INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Septic shock is the commonest cause of death in the intensive care unit (1). Death is attributed to refractory hypotension or progressive multiple organ failures. In the common feature, unresponsive hypotension is associated with dramatically decreased systemic vascular resistance. Sometimes, myocardial dysfunction contributes further to the circulatory failure. In septic shock, the cardiovascular system is unable to compensate the hypotensive stress. The normal compensatory response to hypotension includes an immediate baroreceptor receptor sensing of the hypotension which initiates an autonomic response; namely, sympathetic outflow to both heart and peripheral vessels is increased and would serve to restore blood pressure to normal. Experimental models of septic shock suggested that the baroreceptor mechanism which controls the efferent peripheral sympathetic activity is compromised (2). Human studies have convincingly shown that sepsis is characterized by an increased sympathetic outflow (5, 6). However, these observations provide no information on autonomic control of the cardiovascular system.

Power spectrum analysis of periodic heart rate (HR) and blood pressure fluctuations provides an indirect but noninvasive probe of autonomic cardiovascular control in a wide range of physiological perturbations (7). The method of spectral analysis is used for the detection and quantitative description of periodicities in cardiovascular fluctuations by decomposition of the original time series into sinusoidal functions of different frequencies. Three major components are usually considered: one at respiration frequency (high-frequency [HF], around 250 MHz), another around 100 MHz (low-frequency [LF]), and a third in a very-low-frequency region ([VLF], around 30 MHz) (11). In HR spectra, the HF component is linked to respiration and mediated predominantly by cardiac vagal activity, whereas the LF and VLF components are mediated by sympathetic, parasympathetic, and renin-angiotensin system activity (7). Though there is no definite evidence that the autonomic nervous system exclusively affects the ratio LF_HR/HF_HR of HR variability, this ratio is commonly used as an index of the balance between sympathetic and parasympathetic modulation of the sinoatrial node (7, 12). In addition the ratio LF_HR/HF_HR strongly correlated with direct measurement of sympathetic efferent activity (12). In blood pressure spectra, HF is mainly related to mechanical effects of respiration on cardiovascular control (13), and LF to sympathetic modulation (8). This approach of autonomic cardiovascular control has been previously used in experimental models of sepsis (14), experimental human endotoxemia (15), and in septic patients (16). The results of these studies suggested an impaired sympathetic modulation on heart and vessels.

Using spectral components of cardiovascular variability, we evaluated the sympathovagal interaction at the onset of septic shock in comparison to patients with sepsis syndrome and upright tilted healthy subjects.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

The protocol was approved by our institutional review board. Informed consent was obtained from control subjects and informed assent from patients' closest relatives.

Study Populations

Patients. We studied 10 consecutive adult patients with sepsis syndrome (developed for less than 48 h) and 10 consecutive adult patients seen at the early stage of septic shock (developed for less than 4 h).

The definition for sepsis syndrome was clinical evidence for a site of infection and two or more of the following criteria: (1) oliguria below 0.5 ml/kg for at least 1 h, (2) heart rate above 90 beats/min, (3) core temperature above 38.3° C or below 35.6° C, (4) need for mechanical ventilation, (5) arterial lactates above 2 mmol/L.

The definition for septic shock was sepsis syndrome plus sustained (at least 1 h) decrease in systemic mean blood pressure below 65 mm Hg after fluid replacement with hydroxyethyl starch with a pulmonary capillary wedge pressure between 10 and 18 mm Hg as a target goal.

The noninclusion criteria included: (1) age below 18 yr; (2) nonsinus rhythm; (3) pregnancy; (4) acute myocardial infarction; (5) pulmonary embolism; (6) past medical history of chronic cardiovascular, pulmonary, or neurological diseases; (7) diabetes; (8) any other condition that may have impaired the autonomic control of the cardiovascular system (9).

Control subjects. Six nonsmoker subjects (3 males, 43 ± 11 yr) volunteered to take part in the study. The subjects were deemed healthy on the basis of thorough medical examination, 12-lead electrocardiogram, and routine laboratory tests. They were free from any medication and were not allowed to drink alcoholic beverages or caffeine-containing products from 24 h prior to or during the entire period of the trial.

Investigated Parameters

Power Spectrum Analysis

Device. Finger arterial blood pressure and pulse intervals (transformed in HR beats automatically) were measured using a Finapres device (model 2300; Ohmeda, Trappes, France), according to the photoelectric method for the measurement of blood pressure in the finger (19). The analog output from the Finapres device was connected to an analog-to-digital converter (AS1; Notocord Systems, Igny, France) to allow data acquisition, storage, and analysis using a microcomputer (PC, Onyx DE433). In all subjects, a plethysmographic cuff including an infrared detector was fitted to the middle phalanx of the third finger, which was kept at the heart level.

