Predicting preload responsiveness using simultaneous recordings of inferior and superior vena cavae diameters

This prospective study was conducted in the ICU of a university hospital (Hôpital Purpan, Toulouse, France). The study was reviewed and approved by the Institutional Review Board (Comité de Protection des Personnes Hospices Civils de Limoges, France, approval CPP10-008a/2010-A00616-33). Written informed consent was obtained from each patient’s next of kin.

Inclusion criteria were mechanically ventilated patients in septic shock (as defined by the Surviving Sepsis Campaign [1]) who required a rapid volume challenge (7 ml/kg of 6% hydroxyethylstarch for 15 minutes) as directed by the attending physician. The physician’s decision was based on the presence of clinical signs of acute circulatory failure (low blood pressure or urine output, tachycardia, mottling) and/or biological signs of organ dysfunction (renal or hepatic dysfunction, lactic acidosis), as well as on the absence of contraindication to a fluid challenge (life-threatening hypoxemia, echocardiographic evidence of right ventricular failure). Excluded were patients with spontaneous respiratory effort and/or cardiac arrhythmias, as well as those in whom an echocardiographic examination could not be performed (that is, contraindication to TEE [12] or inability to perform TTE (n = 4, 8%)).

For each patient, echocardiographic assessments were performed double-blinded simultaneously by two experienced physicians (level 3 echocardiography training) [5] using a Doppler echocardiography device (EnVisor ultrasound system; Philips, Suresnes, France) equipped with a phased array transthoracic probe (3.5 MHz) and a multiplane transesophageal probe (5 MHz). Synchronization of the measurements with the different times of the ventilatory cycle was made possible by displaying the airway pressure curve on the screen of the ultrasound system (Echo Bridge; MAQUET, Rastatt, Germany).

The IVC was examined subcostally in the longitudinal view with the transthoracic probe. Its diameter was measured using the M-mode strictly perpendicular to the vessel and immediately above the juncture with the hepatic vein. Maximal and minimal IVC diameters (Dmax_IVC and Dmin_IVC, respectively) were measured over a single ventilatory cycle. The ΔIVC or the distensibility index of IVC, which reflects the increase of its diameter during mechanical insufflation, was calculated as (Dmax_IVC - Dmin_IVC)/Dmin_IVC and expressed as a percentage [8]. A ΔIVC ≥18% has previously been shown to have the best accuracy for predicting fluid responsiveness [8], and this threshold value was tested in our patients. In addition, we tested another previously published index based on the same measurements (Dmax_IVC - Dmin_IVC)/(Dmax_IVC + Dmin_IVC)/2) = ΔIVC₂), where ΔIVC₂ - 12% was the best threshold value for predicting fluid responsiveness [9].

The SVC was examined from a long-axis view with a transesophageal probe using the two-dimensional view to locate the M-mode beam across its maximal diameter, as previously described [7]. The SVC diameters (Dmax_SVC and Dmin_SVC) were measured over a single respiratory cycle, and the ΔSVC or the collapsibility index of SVC was calculated as (Dmax_SVC - Dmin_SVC)/Dmax_SVC and expressed as a percentage. As a ΔSVC >36% has been previously shown to have good accuracy for predicting fluid responsiveness [7], this threshold value was chosen for testing in our patients.

The left ventricular (LV) stroke volume was measured by using a Doppler technique with a transesophageal probe. The pulse Doppler aortic flow velocity-time integral (AoVTI) was determined at the level of the aortic annulus using a transgastric 120° view and the aortic valve area (SAo) at the level of the aortic annulus. AoVTI was measured only with TEE. The stroke volume was then calculated by multiplying AoVTI by Sao, and the cardiac index (CI) was determined by dividing the product of stroke volume and heart rate by the patient’s body surface area, as described and validated in previous studies [13]. Changes in CI before (T0) and after fluid challenge (T1) or ΔCI, were expressed as percentages. We calculated CI by using only TEE data [13].

Additionally, LV systolic function was measured before and after fluid challenge by calculating the LV fractional area change (LVFAC) as previously described [14]. A LVFAC <40% was considered as a LV dysfunction.

