The Intrathoracic Blood Volume Index as an Indicator of Fluid Responsiveness in Critically Ill Patients with Acute Circulatory Failure: A Comparison with Central Venous Pressure

Abstract

BACKGROUND: 

The intrathoracic blood volume index (ITBVI) and central venous pressure (CVP) are routinely used to predict fluid responsiveness in critically ill patients with acute circulatory failure (systolic blood pressure <90 mm Hg or vasopressor requirement). However, they have never been compared.

METHODS: 

In this prospective interventional study, we included 35 (21 men) mechanically ventilated and sedated patients with acute cardiovascular failure requiring cardiac output measurement (transpulmonary thermodilution technique). Fluid responsiveness was defined as an increase in stroke index (cardiac output/heart rate/body surface area) ≥15%. Receiver operating characteristic curves were generated for ITBVI and CVP.

RESULTS: 

Fluid challenge induced a stroke index increase ≥15% in 18 (51%) patients (responders). At baseline, no studied hemodynamic variables were different between responders and nonresponders. The areas under the receiver operating characteristic curves were 0.64 [95% CI: 0.46–0.80] for ITBVI and 0.68 [95% CI: 0.50–0.83] for CVP, without any statistical difference (P = 0.73). The best cut-off values for CVP and ITBVI were 9 mm Hg (sensitivity = 61%; specificity = 82%) and 928 mL · m⁻² (sensitivity = 78%; specificity = 53%).

CONCLUSION: 

ITBVI is similar to CVP in its ability to predict fluid responsiveness in critically ill patients with acute circulatory failure.

In Brief

IMPLICATIONS: Intrathoracic blood volume index and central venous pressure, when compared in critically ill patients with acute circulatory failure, have a similar ability to predict fluid responsiveness.

In critically ill patients with hypotension, intravascular volume expansion is a cornerstone of hemodynamic therapy, increasing cardiac output (CO) or stroke volume in 40% to 70% of patients.¹ However, fluid challenge can induce deleterious peripheral and pulmonary edema, compromising microvascular perfusion and oxygen delivery in patients with right or left ventricular dysfunction.^2,3 Therefore, many studies have focused on variables that could predict a beneficial effect and avoid deleterious effects of fluid administration.¹ It has been shown that dynamic indicators, such as pulsed pressure variation, variation in vena cava diameter and stroke volume variation better predict fluid responsiveness than static indicators: central venous pressure (CVP), pulmonary artery occlusion pressure (PAOP), and indicators derived from a transpulmonary thermodilution curve, such as global end-diastolic volume index and intrathoracic blood volume index (ITBVI).^1,4–8 However, static indicators remain widely used and are recommended to assess cardiac preload when they are associated with the measurement of cardiac index (CI) or ScV_O2 to guide hemodynamic therapy (goal-directed therapy) in patients undergoing cardiac surgery or in patients with septic shock.^9–11 In such patients, algorithms using global end-diastolic volume index or CVP decrease the need for vasopressors and mechanical ventilation, and decrease mortality rates.

The ITBVI was developed with the use of the PiCCO method and is considered an indicator of cardiac preload. ITBVI has been shown to be better correlated with the change in CI after fluid challenge than CVP and PAOP.^12–14 However, use of the PiCCO method to manage critically ill patients has been shown to decrease the need for vasopressors and mechanical ventilation despite a larger infused volume.¹⁵ The ability of ITBVI and CVP to predict fluid responsiveness has never been compared in critically ill patients.^4,7,8 Therefore, the aim of the present study was to assess and compare the ability of ITBVI and CVP to predict fluid responsiveness in critically ill patients with acute circulatory failure.

METHODS

As the present prospective study did not alter patient management, the local Ethics Committee stated that written informed consent was not necessary. However, the next of kin were systematically orally informed about the study.