Signal processing and power spectral analysis of HR and blood pressure recordings. The data acquisition and analysis methods have been previously described (20). Briefly, two 5-min recordings were performed. All records were edited and visually checked. All nonstationary signals (artifacts or ectopic beats) were removed and the resulting missing data (0.15 ± 0.10%, per record) replaced by interpolating between the 3 preceding and the 3 succeeding intervals. The blood pressure signal was digitized using a 12-bit analog-to-digital converter at a rate of 500 Hz and processed by an algorithm based on feature extraction to detect and measure the characteristics of an arterial pressure cycle with its maximum in a 1-s window (Anapres 3.0; Notocord Systems, Croissy/Seine, France). The systolic blood pressure (SBP), the preceding diastolic blood pressure (DBP), integrated mean blood pressure and HR, calculated as 60,000 divided by heart period in milliseconds were stored. In this study, pulse interval and pulse rate (reciprocal of pulse interval) were taken as surrogate for heart period and HR, respectively. The evenly spaced sampling allowed direct spectral analysis of each distribution using a fast Fourier transform algorithm on a 256-point stationary time series, corresponding to a period of 4 min 16 s at a 1-Hz sampling rate. Thus, each spectral component (band) corresponded to a harmonic of 1,000/256 MHz, i.e., 3.91 MHz. The first band (0 to 3.91 MHz) corresponded to baseline. The frequency of oscillation scale was analyzed up to 500 MHz. All spectra were analyzed as power spectra (i.e., amplitude squared) for HR and blood pressure and expressed in beats/min²/Hz and mm Hg²/Hz, respectively. The integration of the values of consecutive bands was computed to estimate the various components of the variability. The total area under the curve (AUC) corresponded to the overall variability and was obtained by the integration of all spectral bands from 20 to 500 MHz. The LF and HF components were obtained from the integration of spectral bands from 60 to 140 MHz and 200 to 350 MHz, respectively. The mean of each variable, the AUC, and the absolute and relative (LFnu and HFnu, absolute value multiplied by 100 and divided by the AUC, in percent) spectral density estimates in LF and HF bands for blood pressure and HR were calculated for control subjects and patients. The ratio between LF and HF component from HR spectra (LF_HR/HF_HR) was used as an index of sympathovagal balance. The square root of the ratio between LF component from HR and systolic blood pressure (termed  coefficient, beats/ min/mm Hg) was taken as an index of the baroreflex sensitivity (21).

Biological parameters. Blood lactates (mmol/L) were determined using standard technique. Plasma renin activity (ng/L/min) and plasma aldosterone (ng/100 ml) were determined by radioimmunoassay (22, 23) and plasma norepinephrine (pg/ml) and epinephrine (pg/ml) by high-performance liquid chromatography (HPLC) (24). For nitrates/ nitrites measurements (µM), plasma samples were deproteinized by adding 1/20th volume of zinc sulfate (300 g/L). Subsequent to centrifugation at 5,000 g for 20 min, 5 µl supernatant was analyzed by using the nitrates/nitrites assay (Boehringer Mannheim, Germany) as recommended by the manufacturer. This colorimetric method was adapted on a Progress analyzer (Konelab SA, Lisses, France).

Experimental Protocol

Patients were studied at bedside after fluid replacement, and before sedation and catecholamines infusion. All patients were mechanically ventilated in the assist-control mode with a respiratory rate set at 18 cycles per minute on the ventilator. Blood samples were drawn from central veins. Thereafter, spectral analysis of HR and blood pressure were determined.

Control subjects were studied in a quiet room at 21° C. The volunteer arrived in the laboratory 1 h after a light breakfast and immediately laid recumbent (on a tilt table) to normalize all parameters. An indwelling catheter with a heparin lock was inserted into a forearm vein in the right arm, and recording transducers were attached for an adaptation period of 30 min. The subjects were instructed to close their eyes, relax, breathe at approximately 18 cycles per minute, and not to cough, sigh, move, or sleep during the measurements. After a 1-h rest period, blood samples were drawn and spectral analysis of HR and blood pressure were obtained before and after a 30-min orthostatic tilt up to 70°.

Statistical Analysis

All data were expressed as means ± SD. Quantitative parameters were compared within control subjects before and after tilt testing by paired Student's t tests. Quantitative parameters were compared between patients with septic shock, patients with sepsis syndrome, and lying normal control subjects in a first analysis of variance (ANOVA) (with Bonferroni's adjustments for multiple comparisons). Then, quantitative parameters were compared between septic shock patients, sepsis syndrome patients, and tilted normal control subjects in a second ANOVA (with Bonferroni's adjustments for multiple comparisons). For categorical variables chi-square statistics were used. The statistical power of significance was 0.05.

