Introduction

Several studies have demonstrated that the standard techniques for preload monitoring, such as measuring cardiac filling pressures, often fail to give adequate information upon fluid responsiveness in mechanically ventilated patients [1]. However, specifically the decision whether a hemodynamically instable patient will respond to fluid loading with an increase in stroke volume and cardiac output determines the initial treatment strategy whether to administer fluid or to use catecholamines. Growing knowledge about the specific interactions of the cardiovascular system and the lungs under mechanical ventilation have opened up unique opportunities to determine fluid responsiveness [1, 2]. The principle of this functional preload monitoring is based on the common finding of extensive undulations in systolic arterial pressure under mechanical ventilation. The cyclic increases in intrathoracic pressure due to the mechanical breath cause specific changes in biventricular pre- and afterload. This finally leads to variations in left ventricular stroke volume and consequently in cyclic changes of systolic arterial pressure [3]. Retrospective off-line analysis of this systolic pressure variation (SPV), its delta down component, and pulse pressure variation (ΔPp) have been extensively investigated and validated to be useful predictors of fluid responsiveness [4, 5, 6]. Arterial pulse contour analysis, a technique described by Frank [7] as early as 1930, allows continuous and real-time monitoring of cardiac output and stroke volume. Further technical developments and modifications of the algorithm and the implementation of computerized analysis [8] now allow monitoring also the respiration-induced left ventricular stroke volume variation (SVV), which causes SPV and ΔPp. Although no extended experimental validation of pulse contour stroke volume and SVV has so far been published, the close agreement between pulse contour cardiac output measurement and the clinical gold standard thermodilution as well as the close relationship of pulse contour SVV and retrospective SPV suggests an adequate accuracy of this method for quantifying this hemodynamic phenomenon [8, 9]. Further, SVV has been reported to be useful in predicting volume responsiveness and monitoring the hemodynamic effects of volume resuscitation [10, 11]. However, in addition to the effective blood volume and therefore cardiac preload, the extent of these specific interactions between the lungs and the cardiovascular system may also be affected by ventilation patterns such as the magnitude of tidal volume (Vt) [12, 13]. Thus we hypothesized that SVV measured by arterial pulse contour analysis is also affected by the depth of Vt. Therefore we studied SVV in patients in the early period after cardiac surgery both during volume responsiveness and immediately after fluid loading under different Vt.

Materials and methods

With the approval of the institutional review board and after obtaining participants' informed consent, 20 patients were studied after elective aortocoronary bypass grafting. Prior to surgery each patient had a central venous catheter and a 5-F thermistor-tipped arterial catheter (PV2025L20, Pulsion Medical Systems, Munich, Germany) inserted into the femoral artery and advanced into the distal abdominal aorta. This catheter allows continuous monitoring of arterial pressure, left ventricular stroke volume, and cardiac output by pulse contour analysis and the discontinuous measurement of cardiac output and stroke volume by arterial thermodilution [8]. The catheter was connected to a bedside hemodynamic monitor (PiCCO 4.12, Pulsion Medical Systems). Stroke volume was measured by arterial thermodilution (always performed in triplicate). Raw data were indexed using body surface area to calculate thermodilution stroke volume index (SVI). After initial calibration of pulse contour cardiac output by thermodilution cardiac output the pulse contour cardiac output and stroke volume were recorded continuously. SVV demonstrating the variation in the beat-to-beat stroke volume around the mean during the respiratory cycle was assessed using an algorithm that utilizes a continuously sliding time window of 30 s to calculate the mean stroke volume. This time window is divided into four periods of 7.5 s; within each window the highest and the lowest value of stroke volume were used to calculate SVV. Accordingly, ∆Pp expressing the variation in pulse pressure (systolic−diastolic arterial pressure) around the mean were calculated [6].

Measurements were performed postoperatively within the first 2 h after admission to the intensive care unit. Patients were sedated and mechanically ventilated using Vt of 10 ml/kg body weight and a positive end expiratory pressure of 5 cmH2O. Drug therapy, if any, was not changed within the study period. After initial measurement of SVI by thermodilution ventilator settings were changed in a randomized order. Vt of 5, 10, and 15 ml/kg body weight were applied with a frequency adaption to keep pCO2 (arterial blood gas analysis, Radiometer, Copenhagen, Denmark) within the range of 35–40 mmHg. The ratio of inspiration to expiration was not changed. After adjustment of each Vt an equilibration period of 5 min was allowed. Then SVV and ∆Pp at the respective Vt values were measured over a period of another 5 min. In parallel, central venous pressure (CVP) and thermodilution SVI were measured in triplicate.

Patients then underwent fluid loading using 10 ml hetastarch 6%, 130 kDa times body mass index. If thermodilution SVI increased by more than 10%, fluid loading was repeated until no further increase in SV was achieved. Then measurements at the respective Vt were repeated as described above.

Statistics

Results are shown as mean SEM. One-way analysis of variance was performed for repeated measurements. Pearson's product moment correlation and linear regression analysis were used. p<0.05 was considered significant.