From January 1 to December 1 2005, mechanically ventilated and sedated patients (Ramsay score¹⁶ = 4–6) without spontaneous breathing (assessed by visual observation of the airway pressure–time curve), with acute circulatory failure in whom CO was monitored with the PiCCO technique (Pulsion Medical Systems AG, Munich, Germany) were eligible to participate in the study. Acute circulatory failure was defined as systolic blood pressure <90 mm Hg or the need for vasopressors (dopamine >5 μg · kg⁻¹ · min⁻¹ or epinephrine or norepinephrine >0.1 μg · kg⁻¹ · min⁻¹) to maintain a systolic blood pressure >90 mm Hg.⁴ The combination of acute circulatory failure with a clinical infection, and the presence of systemic inflammatory response syndrome, defined septic shock.¹⁷

Inclusion and Exclusion Criteria

Eligible patients with acute circulatory failure and PiCCO monitoring were prospectively included when a fluid challenge was indicated by clinical signs of inadequate tissue perfusion, defined as oliguria <0.5 mL · kg⁻¹ · h⁻¹ for >2 consecutive hours, tachycardia >100 bpm, presence of skin mottling. Patients were excluded if they had cardiac arrhythmia or known tricuspid insufficiency, if they had impaired ventricular dysfunction and the physician assessed that a fluid challenge could be deleterious (cardiogenic pulmonary edema, PAOP >18 mm Hg or left ventricular ejection fraction <40% assessed by echocardiography), if they were moribund or parturient, or if they were <18 yr of age.

Fluid Challenge Procedure and Fluid Challenge Responsiveness

Fluid challenge was performed with hydroxyethyl starch (Voluven®, Fresenius Kabi, Louviers, France). As there is no clear consensus regarding the optimal volume for fluid challenge,¹⁸ the physician was allowed to choose the volume (250 or 500 mL) according to his clinical assessment of the potential risk of pulmonary edema. The fluid challenge was always given IV via a specific venous line at a constant rate (999 mL · h⁻¹) using an infusion pump.¹⁸ Fluid responsiveness was defined as an increase in stroke index (SI = ratio of CI and heart rate) ≥15%, separating the studied population into responders (R) and nonresponders (NR).⁴

Measured Variables and Time of Measurement

Patient Characteristics

Age, sex, height, weight, APACHE II score,¹⁹ and the number of organ dysfunctions using the ODIN score (Organ Dysfunctions and/or Infection)²⁰ were recorded at admission.

Mechanical Ventilation Variables

Tidal volume (mL · kg⁻¹ of ideal body weight), respiratory rate (per min) and the level of positive end-expiratory pressure were recorded.

Hemodynamic Variables

Heart rate (bpm), mean arterial blood pressure, CVP (mm Hg) and ITBVI (mL · m⁻²), CI (l min⁻¹ · m⁻²) and SI (mL · m⁻²) were measured or calculated before fluid challenge (baseline = T₀) and within 10 min after the end of the fluid challenge (T₁). The CO was calculated using the mean of three measurements after injection of 15 mL cold saline with an adequate thermodilution curve on the monitor screen (no rapid recirculation). CVP and mean arterial blood pressure were measured invasively with a zero referenced to the middle axillary line. CVP was measured at end-expiration. ITBVI was automatically calculated with the PiCCO technique adapted to the monitor used in the unit (Intellivue MP 160, Philips, Eindhoven, The Netherlands).

Statistical Analysis

Statistical analysis was performed with SAS© (SAS Institute, Cary, NC). The quantitative and qualitative variables are expressed as medians with 5th and 95th percentiles and by frequencies, respectively. For comparisons between R and NR and the effects of fluid challenge on hemodynamic variables, a Mann-Whitney test and a χ² or Fisher’s exact test were used when appropriate.

Receiver operator characteristic (ROC) curves were constructed to evaluate the capacity of ITBVI and CVP to predict fluid responsiveness. When the ROC curve area was more than 0.5, the best cut-off value was calculated and defined as the point of the ROC curve nearest to the ideal point (sensitivity = specificity = 1). Since no previous study directly has compared CVP and ITBVI to predict fluid responsiveness, we assumed that ITBVI could be of clinical interest when the area under the curve of its ROC curve value was >0.80, corresponding to the higher value of the 95% confidence interval of the area under the curve of ROC curve of CVP in previous studies.^4,5,7,8,21 For this purpose, and for an expected rate of fluid responsiveness = 50% and assuming an α error of 0.05 at a power of 0.20, 28 patients had to be included.