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Description of the Population

There was no significant difference for age and sex ratio between patients with septic shock, patients with sepsis syndrome, and healthy volunteers. The patients had no serious past medical history as assessed by the MacCabe and Knaus scores (Table 1). As expected, compared with sepsis syndrome, in septic shock, mean blood pressure was lower and organ system failure score, Simplified Acute Physiology Score (SAPS) II, HR, and arterial lactate levels were higher. Septic shock patients have high cardiac output, pulmonary capillary wedge pressure between 10 and 18 mm Hg after fluid replacement (30 ± 10 ml/kg of hydroxyethyl starch), severe hypoxemia, and bilateral alveolar syndrome at the chest X-ray (Table 2). Two patients with shock, although with no previous cardiac history have low cardiac index (i.e., 2.7 L/min/m²). The two of them died and necropsy studies revealed normal large coronary arteries, dilation of the left ventricle, and no sign of myocardial infarction.

The site of infection was the lung, abdominal and soft tissues in 12, five, and three patients, respectively. Gram-negative bacteria accounted for infection in 10 cases, and Staphylococcus aureus and -hemolytics streptococci in, respectively, seven and three cases. None of the patients with sepsis syndrome deteriorated to shock. Two patients with sepsis syndrome died from refractory hypoxemia and five patients with septic shock died from refractory hypotension (three cases) or multiple organ failure (two cases). The survival rate was 80% in sepsis syndrome and 50% in septic shock.

Spectral Analysis of HR and Blood Pressure Variability

Normal control subjects before and after upright tilt testing. Compared with lying position, during upright tilt, overall variability of HR, SBP, and DBP significantly increased (p = 0.05, p = 0.02, and p = 0.02, respectively). Simultaneously, LF_HR and LFnu_HR increased (p = 0.02 and p = 0.01, respectively), HF_HR and HFnu_HR decreased (p = 0.008 and p = 0.002, respectively), with a subsequent increase in the ratio LF_HR/ HF_HR (p = 0.01). As compared with lying position, during upright tilt, LF_SBP and LFnu_SBP increased significantly (p = 0.03 and p = 0.04, respectively), as well as LF_DBP and LFnu_DBP (p = 0.02 and p = 0.03, respectively). Finally,  coefficient increased slightly (p = 0.09) (Table 3; see also Figure 1).

View larger version (18K): [in this window] [in a new window]   Figure 1.   Original spectra from one control subject in the lying position (left panels) and after 30 min in the upright position (middle panels) and from one patient with septic shock (right panels). SBP = systolic blood pressure; DBP = diastolic blood pressure; HR = heart rate; PSD = power spectral density. Patient's overall variability of HR, SBP, and DBP is dramatically reduced when compared with that of the control subject in the lying or in the upright positions.

Comparisons between patients with septic shock, sepsis syndrome, and lying normal control subjects. A significant difference was observed between the three groups for LFnu_HR (p = 0.04), AUC_DBP (p = 0.01), and LF_DBP (p = 0.01). AUC_DBP and LF_DBP were lower in septic shock than in sepsis syndrome (p = 0.01 and p = 0.06, respectively) or normal control subjects (p = 0.05 and p = 0.05, respectively). In addition, AUC_HR, LF_HR, LFnu_HR, the ratio LF_HR/HF_HR, and  coefficient were lower in septic shock than in sepsis syndrome, although the differences did not reach statistical significance. Finally, LFnu_HR was higher in patients with sepsis syndrome than in lying normal control subjects (p = 0.03) (Table 3).

Comparisons between patients with septic shock, sepsis syndrome, and upright tilted normal control subjects. A significant difference was observed between the three groups for AUC_HR (p = 0.007), LF_HR (p = 0.002), LFnu_HR (p = 0.05), HFnu_HR (p = 0.01), and the ratio LF_HR/HF_HR (p = 0.0002). AUC_HR, LF_HR, LFnu_HR, and the ratio LF_HR/HF_HR were lower in patients with septic shock than in tilted normal control subjects (p = 0.01, p = 0.01, p = 0.03, and p = 0.0005, respectively) or in patients with sepsis syndrome, although the difference failed to reach statistical significance. AUC_HR, LF_HR, and the ratio LF_HR/HF_HR were significantly lower in patients with sepsis syndrome than in tilted normal control subjects (p = 0.04, p = 0.01, and p = 0.001, respectively). Finally, HFnu_HR tended to be higher in patients with septic shock and sepsis syndrome than in tilted normal control subjects (Table 3).

