Abdominothoracic esophagectomy for esophageal cancer is a major surgical procedure with high rates of morbidity and mortality [1]. According to our experience, approximately 60% of patients who undergo esophagectomy develop hypotension requiring subsequent fluid volume loading during the first 15-h postoperative period, even though cardiovascular states immediately after surgery are relatively stable and/or postoperative bloody drainage is minimal [2].

Over the last decade, respiratory variations or dynamic preload variables such as stroke volume variation (SVV) and pulse pressure variation (PPV) have been reported to be better predictors of fluid responsiveness than commonly monitored static preload variables such as cardiac filling pressures [3, 4], even though SVV or PPV measurements generally require mechanical ventilation (tidal volume >8 mL/kg) in the absence of spontaneous breathing and/or cardiac arrhythmias [4, 5]. However, Cannesson et al. [6] recently showed that approximately 25% of patients undergoing prediction of fluid responsiveness are in the ‘grey zone’ of PPV and that a ‘black-or-white’ decision based on the receiver-operating characteristic (ROC) curve approach does not fit the reality of clinical or screening practice.

To our knowledge, only one study has measured SVV after esophagectomy, suggesting that SVV is clinically relevant as a guide of fluid volume management [7]. However, as this measurement was made in the presence of spontaneous respiratory activity (pressure support ventilation), it would be of limited use in evaluating fluid responsiveness. Abdominothoracic esophagectomy can lead to hemodynamic instability soon after the operation [8] and may also modify the original thoracic structure, decreasing the constraints of the chest wall imposed on the heart and the lungs and altering cyclic changes in intrathoracic pressure on heart-lung interactions. We therefore hypothesized that studies into fluid responsiveness would be of limited value during such hemodynamically unstable states.

The initial distribution volume of glucose (IDVG) has been proposed as a representative of the central extracellular fluid (ECF) volume status without significant modification of glucose metabolism [9, 10], and can be measured simply and rapidly in any intensive care unit (ICU) by injecting a small amount of glucose (5 g) and determining plasma glucose changes 3 min post injection [11]. Measurements can also be repeated at 30-min intervals without sustained increases in plasma glucose [12]. We previously reported that IDVG, rather than intrathoracic blood volume (ITBV) or central venous pressure (CVP), is closely correlated with cardiac output (CO) during hypotension and subsequent fluid volume loading early after esophagectomy [8]. Moreover, IDVG was recently reported to predict hypovolemic hypotension early after abdominal aortic surgery [13] and to have an inverse correlation with PPV after the induction of anesthesia in neurosurgical patients [14]. Accordingly, IDVG has the potential to be a useful marker of fluid responsiveness.

This study aimed to evaluate the ability of currently available preload variables such as SVV, PPV, CVP, and ITBV as well as IDVG to predict fluid responsiveness early after admission to the ICU following abdominothoracic esophagectomy.

Ethical approval for this study was provided by the Ethical Committee of Hirosaki University Graduate School of Medicine, Hirosaki, Japan, and each patient gave written informed consent before surgery. The study was planned to consist of at least 31 hypotensive patients, since a sample size of 31 is required to detect differences of 0.10 between areas under the ROC curve (5% type I error rate, 80% power, two-tailed test) [15]. Patients with aortic aneurysms and/or sustained arrhythmias including atrial fibrillation were excluded from the study. Those with diabetes mellitus and/or cardiovascular diseases such as hypertension without apparent ischemic heart disease were included. Preoperative echocardiographic measurements of left ventricular ejection fraction were required to exceed 60%.

Each patient underwent radical surgery for esophageal cancer that was performed using a right thoracoabdominal approach together with extensive resection of adjacent lymph nodes, subcarinal lymph nodes and/or cervical lymph nodes. Fluid and cardiovascular management decisions during anesthesia, including amounts of crystalloid solution and use of colloidal solutions, blood products or vasoactive drugs, were made by individual anesthesiologists. No patients received a continuous infusion of vasoactive drugs during anesthesia. Neither vasoactive drugs nor blood products were administered when patients arrived at the ICU. All patients postoperatively received controlled mechanical ventilation (tidal volume >8 mL/kg of ideal body weight), with peak airway pressure above positive end-expiratory pressure (10–15 cmH₂O, respiratory rate 12-15/min) with a low positive end-expiratory pressure (<5 cm H₂O) and continuous infusions of propofol (2–3 mg/kg/h) and morphine (0.4-0.8 mg/h) at least until the completion of the study. Supplemental midazolam (2–6 mg/h) was infused to achieve complete control of ventilation without spontaneous respiration during the study period. The infusion rate of these sedatives and analgesics was kept constant from at least 30 min before and during the study period. Although each patient had a thoracic epidural catheter for postoperative analgesia, epidural analgesia was only started after completion of the study.

