individuals and in those recovering from acute myocar- dial infarction'1"3'. Decreased heart rate variability, measured in the time'4"6' and frequency domain'7"9'.
European Heart Journal (1997) 18, 789-797
Heart rate variability as a means of assessing prognosis after acute myocardial infarction A 3-year follow-up study M. Quintana*, N. Storckf, L. E. Lindbladf, K. Lindvall* and M. Ericson* Karolinska Institute at the *Department of Cardiology, ^Department of Clinical Physiology, South Hospital, I Royal Institute of Technology, Stockholm, Sweden
Aims The present study evaluated the prognostic value of heart rate variability after acute myocardial infarction in comparison with other known risk factors. The cut-off points that maximized the hazards ratio were also explored. Patients and methods Heart rate variability was assessed with 24 h ambulatory electrocardiography in 74 patients with acute myocardial infarction, 4 ± 2 days after hospital admission and in 24 healthy controls. Patients were followed for 36 ±15 months.
cardiac events (defined as death, non-fatal infarction or revascularization) had a lower heart rate variability than those without (P=003 and P=003, respectively). With multivariate regression analysis, decreased heart rate variability independently predicted mortality and death or non-fatal infarction. The presence of a left ventricular ejection fraction =ns).
Clinical characteristics, pharmacological therapy and heart rate variability A description of the studied population is presented in Table 1. Both time and frequency domain measures of heart rate variability were compared with the patients' clinical characteristics and pharmacological therapy. However, only the values corresponding to the very low frequency power are shown since similar values were found with the other frequency bands. Patients with vs without diabetes mellitus, and with vs without previous myocardial infarction had lower heart rate variability. Patients with Q wave infarction had a higher heart rate variability than those with a non-Q wave infarction. Eur Heart J, Vol. 18, May 1997
Patients treated with aspirin and heparin showed higher heart rate variability than those without these drugs. The correlation coefficient between ejection fraction and VLF power was r=0-44, />=00004.
Outcomes and heart rate variability Table 2 shows measures of heart rate variability for all patients (survivors and non-survivors) as well as for the controls. All measures obtained in the frequency domain (except for HF power) were statistically lower in nonsurvivors than in survivors. This difference was most pronounced when comparing non-survivors vs controls, but was still statistically significant when comparing survivors vs controls. Measurements of heart rate variability in the time domain were significantly higher (variance, SDNN, SDANNi) in survivors vs nonsurvivors, and in non-survivors vs controls. The only measure that was significantly higher in survivors vs controls was SDNN. Table 3 summarizes the values of frequency domain heart rate variability according to the presence of different end-points. Patients suffering short-term events had lower values for VLF power than patients without short-term events. All measures of heart rate variability (except those for HF power) were lower in non-survivors than in survivors. The same applied for patients with and without reinfarctions or non-fatal reinfarctions.
Survival and multivariate regression analysis Using Cox's model, the dichotomization points that maximized the hazards ratio were found to be localized
Assessing prognosis after AMI with HRV
793
Table 3 End-points and measures of heart rate variability in the frequency domain. All measurements are expressed as natural logarithmic values (Ln) ± SD Variables
End-points
(+) n=18 Ln Ln Ln Ln
total power VLF power LF power HF power
6-5 ±1-2 5-8 ±0-9 4-9 ± 1-2 4-6 ±1-5 (+) n=16
Ln Ln Ln Ln
total power VLF power LF power HF power
6-3 ±1-1 5-7 ±0-9 4-8 ±1-2 4-6±l-3 (+)n =9
Ln Ln Ln Ln
total power VLF power LF power HF power
6-2 5-6 4-6 4-4
±11 ±10 ±11 ±1-3
( + ) n = 38 Ln Ln Ln Ln
total power VLF power LF power HF power
6-8 6-2 5-3 4-8
±0-9 ±0-9 ±11 ±1-3
(+)n=ll Ln Ln Ln Ln
total power VLF power LF power H F power
6-5 ± 1-3 5-7 ± 11 5-0 ±1-4 4-8 ± 1-6
Deaths* ( - ) n = 56
P
7-2 ±0-9 6-6 ±0-8 5-8± 11 5-1 ± 1-2 All reinfarctions* ( - ) n = 58 7-2 ±0-9 6-6 ±0-8 5-8 ± 11 51 ± 1-3
2-7 3-3 2-7 1-4
002 0005 001 ns
t
P
3-3 40 31 1-3
0006 0001 0006 ns
Non-fatal infarctions* ( ( - ) n = 65 7-2 ±0-9 6-5 ±0-8 5-7 ± 1 1 51 ± 1-3 Cardiac events* ( - ) n = 36 7-3 6-6 5-9 5-2
±0-9 ±0-9 ±11 ±1-3
Short-term events! ( - ) n = 63 7-1 ±0-9 6-5 ±0-8 5 7 ±11 5-0 ± 1-3
P
2-7 2-8 2-8 1-3
0-03 003 001 ns
;
P
20 2-3 21 1-5
004 0-02 0-03 ns
(
P
1-9 2-9 1-8 0-5
ns 004 ns ns
*End-points occurring during long-term follow-up; fshort-term events: death (n = 6), non-fatal infarction (n = 2) and revascularization procedure (n = 3) occurring within the first month after the index infarct; (+) = presence of end-point, ( —)=absence of end-point. Abbreviations as in Table 2.
as follows: the 27th percentile for total power spectrum (Ln 6-4815 or 653 ms2), the 27th percentile for VLF power (Ln 5-9915 or 400 ms2), the 32nd percentile for LF power (Ln 5-2575 or 192 ms2), and the 27th percentile for HF power (Ln 41271 or 62 ms2). Survival curves were calculated using these points. For this purpose the study population was divided in two groups (patients with values < the cut-off points and patients with values > the cut-off points), Fig. l(a)-(d)Table 4 shows the results of uni- and multivariate regression analysis. Only the variables showing a statistically significant predictive value during univariate analysis are shown. Among measures of heart rate variability, we selected that which added the best predic-
tive value during multivariate analysis (Ln VLF power). The model was also tested introducing the Ln VLF power values without dichotomization, that is, as a continuous variable, and the following beta (/?) and P values were found: /?= -0-82, />= the cutoff points. Sens=Sensitivity; Spec=Specificity.
showed lower values of heart rate variability measured in the frequency domain, with the exception of the values for the Ln HF power (Table 3). Mortality and death or non-fatal reinfarction were independently predicted by frequency domain measures of heart rate variability. A history of systemic hypertension and a low left ventricular ejection fraction (
