Heart rate variability displays 1/ f noise in critical illness and correlates with severity of multiple organ dysfunction

Abstract

It has been postulated that multiple organ dysfunction (MOD) results from an uncoupling of connections between organs, which can be viewed as biological oscillators, with subsequent loss of complex' or adaptive behaviour of the body as a whole. If so, this process may be demonstrated by loss of variability in individual organs. To test this hypothesis, we measured heart rate variability (HRV) in a cross section of ICU patients, using a non-linear model based on variations occurring in many complex naturally occurring systems. These variations obey a power law (known as 1/f noise) when plotted as frequency distributions with log-log axes. The power relationship is such that aEb = f, where E is the amount of variation, f is the frequency of observations, and b and log (a) refer to the slope and intercept, respectively. We postulated that increasing MOD would be associated with loss of HRV. A 5-min, real-time ECG recording was downloaded to a personal computer at a sample frequency of 500 Hz utilizing an analogue to digital data translation board. Instantaneous heart rate was calculated for each RR interval, artefact screened and excluded using a Poincare plot, and the mean heart rate was calculated. Absolute variations from the mean were plotted on a 20-bin histogram, and from this a frequency distribution was plotted in log10-log10 space, producing a unique regression line for each measurement. Patients were grouped according to number of organs failing (0-1, 2, [>=]3) along the lines of mortality risk. A total of 109 measurements, made on 52 children, median age 7 months (range 2 days to 16 yr) were spread across organ failure groups such that: 0-1 organ failures (n = 56), 2 (n = 32), and [>=]3 (n = 21) (Figure 2). Breakdown of individual organ failures gave: respiratory (n = 67), cardiac (n = 48), haematologic (n = 19), liver (n = 16), renal (n = 13), neurologic (n = 7). Regression analysis gave an overall mean ([IMG]f1.gif" BORDER="0">) r2 of 0.87 (0.09), with no difference between organ failure groups (P = 0.46). The mean ([IMG]f1.gif" BORDER="0">) slope for all patients was -1.80 (0.29), and again was similar regardless of the degree of MOD (P = 0.24). The three groups showed significant differences for both the x- and y-intercepts, with decreasing HRV (i.e. left-shifted regression lines with decreasing intercept) seen amongst those with greatest MOD (ANOVA P<0.001 for all groups). A multiple regression model showed this to be independent of the potential confounding factors: inotrope, opiate or benzodiazepine use. (1) HRV exhibits 1/f noise, as shown by the high r2 value, and consistency of slope. (2) HRV can be quantified by change in intercept of the regression line. (3) Loss of HRV (evidenced by a left-shifted regression line) is seen with increasing MOD. (4) This effect appears to be independent of inotrope, opiate or benzodiazepine use. (5) These findings are consistent with MOD representing an uncoupling of inter-organ communication, and this model may be used as a marker for disease severity.