Results

All 20 patients initially included tolerated the study well. At baseline SVV at Vt of 5 ml kg (7.1±0.6%) differed significantly from SVV at Vt of 10 ml/kg (14.8±1.8%) and SVV at Vt of 15 ml/kg (21.4±1.9). Further, ∆Pp at Vt of 5 ml/kg (9.1±0.9%) differed significantly from ∆Pp at Vt of 10 ml/kg (14.4±1.6%) and ∆Pp at Vt of 15 ml/kg (20.5±1.9%). Seven patients did not respond to the first volume loading with an increase in SVI of more than 10% and were excluded from further analysis. In the remaining 13 patients 32 volume loading steps (median two per patient) were performed. Hemodynamic and ventilation data before volume loading and after the last responsive volume loading step of these 13 patients are listed in Table 1. Both before and after volume loading SVV and ∆Pp at Vt of 5 and 15 ml/kg differed significantly from that at 10 ml/kg. As a result of volume loading SVV and ∆ Pp at the respective Vt were significantly lower than the values of SVV and ∆Pp prior to volume loading (Table 1). CVP did not differ significantly under the varying Vt. However, CVP was significantly higher after volume loading than at the respective Vt during volume responsiveness. Before volume loading SVV and ∆Pp were significantly correlated with the absolute values of Vt (SVV: R=0.57, p<0.001: ∆Pp: R=0.41, p<0.01; Fig. 1a, b). No correlation was found between SVV or ∆Pp and peak or mean inspiratory airway pressure. After volume loading SVV and ∆Pp were significantly correlated with the absolute values of Vt (SVV: R=0.67, p<0.0001; ∆Pp: R=0.58, p<0.001; Fig. 1c, d).

Table 1. Ventilatory and hemodynamic data (V t tidal volume, P IP peak inspiratory airway pressure, P aw mean airway pressure, MAP mean arterial pressure, SVI thermodilution stroke volume index, SVV pulse contour left ventricular stroke volume variation, ∆Pp pulse pressure variation, CVP central venous pressure)
Full size table
Fig. 1.
figure 1

Linear regression analysis of depth of tidal volume (V t ), stroke volume variation (SVV), and pulse pressure variation (∆Pp) before (a, b) and after volume loading (c, d)

Full size image

Discussion

Optimizing cardiac preload remains crucial for hemodynamic stabilization of critically ill patients. As demonstrated in several studies, functional preload parameters, such as SPV, ΔPp, and SVV, measuring the hemodynamic response to the specific interactions of the heart and the lungs under mechanical ventilation are highly useful in predicting volume responsiveness in mechanically ventilated patients at bedside [5, 6, 10, 11]. However, as shown earlier for the off-line parameter SPV, the hemodynamic effects of these heart lung interactions are potentially affected by the magnitude of Vt [12, 13]. We therefore investigated in mechanically ventilated cardiac surgery patients the influence of Vt on SVV measured by arterial pulse contour analysis, both during fluid responsiveness and after volume loading. Finally only data of those patients were included having responded to volume loading with an increase in SVI by more than 10%. Thus we ensured that all patients included were initially in a state of fluid responsiveness, i.e., on the volume dependent part of the Starling curve [14].

At the respective tidal volumes of 5, 10, or 15 ml/kg body weight SVV was greater during volume responsiveness than after volume loading. These data are in accordance with previous investigations on SVV [11, 12]. Further, SVV was greater when using higher Vt, both during volume responsiveness and after volume loading. These data explain the apparently conflicting differences between results of earlier studies in which lower values of SVV were reported at Vt of 10 ml/kg [11] than at higher Vt [12].

Furthermore, both before and after fluid loading there was a significant correlation between SVV and the absolute values of Vt (Fig. 1a, c), demonstrating the predominant contribution of Vt on SVV. After fluid loading this correlation was even closer, although SVV was reduced (Fig. 1c). This demonstrates that on the shallower part of the left ventricular Starling curve, i.e., when left ventricular filling is adequate, SVV is nearly determined solely by Vt. On the other hand, during volume responsiveness, i.e., on the steeper part of the Starling curve, poor ventricular filling (and thus low preload) additionally contributes to SVV.

However, measuring SVV by arterial pulse contour analysis in its present form does not allow a possible initial augmentation in stroke volume during early inspiration to be differentiated, which might specifically contribute to SVV during the phases of sufficient ventricular filling or potential ventricular overload. This differentiation has been reported for SPV by analysis of its delta up component, which, however, requires a short phase of apnea for determination [13]. Moreover, the present study did not focus on the specific changes in right ventricular hemodynamics under positive pressure ventilation and its contribution to these heart-lung interactions [15].

The present study found no significant correlation between peak airway pressure and SVV as tidal volume was varied, which could be theoretically expected. This might be explained by differing chest wall and pulmonary compliances among subjects. Thus, changing tidal volume potentially had different effects on pleural pressure and hence SVV. Another potentially contributing factor was the lacking pericardial constraint in these patients due to the surgical procedure. Therefore the effect of a modulation in left ventricular afterload which is potentially caused as well by a rise in intrathoracic pressure during mechanical inspiration might have been different from the "physiological" state of an intact pericardium. However, this observation needs further and focused investigation.

In summary, we found that in addition to by intravascular volume state SVV is influenced by the depth of tidal volume under mechanical ventilation. Therefore this effect must be regarded when using SVV for functional preload monitoring.