Both 4.3% glucose solution with electrolytes and lactated Ringer’s solution were infused simultaneously at a constant rate of 1.5 mL/kg/h and 1.0 mL/kg/h, respectively, for at least 12 h after surgery. No vasoactive drugs were administered throughout the study period. One patient required continuous infusion of insulin (1 U/h) throughout the study period.

Initially, 45 consecutive patients with no arrhythmias after esophagectomy were recruited to the study. Of these, two were excluded because of technical problems in measuring CO, and the development of hypotension soon after admission to the ICU, requiring fluid volume loading during the first measurement. Thus, 43 consecutive patients were finally enrolled into the study. Of these, 23 patients were responsive to fluids and 20 were not.

The area under the ROC curve for evaluation of the ability to predict fluid responsiveness was highest for CVP (0.690) and lowest for ITBVI (0.584), but this difference was not statistically significant (Figure 1, Table 4). Pre-loading IDVGI was inversely correlated with △CI (r = −0.523, P <0.001), and pre-loading SVV and CVP were slightly correlated with △CI (r = 0.348, P = 0.026 and r = −0.307, P = 0.046, respectively) (Figure 2). Neither preloading PPV nor ITBVI were correlated with △CI (r = 0.288, P = 0.068 and r = −0.148, P = 0.345, respectively).

Most fluid responsiveness studies have been performed in the absence of severe hypotension. For ethical reasons, however, it is not appropriate to give fluid volume loading to normotensive patients, particularly non-responders in this study who received a large fluid volume during anesthesia and surgery, unless some indication is present.

Major surgical procedures result in a shift of the ECF from the central to the peripheral compartment as well as generalized capillary protein and water leakage both intra- and postoperatively [21]. Indeed, two patients in the present study recorded decreased CI despite fluid volume loading. Furthermore, most patients in this study required additional volume loading and/or an infusion of noradrenaline to overcome repeated hypotension throughout the first postoperative day. As hypotension development can be affected by hypovolemia and changes in peripheral vascular resistance, we believe that our study is ethically appropriate even in the absence of severe hypotension early after abdominothoracic esophagectomy.

The present study demonstrates that none of the tested variables can accurately predict fluid responsiveness soon after abdominothoracic esophagectomy, as assessed by the area under the ROC curve, even though definitions of fluid responsiveness may have a major impact on the results of SVV or PPV validity [22]. Fluid responders in this study were defined by an increase in CI >15% after fluid volume loading as reported previously [17, 18]. When CI or stroke volume index (SVI) for evaluating fluid responsiveness set the threshold value at 10%, the number of fluid responders and non-responders differed (31 versus 12 for CI at 10%, and 33 versus 10 for SVI at 10%) suggesting that ROC curve analysis is inadequate since it assumes an expected rate of fluid responsiveness of 50% [6, 23]. When the threshold value for SVI was set at 15%, only SVV reached the lowest edge of a ‘good’ diagnostic value (0.757), but there was no statistical difference among tested variables indicating no obvious difference from the present results(Additional file 1: Figure S1). Accordingly, we believe that a threshold value of 15% for CI in this study was adequate, even though the ‘grey zone’ approach has been proposed to avoid the binary constraints of a ‘black-or-white’ decision of the ROC curve approach [6].

