Tidal volume challenge–induced hemodynamic changes can predict fluid responsiveness during one-lung ventilation: an observational study

Background To evaluate the ability of tidal volume challenge (VTC)-induced hemodynamic changes to predict fluid responsiveness in patients during one-lung ventilation (OLV). Methods 80 patients scheduled for elective thoracoscopic surgery with OLV were enrolled. The inclusion criteria were: age ≥ 18 years, American Society of Anesthesiologists physical status I-III, normal right ventricular function, normal left ventricular systolic function (ejection fraction ≥55%), and normal or slightly impaired diastolic function. The study protocol was implemented 15 min after starting OLV. Simultaneous recordings were performed for hemodynamic variables of diameter of left ventricular outflow tract, velocity time integral (VTI) of aortic valve, and stroke volume (SV), and ΔSV-VTC, ΔVTI-VTC, and ΔMAP-VTC were calculated at four time points: with VT 5 mL/kg (T1); after VT increased from 5 mL/kg to 8 mL/kg and maintained at this level for 2 min (T2); after VT was adjusted back to 5 mL/kg for 2 min (T3); and after volume expansion (250 mL of 0.9% saline infused over 10–15 min) (T4). Patients were considered as responders to fluid administration if SV increased by ≥10%. Receiver operating characteristic (ROC) curves for percent decrease in SV, VTI, and MAP by VTC were generated to evaluate their ability to discriminate fluid responders from nonresponders. Results Of the 58 patients analyzed, there were 32 responders (55%) and 26 nonresponders (45%). The basic characteristics were comparable between the two groups (p > 0.05). The area under the curve (AUC) for ΔSV-VTC, ΔVTI-VTC, and ΔMAP-VTC to discriminate responders from nonresponders were 0.81 (95% CI: 0.68–0.90), 0.79 (95% CI: 0.66–0.89), and 0.56 (95% CI: 0.42–0.69). The best threshold for ΔSV-VTC was −16.1% (sensitivity, 78.1%; specificity, 84.6%); the best threshold for ΔVTI-VTC was −14.5% (sensitivity, 78.1%; specificity, 80.8%). Conclusion Tidal volume challenge–induced relative change of stroke volume and velocity time integral can predict fluid responsiveness in patients during one-lung ventilation. Clinical Trial Registration: Chinese Clinical Trial Registry, No: chictr210051310.


Introduction
Perioperative hemodynamic optimization is essential to prevent tissue underperfusion and fluid overload in patients with acute circulatory failure in the intensive care unit and operating room (1). Numerous studies have shown that, in critically ill and surgical patients, unnecessary perfusion increases morbidity, mortality, and length of hospital stay (2)(3)(4)(5)(6)(7)(8)(9)(10). Only in cardiac preload-dependent conditions, i.e., the rising part of the Frank-Stalin curve, does stroke volume (SV) increase after volume expansion. However, this occurs in only about 50% of intensive care and surgical patients (11,12). In preload-independent conditions, i.e., the plateau branch of the Frank-Staling curve, volume expansion does not increase SV and may even increase the risk of adverse effects.
A proper fluid management strategy is crucial for patients undergoing thoracic surgery. Oxidative stress during one-lung ventilation (OLV) and after pulmonary resuscitation may cause pulmonary edema (13,14). Pulmonary surgery can also lead to local edema via release of proinflammatory cytokines. In 2019, the Enhanced Recovery after Surgery Society and the European Society of Thoracic Surgeons recommended restrictive fluid therapy during lobectomy (15). However, there is concern that excessive restriction of fluid infusion may lead to hypovolemia and impaired tissue perfusion, organ dysfunction, and acute kidney injury. Therefore, assessment of fluid responsiveness is extremely important for hemodynamic management during OLV. As is well known, volume responsiveness assessment is the cornerstone of fluid therapy.
Traditional static indicators such as central venous pressure and pulmonary artery wedge pressure do not accurately reflect volume responsiveness (16,17). Dynamic indicators such as stroke volume variation (SVV) and pulse pressure variation (PPV) are currently recognized as the best objective indicators of volume responsiveness of patients on conventional mechanical ventilation (16,18). But the pathophysiological changes caused by opening the nonventilated side of the chest cavity during thoracic surgery, OLV, and lung-protective ventilation strategies with small tidal volumes can limit the usefulness of these dynamic indicators (19,20). Changes in intrathoracic pressure affect the amount of blood returned to the heart, which in turn leads to changes in hemodynamics. Some authors have applied functional hemodynamic tests in the attempt to overcome the above limitations (21-23). One of these tests-the tidal volume challenge (V T C) testwhich assesses changes in hemodynamics following an increase in the tidal volume from 6 mL/kg to 8 mL/kg for 1 min, has been shown to reliably predict fluid responsiveness in patients undergoing neurosurgery (24). To our knowledge, no studies have evaluated the usefulness of V T C for predicting fluid responsiveness during OLV. We hypothesized that the hemodynamic changes induced by temporarily increasing tidal volume from 5 mL/kg to 8 mL/kg could predict fluid responsiveness in patients receiving OLV. This study was designed to determine whether tidal volume challenge (V T C)-induced hemodynamic changes can predict fluid responsiveness in patients during OLV.

