency), and history of prior MI and history of congestive heart failure each remained in the model.
Conclusion: Nonlinear HRV is associated with mortality post-MI. However, as with traditional HRV, this is diluted by CABG surgery post-MI and by diabetes. Results suggest that decreased long-term HRV and increased randomness of heart rate are each independent risk factors for mortality post-MI.
Beginning in 1987 with the results of the Multi-Center Post-Infarction Project (MPIP) and numerous subsequent trials, it has been clearly demonstrated that traditional time-domain and frequency-domain indices of heart rate variability (HRV) measured in the peri-infarction period identify postmyocardial infarction (post-MI) patients at increased risk for mortality. Recently, techniques from nonlinear dynamics have been applied to peri-infarction patients. The index power law slope was applied to the MPIP data and found to be a better predictor of mortality than any time- or frequency-domain HRV index. In other studies, the short-term fractal scaling exponent was a better predictor of mortality than traditional HRV measures. Thus, limited data suggest that nonlinear HRV may provide superior risk stratification to traditional HRV post-MI.
In risk stratification studies, HRV almost always is measured in the peri-infarction period. Time- and frequency-domain HRV are known to increase in many patients during recovery from MI. The predictive value of HRV measured at a later time is less clear. The Cardiac Arrhythmia Pilot Study (CAPS), the pilot study for the Cardiac Arrhythmia Suppression Trial (CAST), reported that HRV measured in the frequency domain 1 year post-MI continues to predict mortality during an approximately 2-year follow-up.
CAST was an historic post-MI study of patients with impaired left ventricular ejection fraction and high-grade ventricular arrhythmias that were suppressed by class IC drugs. Holter recordings were obtained at a large range of times post-MI. We previously reported that time-domain indices of HRV, such as SDNN (standard deviation of all normal-to-normal interbeat intervals), which are strong predictors of mortality in the peri-infarction period, were not associated with mortality when the entire CAST group was taken together. However, after patients with diabetes and/or coronary artery bypass graft (CABG) surgery post-MI were excluded from the analysis, decreased SDNN was strongly associated with mortality (P = 0.006). Similarly, in the frequency domain, when patients with diabetes or CABG post-MI were excluded, as was the case with peri-infarction post-MI studies, decreased ln total power (P < 0.001) and decreased ln ultra low frequency power (P < 0.001) had the strongest associations with mortality.
In the current investigation, we compared the association with all-cause mortality in the CAST of 24-hour, daytime, and nighttime nonlinear HRV indices: power law slope, SD12 (sometimes called the Poincaré dimension), and short-term fractal scaling exponent, with previously computed associations for standard time- and frequency-domain HRV. Because of the confounding effect of including patients with CABG post-MI and diabetes on the association of traditional HRV and mortality, these associations were again determined for the entire dataset, excluding CABG post-MI and excluding both CABG and diabetes. Finally, the independent association of traditional and nonlinear HRV with mortality was determined.
Source...