Measurements of Dmax_SVC, Dmax_IVC and AoVTI were performed in triplicate over three consecutive respiratory cycles. The results are expressed as the mean of these three measurements. The mean interobserver and intraobserver variabilities in the measurement of Dmax_SVC, Dmax_IVC and AoVTI were 8 ± 7% and 5 ± 6%, 9 ± 9% and 6 ± 8%, and 8 ± 6% and 5 ± 4%, respectively.

All patients were sedated and mechanically ventilated in a volume-controlled mode with a tidal volume of 8 to 10 ml/kg. Two sets of measurements were taken. The first was prior to volume expansion, and the second was immediately after volume expansion. Ventilatory settings as well as dosages of vasopressive drugs were held constant throughout the study. All Doppler echocardiographic measurements were taken offline from videotape recordings.

The effects of volume expansion on hemodynamic parameters were assessed using a nonparametric Wilcoxon rank-sum test. Assuming that a 15% change in CI was required for clinical significance [15],[16], patients were separated into responders (R) and nonresponders (NR) on the basis of a change in cardiac output ≥15% and <15%, respectively, following the volume challenge. The comparison of hemodynamic parameters prior to volume expansion in R and NR patients was performed using a nonparametric Mann-Whitney U test.

Receiver operating characteristic (ROC) curves were generated for ΔSVC and ΔIVC, with the discriminating threshold varied for each parameter. The areas under the ROC curves (AUC) for ΔSVC and ΔIVC were compared using the nonparametric test published by DeLong et al. [17]. The sensitivity, specificity, positive predictive value and negative predictive value of ΔIVC and ΔSVC for predicting fluid responsiveness were calculated. The best cutoff of ΔIVC and ΔSVC values were defined by the best cutoff of the sensitivity and specificity of each index. Correlations between ΔSVC and ΔCI, ΔIVC and ΔCI, and ΔSVC and ΔIVC were assessed using Spearman’s ρ coefficient. Linear correlations were tested using the Spearman’s rank method. Statistical analysis was performed using R software version (2.15.1; R Project for Statistical Computing, Vienna, Austria). All P-values were two-sided, and a P-value of 0.05 was considered significant.

Forty-four patients with sepsis or septic shock were included over an 11-month period. Twenty-six patients (59%) were R. Ten patients (22.7%) died during their ICU stay. Characteristics of the study patients and comparisons between R and NR at baseline are shown in Table 1. Hemodynamic and echocardiographic data in R and NR before and after fluid challenge (T0 and T1, respectively) are shown in Table 2. At baseline (T0), CI was not significantly different between R (2.3 L . min^-1 - m^-2 (95% CI: 1.3 to 3.4)) and NR (2.4 L . min^-1 - m^-2 (95% CI: 1.3 to 3.9)) (P = 0.841). Overall (R and NR), heart rate, mean arterial blood pressure and central venous pressure increased significantly after volume expansion (P < 0.005 for all comparisons).

The AUC for ΔSVC and ΔIVC regarding assessment of fluid responsiveness showed that ΔSVC showed better accuracy compared to ΔIVC (0.74 (95% CI: 0.59 to 0.88) versus 0.43 (95% CI: 0.25 to 0.61) (P = 0.012) (Figure 3). No significant correlation between ΔSVC and ΔIVC was found (r = 0.005, P = 0.98). ΔSVC and ΔIVC were significantly lower after volume expansion (P < 0.001), whereas changes for Dmax_SVC and Dmax_IVC were not significant (P = 0.16 and P = 0.06, respectively). Despite the significant difference between R and NR, the AUC of D_maxSVC remained low (0.67 (95% CI: 0.51 to 0.85)).

We assessed right ventricular function in all of the cases and detected three cases of right ventricular failure (right/left >0.6). In none of the three cases did we observe a difference in reactivity between IVC and SVC.