The open-chest surgical procedure was performed with a right thoracoabdominal approach and it is likely that the extensive resection of adjacent lymph nodes, subcarinal lymph nodes, cervical lymph nodes and esophageal substitution such as gastric advancement would modify the intrathoracic structure. Furthermore, postoperative left pleural effusion was common. Two of the 43 patients experienced continuous air leakage from the chest drainage tube, so were excluded from the SVV and PPV study to avoid potential inaccuracies [20]. Indeed, these two patients had a low SVV (both 7%) and PPV (11% and 7%), despite an obvious increase in △CI (71% and 34%). Although a closed-chest condition after open-chest coronary artery bypass graft surgery enables the assessment of fluid responsiveness [24], changes in these thoracic structures and reduction of the pericardial constraint may have abated the effects of cyclic changes in intrathoracic pressure to heart-lung interactions even in the absence of an open-chest condition after esophagectomy. Such pathophysiology may lead to inaccuracies in respiratory variation results.

Of the variables tested, only pre-loading IDVGI had an inverse correlation with △CI following fluid volume loading in this study (power = 0.956). We used 250 mL of 10% dextran 40 solution for fluid volume loading, which has an oncotic pressure of 40 mmHg. Subsequent increments in plasma volume can exceed its infusion volume by up to 1.5 times while depleting the interstitial fluid volume [25], which could have an important impact on IDVG after fluid volume loading since IDVG represents both intravascular volume and the interstitial fluid volume of highly perfused tissues [10]. An inconsistent increase in IDVG associated with unchanged mean arterial pressure after colloid loading has also been reported following cardiac surgery [26]. Additionally, Harvey et al. [27] showed that IDVG and systolic area variability could not predict fluid responsiveness following cardiac surgery. Presumably, the presence of hemodynamically unstable states caused by internal bleeding, temperature change, alteration in vasomotor tone, or fluid shifts between compartments during volume loading rather than methodological flaws of IDVG would also play a role in these inconsistent results [28].

Although this study showed the limited predictive value of tested variables for fluid responsiveness, measurement of IDVG is desirable when either hypotension or decreases in arterial blood pressure occur early after abdominothoracic esophagectomy since IDVG has the highest correlation with CO, as reported previously [8]. When a small IDVG (<110 mL/kg) is observed, fluid volume loading is indicated and vice versa. However, an infusion of noradrenaline is indicated when a large IDVG (>130 mL/kg) is observed.

The present study has a number of limitations. First, most studies on fluid responsiveness evaluated post-fluid volume loading variables at the completion of fluid volume loading or soon after its completion [6, 18, 29]. In this study, however, its measurement was performed 10 min after completion of fluid volume loading as reported previously [8], since a minimum interval of 30 min is required for repeated IDVG measurements to avoid sustained hyperglycemia [12]. Consequently, many of the important signals may have been lost during this 10-min period, particularly during hemodynamic unstable states. However, the magnitude of an increase in CI after fluid volume loading in fluid responders of this study was comparable or even greater than other fluid responsive studies [6, 30], supporting the idea that the poor predictive values of tested variables were not attributable to insufficient signals in post-fluid loading measurements.

Second, we used 10% low molecular weighted dextran for fluid volume loading as reported previously [8], although 6% hydroxyethyl starch (medium molecular weight: 200,000 Da) is now widely used for this purpose and may have been more desirable in our study. However, considering the almost equivalent effects of either colloid on plasma volume expansion [25], it is likely that we can extrapolate our results to studies using 6% hydroxyethyl starch.

Third, we administered a fixed amount of dextran rather than an amount based on body weight. However, as determined by preoperative body weight, its variability was lower than that of other fluid responsiveness studies [6, 20, 23], reflecting insufficient preoperative nutritional status. As a lower preoperative body weight was observed in the majority of patients prior to diagnosis of esophageal cancer, most were administered approximately 4 mL/kg dextran.

Fourth, we did not test fluid responsiveness in the presence of hypotension or cardiac compromised conditions. Nevertheless, preload variables might be useful during hemodynamically unstable states such as hypotension when either fluid volume loading or an infusion of vasoactive drugs is required. Further studies are therefore required to assess this. Finally, we did not evaluate simultaneous echocardiography so an adequate view of cardiac chambers could not be consistently obtained following surgery for esophageal cancer.

Our data demonstrate that none of the dynamic, static, or volumetric variables measured in this study can be accurately used as a predictor of fluid volume responsiveness early after abdominothoracic esophagectomy.