No significant difference could be seen between the three groups for the spectral components of SBP (Table 3).

A significant difference was observed between the three groups for  coefficient (p = 0.003). The latter was lower in patients with septic shock and sepsis syndrome than in tilted normal control subjects (p = 0.01 and p = 0.05, respectively). Finally,  coefficient was lower in septic shock than in sepsis syndrome, although the difference failed to reach statistical significance (Table 3).

A significant difference was observed between the three groups for AUC_DBP (p = 0.003), LF_DBP (p = 0.001), and LFnu_DBP (p = 0.008). AUC_DBP was lower in patients with septic shock than in sepsis syndrome patients (p = 0.01) or tilted normal control subjects (p = 0.005). LF_DBP and LFnu_DBP were lower in patients with septic shock (p = 0.001 and p = 0.01, respectively) and sepsis syndrome (p = 0.01 and p = 0.01, respectively) than in tilted normal control subjects (Table 3).

Neurohormones

Controls before and after upright tilt testing. Compared with lying position, during upright tilt, plasma levels of norepinephrine, epinephrine, and aldosterone and plasma renin activity increased significantly (respectively, +553%, p = 10⁴; +227%, p = 10⁴; +2,037%, p = 4 × 10⁴; and +1,152%, p = 5 × 10⁵) (Figure 2).

View larger version (23K): [in this window] [in a new window]   Figure 2.   Results for circulating neurohormones and nitrates/nitrites in lying and tilted normal volunteers, patients with sepsis syndrome and patients with septic shock. Nitrates/nitrites measurements were not performed in tilted volunteers. The symbols indicate statistically significant differences (ANOVA plus Bonferroni's adjustments) for comparisons between the lying and the upright position in healthy volunteers (*), lying healthy volunteers and patients with sepsis syndrome () or septic shock (), tilted healthy volunteers and patients with sepsis syndrome (§), and patients with sepsis syndrome and septic shock (¶).

Comparisons between patients with septic shock, sepsis syndrome, and lying normal control subjects. A significant difference between the three groups was observed for all neurohormones and for nitrates/nitrites levels. Compared with lying control subjects, in septic shock and in sepsis syndrome, plasma levels of norepinephrine, epinephrine, aldosterone, and nitrates/ nitrites, and plasma renin activity were increased (Figure 2).

Comparisons between patients with septic shock, sepsis syndrome, and upright tilted normal control subjects. A significant difference between the three groups was observed for plasma levels of norepinephrine and aldosterone (p = 0.004 and p = 0.007, respectively). Namely, plasma levels of norepinephrine and aldosterone were significantly lower in patients with sepsis syndrome (492 ± 280 and 42 ± 51 pg/ml, respectively) than in tilted normal control subjects (1,268 ± 296 pg/ml, p = 0.01, and 175 ± 42 pg/ml, p = 0.01, respectively) or in patients with septic shock (878 ± 437 pg/ml, p = 0.05 and 127 ± 108 pg/ml, p = 0.05, respectively). In patients with septic shock, plasma levels of norepinephrine, epinephrine, and aldosterone and plasma renin activity were not significantly different than in tilted control subjects (Figure 2).

Correlation between Power Spectrum Components, Hemodynamics, Plasma Nitrates/Nitrites, and Neurohormones Levels in Septic Shock

In septic shock, LF_HR/HF_HR, , and LF_DBP correlated with mean blood pressure (r = 0.67, p = 0.04, r = 0.64, p = 0.03, and r = 0.88, p = 0.0008, respectively), and with plasma norepinephrine levels (r = 0.65, p = 0.03, r = 0.79, p = 0.006, and r = 0.69, p = 0.03, respectively). In addition, LF_DBP correlated with nitrates/nitrites (r = 0.66, p = 0.04).

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

This study shows that the instantaneous interaction between sympathetic and vagal activity is altered in the early course of severe sepsis because of a decrease in sympathetic activity, and that the impaired autonomic control to heart and vessels may contribute to circulatory failure.