A comparison of the ROC curves of ITBVI and CVP was performed using a nonparametric method for paired data.²² Moreover, the relative changes (%) in SI, CVP and ITBVI before and after fluid challenge were calculated. A correlation was then sought between relative changes in SI and CVP and relative changes in SI and ITBVI.

A P value of <0.05 was considered as statistically significant.

RESULTS

Of 48 patients eligible for the study, 10 could not be included because of cardiac arrhythmia (n = 8) or moribund status (n = 2) and three were excluded because of protocol violation (too much delay between the end of the fluid challenge and the measurement at T₁). Therefore, 35 patients (21 men) were included in the study. The cause of acute circulatory failure was septic shock in 24 (69%) patients, hemorrhagic shock in 2 (6%) patients, and systemic inflammatory response syndrome in 9 (26%) patients (Table 1). Fluid challenge induced a SI increase ≥15% in 18 (51%) patients (R). Table 2 shows the comparison between R and NR. No statistical difference was shown between R and NR for CVP and ITBVI. Baseline individual values of R and NR for CVP and ITBVI are shown in Figure 1. The areas under the ROC curves of ITBVI and CVP were not statistically different (0.64 [95% CI: 0.46–0.80] vs 0.68 [95% CI: 0.50–0.83], P = 0.73) (Fig. 2). The best cut-off values for CVP and ITBVI were 9 mm Hg (sensitivity = 61%; specificity = 82%) and 928 mL · m⁻² (sensitivity = 78%; specificity = 53%), respectively. With CVP = 5 mm Hg and ITBVI = 691 mL · m⁻² there was 100% specificity in prediction of fluid responsiveness.

Table 1: Causes of Acute Circulatory Failure

Table 2: Comparison Between Responders and Nonresponders

Figure 1.: Individual values of intrathoracid blood volume index (ITBVI) (mL · m⁻²) and central venous pressure (CVP) (mm Hg) according to the fluid responsiveness with the best cut-off values (horizontal line) (ITBVI = 928 mL · m⁻²; CVP = 9 mm Hg). Individual values of ITBVI and CVP according to fluid responsiveness.

Figure 2.: Receiver operating characteristic curves of intrathoracid blood volume index (ITBVI) and CVP. Area under the curve: ITBVI 0.64 (95% CI: 0.46–0.80)CVP 0.68 (95% CI: 0.50–0.83) P = 0.73 receiver operating characteristic curves of ITBVI and CVP.

The relative changes in SI and CI were correlated with relative changes in ITBVI (r = 0.59, P = 0.001; r = 0.66, P = 0.0001, respectively), but no correlation was found between relative changes in SI and CI and relative changes in CVP (r = −0.07, P = 0.70; r = 0.10; P = 0.54) (Fig. 3).

Figure 3.: Top: correlation between changes in central venous pressure (CVP) (left) (%) (r = −0.07, P = 0.70) or changes in intrathoracid blood volume index (ITBVI) (right) (%) (r = 0.59, P = 0.001) and changes in stroke index (%) after fluid challenge. Bottom: correlation between changes in CVP (left) (%) (r = 0.10; P = 0.54) or changes in ITBVI (right) (%) (r = 0.66, P = 0.0001) and changes in cardiac index (%) after fluid challenge.

DISCUSSION

The main finding of the present study involving 35 critically ill patients with acute circulatory failure is that ITBVI was not superior to CVP for predicting fluid responsiveness. Nevertheless, low values of CVP (≤5 mm Hg) and ITBVI (≤691 mL · m⁻²) have 100% specificity in prediction of fluid responsiveness.