Overall, these results are in agreement with our main hypothesis of a dissociation between the ability of these dynamic vena cavae measurements to predict preload responsiveness. The better predictive value of ΔSVC could be due to a comparatively greater mechanical insufflation-induced decrease in venous return at the intrathoracic level compared to the intra-abdominal level. The greater impact of intrathoracic pressure variation could be related to (1) a greater increase in right atrial pressure (that is, in the back pressure to venous return) [10], (2) a greater increase of the right ventricular impedance due to the collapse of poorly filled alveolar vessels [18] or (3) the occurrence of a venous waterfall phenomenon between the extrathoracic and intrathoracic vena cavae segments [19]. Because our study was designed to be part of routine clinical practice, we were unable to determine which of the mechanisms described above was predominant. Furthermore, it must be highlighted that both indices were found to be less sensitive and less specific than previously reported. In our study, a ΔSVC >36% predicted fluid responsiveness with high specificity (100%) and high positive predictive value (100%), but with poor sensitivity (42%). Several explanations for such discrepancies between the present study and previously published work are possible [7],[20]. First, the mechanical ventilation settings were not similar. The respiratory rate and PEEP were higher in our study than in the study by Vieillard-Baron et al. [7] (20 breaths/min and 5 to 11 cmH₂O vs. 15 breaths/min and 5 to 7 cmH₂O). The ability of the SVC to collapse in the thorax is influenced by intrathoracic pressure and volume, and different HR/RR ratios may impact its reliability to predict fluid responsiveness [21]. In our study, one-third of patients had acute respiratory distress syndrome (ARDS), and 40.9% were ventilated with low tidal volumes and had HR/RR ratios <3.6. The parameters used to assess fluid responsiveness in this patient group have been questioned [22], and the low tidal volumes required in our patients with ARDS and low pulmonary compliance may have impacted the ability of ΔSCV to predict fluid responsiveness, indicating the potential limitation of the use of this index in such patients. We mention elsewhere that, in our present study, 26 patients (60%) were ventilated with a tidal volume >8 ml/kg. Although this may seem high by today's standards, it was, at the time of the first patient’s inclusion in March 2011, recommended to ventilate only patients with acute lung injury or ARDS with a low tidal volume (<8 ml/kg) [1]. Second, Vieillard-Baron and colleagues defined R as an increase >11% in CI, whereas we selected 15% to be consistent with data reported in the current literature [8],[9],[15],[16],[23],[24]; however, the sensitivity remained poor (39%), even when we used ΔCI - 11%. A larger proportion of our cohort received vasopressor support (75% versus 50%), and it has been shown that norepinephrine can affect fluid challenge [25].

Our results show that the AUC for ΔSVC regarding assessment of fluid responsiveness was low (0.74 (95% CI: 0.59 to 0.88)). Contrary to the findings of other investigators, we discovered that, in real-life ICU practice conditions, the AUC of ΔIVC seems not to be a reliable predictor of fluid responsiveness and that its TEE counterpart, ΔSVC, shows a poor fluid responsiveness except for the high variation levels.

Regarding the IVC, in contrast with previous researchers, we found a poorer sensitivity (42%) and specificity (39%), despite the fact that we used the same threshold of ΔIVC (≥18%) [8]. Several physiological hypotheses should be considered. For instance, because the IVC is mainly intra-abdominal, its ability to distend could be limited by an increase in intra-abdominal pressure, especially in postoperative abdominal surgery patients. In a recent study, the impact of intra-abdominal pressure on IVC diameter was evaluated in mechanically ventilated pigs. The results showed that IVC diameters are affected by intra-abdominal pressure and that fluid responsiveness should not be estimated from retrohepatic IVC diameter in cases of high intra-abdominal pressure [26]. Our results show a weak AUC for ΔIVC (0.43 (95% CI: 0.25 to 0.61)), suggesting that ΔIVC may not be a consistently reliable predictor of fluid responsiveness.

Our results, as compared to those of previously published studies, suggest that in ARDS patients, a standard ventilation strategy, high vasopressor infusion rate and/or abdominal surgery may alter the ability of ΔSVC and ΔIVC to predict fluid responsiveness and that these predictive indices should be investigated extensively and refined before any generalized use can be recommended. Furthermore, the selection of a single cutoff point for making clinical decisions may be too simplistic. A "gray zone" approach applied to the pulse pressure variations for prediction of fluid response in mechanically ventilated patients under general anesthesia was recently suggested by Cannesson et al.[27]. This "gray zone" approach has not been used in our study, because the size of our cohort did not permit such statistical analysis.