We studied two populations of sepsis, i.e., patients with sepsis syndrome and patients seen at the early phase of septic shock, when the autonomic control of the cardiovascular system is expected to be highly solicited. In this context, several factors may interfere with HR or blood pressure fluctuations. First, a reduction of venous return to the thorax may alter the respiratory fluctuation in SBP (25, 26). Therefore, patients were studied after fluid replacement with pulmonary wedge pressure between 10 and 18 mm Hg as a target goal. The absence of difference in the high-frequency peak (central frequency and amplitude) from SBP spectra makes likely that preload was comparable between the three groups (26). Second, in healthy volunteers, the infusion of catecholamine may alter the low-frequency fluctuations of blood pressure (27) and has no significant effect on the frequency components of HR (27, 28). In critically ill patients, previous studies suggested that catecholamine infusions did not play a determinant role in spectral analysis changes (16). In our experiments, short-term recordings of HR and blood pressure were obtained before infusion of catecholamines. Third, all patients were mechanically ventilated, whereas healthy subjects were breathing spontaneously. The design of our experiments allowed us to ensure comparable respiratory rates in patients and in control subjects. However, a forcing input like mechanical ventilation may have altered, per se, the spectral profile. Finally, none of the patients received anesthetic drugs that are known to reduce HR variability (29). Subsequently, patients may have been more anxious than healthy subjects. The two latter factors may limit the comparisons between control subjects and patients but not comparisons between sepsis syndrome and septic shock.

The noninvasive beat-to-beat finger blood pressure measurement using the Finapres device has been shown to be reliable as compared with intra-arterial blood pressure monitoring in various conditions in humans (30) including septic shock (31). In four of the 10 septic shock patients, concomitant intra-arterial (through a radial line) measures of SBP strongly correlated with the measures recorded with the Finapres device (data not shown).

In this experiment, cardiovascular variability has been studied using frequency-domain measurements obtained from short-term recordings, allowing the selection of periods of steady state with stable hemodynamic status and sinus rhythm. Time-domain measures are moment statistics and provide no insight into the processes influencing the periodicity or regularity of the heart or blood pressure. Over the short term, frequency-domain measurements are remarkably reproducible (32) and more closely linked to direct measurements of sympathetic efferent activity than time-domain measurements (12). In this condition, power spectral analysis of cardiovascular variability provides indices of neural regulation, and the reciprocal relationship between LF and HF from HR signal may assess the sympathovagal balance (7, 12).

Our findings in healthy volunteers were consistent with previous reports (7, 8, 10, 32). In HR and blood pressure variability signals, at rest, two major oscillatory components, i.e., LF and HF, were observed. The upright position was accompanied by elevations in LF component, reflecting enhanced sympathetic drive.

Considerable controversy exists regarding autonomic control of the cardiovascular system in sepsis. Animal studies have reported inappropriate decrease in the peripheral sympathetic outflow induced by intravenous endotoxin (3, 4). Conversely, in different animal models, increased sympathetic activity following endotoxin has been shown (33). Septic shock is usually associated with marked tachycardia, increased circulating catecholamines (5, 6), and norepinephrine spillover (6). In contrast, the response of the heart (34) and vessels (31) to sympathetic activation may be impaired. The discrepancies between the results of these studies may be accounted for by many factors, including: species, use of anesthesia, severity of illness (i.e., sepsis syndrome or septic shock), and type of shock (i.e., endotoxin shock or septic shock). Moreover, these observations may not rule out a failure of the adrenergic neuroeffector transmission.

Using cardiovascular response to orthostatic tilt in healthy subjects as a standard of sympathetic activation, we found that, in human severe sepsis, overall variability and LF component (in normalized units) of HR and DBP, and the ratio LF_HR/HF_HR were altered in relation to the severity of the disease. These data suggest that an impaired sympathetic efferent activity to heart and vessels may accompany the onset of septic shock. Nevertheless, our experiments may not rule out a very initial and transient burst of sympathetic activity.

The cause of blunted LF variability of HR and blood pressure may involve multiple factors. The  coefficient which provides an accurate estimation of the arterial baroreflex (21) was decreased in patients compared with normal control subjects, and in septic shock compared with sepsis syndrome. These results suggested that impaired baroreflex circulatory regulation may contribute to the observed autonomic disturbances. Furthermore, in septic shock,  coefficient correlated with mean blood pressure. We found that both the ratio LF_HR/ HF_HR and LF_DBP were inversely related to plasma norepinephrine levels. Thus, we may speculate that, as in severe heart failure (35), high sympathetic drive may lead to saturation of LF oscillatory systems or that excessive concentrations of circulating catecholamines may compromise central autonomic controls. Finally, the observed inverse relationship between LF_DBP and plasma levels of nitrates/nitrites may indicate that blunted LF variability of blood pressure reflects nitric oxide-induced low vascular responsiveness to catecholamines.

In conclusion, we have shown that, in septic shock, LF component of the oscillations in HR and DBP variability is dramatically reduced. This decrease in LF component may identify septic patients with high level of sympathetic activation. In addition, as in severe heart failure, in septic shock, central autonomic regulatory impairment may account for the discrepancy between high sympathetic drive and low LF component.