The ability of ITBVI to predict fluid responsiveness was compared to that of CVP. Although dynamic variables of fluid responsiveness, such as systolic arterial pressure variations,²³ down²⁴ and pulsed pressure variations⁴ were recently shown to be better indicators of fluid responsiveness than static variables (CVP and PAOP), these dynamic variables cannot be used in patients with cardiac arrhythmia or with spontaneous breathing. In such patients, the static variables remain the sole usable methods. Moreover, CVP is widely used to assess cardiac preload in critically ill patients.^9,10,25,26 Recently, ITBVI was developed and used to assess cardiac preload. Although CVP and ITBVI were shown to be less predictive of fluid responsiveness than dynamic variables, they were used in algorithms to guide vascular loading in order to detect deleterious (too deep) hypovolemia. These algorithms led to decreases in the mortality rates of patients with septic shock and to reductions in the need for vasopressors, catecholamines and mechanical ventilation in patients undergoing cardiac surgery.^10,11

In our study, CVP and ITBVI were not different in R and in NR at baseline. These findings are similar to those reported by Calvin et al.,²⁷ Reuse et al.²⁸ and Michard et al.⁴ for CVP and by Preisman et al.²¹ for ITBVI. ITBVI has been suggested to provide a better estimation of cardiac preload than CVP or PAOP.¹² The global end diastolic volume that was shown to be linearly correlated with ITBVI has been shown to be a better indicator of cardiac preload than CVP.²⁹ A good correlation was shown between the variation of ITBV (or global end diastolic volume) and the variation of CO after fluid challenge,^{12,13,30–35} whereas no correlation was found between the variation of filling pressure and the variation of CI. The present study confirms that there is a greater correlation between relative changes in ITBVI and SI or CI than there is between relative changes in CVP and SI or CI. Indeed, ITBVI and CO are mathematically coupled. Moreover, the relationship between intraventricular volume and intraventricular pressure defines the diastolic curve function (or ventricular compliance curve).³⁶ In the left portion of this diastolic function curve (which is horizontal), a large increase in volume is associated with a slight increase in pressure. The mathematic coupling and the ventricular compliance curve could explain a good correlation between ITBVI and SI, but a lack of correlation between CVP and SI.

We compared the ability of ITBV and CVP to predict fluid responsiveness in critically ill patients with acute circulatory failure, whereas previous studies have focused on other clinical conditions. In 18 patients undergoing cardiac surgery, one study²¹ reported areas under the ROC curves = 0.61 [0.47–0.75] and 0.71 [0.59–0.84] for CVP and ITBVI, respectively, but no comparison was performed. The proposed threshold value was <845 mL · m⁻² for ITBVI and could not be defined for CVP. In the present study, the areas under the ROC curves of ITBVI and CVP were not statistically different. Moreover, in our study the overlap between R and NR (Fig. 1) suggests that no value of CVP and ITBVI could be proposed as an accurate threshold of fluid responsiveness. However, low values of CVP (≤5 mm Hg) or ITBVI (≤691 mL · m⁻²) are specific for fluid responsiveness. These findings tend to confirm the findings reported by Rivers et al. and Goepfert et al.^10,11 suggesting that CVP and ITBVI could be used to exclude severe hypovolemia and in a goal-directed therapy algorithm. However, the use of ITBVI as a marker of cardiac preload has been shown (in a study comparing the PiCCO method to the pulmonary artery catheter) to be associated with a greater fluid balance [independently associated with intensive care unite (ICU) mortality] and fewer ventilator-free days.¹⁵ This suggests that fluid responsiveness does not necessarily mean that fluid is needed. Indeed, using fluid responsiveness indicators to guide volume management may have adverse effects on outcome.

The present study has several limitations. First, the studied population included septic (n = 24), hemorrhagic (n = 2) and vasoplegic (n = 9) patients, who could be considered as a heterogeneous population. However, the included patients reflect the real activity of an ICU. Second, although the amount of fluid was not statistically different between R and NR, some fluid challenges were achieved with small volume. Even if a larger volume could convert some NR to R, the main findings of the present study are not different, having shown that low values of CVP (<5 mm Hg) and/or ITBVI (<691 mL · m⁻²) are predictive of fluid responsiveness, whereas higher values of CVP and/or ITBVI cannot exclude an increase in SI after fluid challenge.

In conclusion, the present study suggests that ITBVI is not superior to CVP in predicting fluid responsiveness in ICU patients with acute circulatory failure.

ACKNOWLEDGMENTS

The authors thank Lana Zoric for reviewing the English language and Professor Daniel De Backer for his help in writing the article.