Patients
After approval by our Institutional Review Board (Comité pour la Protection des Personnes île-de-France VII), we enrolled 80 patients who had circulatory failure of septic origin in our medical intensive care unit. Acute circulatory failure was defined by the presence of one or more of the following signs: (i) systolic arterial pressure ≤90 mm Hg (or decrease in systolic arterial pressure ≥50 mm Hg in known hypertensive patients); (ii) urinary flow ≤0.5 ml/kg/hr for more than 2 hours; (iii) tachycardia ≥100 beats/min; and (iv) skin mottling [
25,
26]. Patients' relatives were informed about the study at the time the patient was included. They were given a choice to refuse the patient's participation at that time. If not, patients were informed as soon as their mental status allowed, and they were given the choice to withdraw their participation in the study.
In 40 of these patients (Group 1), the attending physician decided to administer fluid because the patient showed some criteria predicting fluid responsiveness: increase in cardiac index (CI) ≥10% during a passive leg-raising test [
27]; increase in cardiac index (CI) ≥5% during an end-expiratory occlusion [
28]; or, in cases with no spontaneous triggering of the ventilator and no cardiac arrhythmias, a pulse-pressure variation ≥13% [
29]. In 40 different patients (Group 2), the attending physician had decided to introduce norepinephrine or to increase its dosage because the mean arterial pressure was <65 mm Hg (or 75 mm Hg in known hypertensive patients) [
30] and the diastolic arterial pressure was low [
31].
All patients had a catheter inserted into the internal jugular vein and a catheter inserted into the femoral artery (Pulsiocath for thermodilution; Pulsion Medical Systems, Munich, Germany). The arterial line was divided in two branches, one connected to a PiCCOplus device (PiCCOplus v6.0; Pulsion Medical Systems, Munich, Germany), and the other one connected to a FloTrac/Vigileo device (FloTrac/Vigileo v1.10; Edwards Lifesciences, Irvine, CA). This enabled the two devices simultaneously to analyze the same sample of the arterial pressure curve.
Interventions and measurements
Before each therapeutic intervention (that is, before administering fluid infusion in Group 1 and before introducing/increasing norepinephrine in Group 2), we performed a first set of hemodynamic measurements, including heart rate, systemic arterial pressure, CItd, CIpc, CIpw, and systemic vascular resistance (SVR). We used the values of CIpc and CIpw that were automatically displayed on the screens of the commercial devices. The CIpc and CIpw displayed by the monitors are averaged over a 12-second and 20-second rolling period, respectively. We could not obtain the digital (unprocessed, unfiltered, and nonaveraged) data from the devices. In practice, the CIpc and CIpw values displayed by the monitors were averaged over 10-second periods. The CItd was measured by the PiCCOPlus device by injecting 15 ml of iced saline (<8°C) through the central venous line. The injection was manually performed in triplicate, and the values of CItd were averaged. Immediately after performing thermodilution boluses, the values of CIpc were measured, the arterial line being physically shared in Y through taps toward the Vigileo and the PiCCO devices. Because the thermodilution automatically calibrated the pulse-contour analysis of the PiCCO device, CIpc was identical to CItd at the starting time. The CIpw was carried immediately before thermodilution to avoid interference between the temperature drift and the accuracy of the CIpw. The total SVR was estimated as mean arterial pressure divided by CItd.
After the first set of hemodynamic measurements was completed in Group 1, fluid loading was performed by infusing 500 ml of saline over a 30-minute period. In Group 2, norepinephrine was titrated, targeting a mean arterial pressure of 65 mm Hg [
30] (or 75 mm Hg in previously hypertensive patients). All other treatments were kept unchanged during the therapeutic intervention. In particular, the dosage of norepinephrine was kept constant in patients of Group 1 who already received norepinephrine, and volume expansion was not administered to the patients of Group 2 during the study period.
A second set of hemodynamic measurements was carried out again after the therapeutic intervention (that is, at the end of fluid administration in Group 1 and 5 minutes after stabilization of mean arterial pressure was obtained in Group 2). This set included heart rate, systemic arterial pressure, CIpc, CIpw, CItd, and SVR. The values of CIpc and of CIpw were recorded before the CItd. Therefore, the value of CIpc was not automatically calibrated by the thermodilution. For that recording, the arterial line was shared in Y through a tap turned on to both the Vigileo and the PiCCO devices.
For 10 patients of Group 1 and 10 patients of Group 2, radial and femoral arterial catheters were simultaneously in place. In these patients, hemodynamic worsening required switching from a simple radial blood-pressure monitoring to a more complete hemodynamic monitoring like transpulmonary thermodilution.
Just after the femoral catheter insertion and before the radial catheter ablation, we took the opportunity to have both catheters in place for testing whether connecting the Vigileo device to the radial or to the femoral arterial line could provide different measures of cardiac output. For this purpose, before and after the therapeutic intervention, immediately after recording the CIpw obtained from the femoral line, the Vigileo device was disconnected from the femoral line and connected to the radial line, and the CIpw obtained from the radial line was recorded.
In these same patients, we also tested whether splitting the arterial line into two branches introduced differing harmonic influences into the system that might influence the monitors. For this purpose, before and after the therapeutic intervention, the CIpc and CIpw were recorded again after turning on the tap sharing the femoral arterial line, such that the lines directed to both devices were alternatively closed.
Statistical analysis
All data were normally distributed (Kolmogorov-Smirnov test), except CItd, CIpc, CIpw, and the dosage of norepinephrine, and are expressed as median [25th to 75th percentile]. In each patient, we performed only one pair of measurements (before/after therapeutic intervention). Comparisons between values recorded before with values recorded after therapeutic interventions were performed in both groups by using a paired Student t test or a paired Wilcoxon test, as appropriate. Comparisons between Group 1 and Group 2 were performed with a two-tailed Student t test or a Mann-Whitney U test, as appropriate.
We compared the relative changes of CIpc and of CIpw with those of CItd during the therapeutic interventions by a Bland and Altman analysis (for absolute changes) and by linear regression analysis (for percentage changes). We tested the ability of CIpc and CIpw to detect an increase in CItd ≥ 15% by constructing receiving operating characteristics (ROC) curves. The area under the ROC curves (expressed as mean (95% confidence interval) were compared by using the Hanley-MacNeil test. This analysis was also separately performed in patients in whom the SVR changed (increased or decreased) by >15% with the therapeutic interventions and in patients in whom the SVR changed (increased or decreased) by <15% with the therapeutic interventions [
32]. The values recorded before therapeutic interventions were not compared, because at this time, the CIpc was, by definition, identical to the CItd value because of calibration. After the therapeutic interventions, we compared the absolute values of CIpc and of CIpw with that of CItd by a Bland and Altman analysis and calculated the percentage error as 2 SD/mean [
33].
The precision of each method was calculated from data obtained from a sample made of the first 20 and the last 20 patients included in the study when arterial pressure was stable. For CItd, we calculated the coefficient of variation (ratio of the standard deviation to the mean) for each set of three consecutive thermodilution boluses and then averaged it for the series of the 20 sets. For CIpc and CIpw, we collected the 10 consecutive values of CI displayed on the monitor. We calculated the coefficient of variation for each set and averaged it for the series of 20 sets [
32].
A P value < 0.05 was considered significant. The statistical analysis was performed by using Statview 5.0 software (Abacus Concepts, Berkeley, CA).