Study design
This study was approved by the hospital ethics committee (approval No: 2021ky244) and registered in the China Clinical Trial Registration Center (Registration No: chictr210051310). Written informed consent was obtained from all patients.

Perioperative management
All patients received standard intraoperative monitoring (heart rate, peripheral oxygen saturation [SpO 2 ], continuous electrocardiographic, and noninvasive blood pressure monitoring). General anesthesia was induced, after preoxygenation, with propofol, midazolam, sufentanil, and atracurium cis-benzene sulfonate, and maintained with sevoflurane, remifentanil, dexmedetomidine, and atracurium cis-benzene sulfonate. The depth of anesthesia was controlled to maintain a bispectral index of 45-60 throughout the duration of surgery.
After the airway was secured, the patient was turned to the lateral position, and OLV was initiated with lung-protective ventilation (tidal volume, 5 mL/kg of ideal body weight [IBW]; positive end-expiratory pressure [PEEP], 5 cm H 2 O). Ventilatory rate was controlled to keep end-tidal CO 2 within 35-40 mm Hg, and peak inspiratory pressure below 30 cm H 2 O. Inspired oxygen concentration was 100% at the start of OLV and gradually decreased, ensuring that SpO 2 remained >95% throughout OLV. After anesthesia induction, a 20G cannula was inserted into the radial artery for invasive blood pressure monitoring, and an esophageal ultrasonic Doppler probe was placed through the mouth. Throughout the study duration, Ringer solution was administered intravenously at a rate of 3 mL·kg −1 ·h −1 .

Study protocol
The study protocol (Figure 1) was initiated only after the beginning of the operation (15 min after starting OLV) to minimize the effect of the surgery type on the measured variables. During the study protocol, CO 2 insufflation was not used in the hemithorax being operated upon. The V T C was performed under hemodynamically stable conditions (i.e., heart rate and blood pressure not fluctuating by more than 10% over 1 min before each measurement) without the administration of vasoactive drugs. Transesophageal ultrasound was used to record the inner diameter (D) of the left ventricular outflow tract (LVOT) and velocity time integral (VTI) of the aortic valve in the long axis view of the gastric fundus, in the middle section of esophagus at four time points: T 1 (with V T 5 mL/kg IBW); T 2 (after V T C application by increasing the tidal volume up to 8 mL/kg IBW for 2 min); T 3 (after Frontiers in Medicine 03 frontiersin.org reducing V T back to 5 mL/kg IBW for 2 min); and T 4 (after volume expansion intravenous infusion of 250 mL normal saline for 10-15 min). Using the measured indices, the following were calculated: 1) SV was automatically calculated by the ultrasonic instrument According to the ΔSV-VE, the patients were separated into two groups: responders (ΔSV-VE ≥10%) and nonresponders (ΔSV-VE <10%) (25).

Statistical analysis
Statistical analysis was performed using SPSS 21.0 (IBM Corp., Armonk, NY, United States). Data were summarized as the means (±standard deviation), medians (interquartile range), or numbers of patients (%). Categorical variables were compared between groups, using the Fisher exact test or the chi-square test. Time-dependent data were compared (between periods in each group and between groups) using the repeated-measures analysis of variance. Preplanned subgroup analysis was conducted using the unpaired t-test if the data were normally distributed or the Mann-Whitney U-test if they were not normally distributed. All statistical tests were two-sided, and p < 0.05 was considered statistically significant. Receiver operating characteristic (ROC) analysis was performed for assessing the abilities of ΔSV-V T C, ΔVTI-V T C, and ΔMAP-V T C to discriminate responders from nonresponders. The best threshold was defined as the point closest to the upper left corner of the ROC curve (26). In addition, the gray-zone approach was applied for ΔSV-V T C and ΔVTI-V T C. The gray-zone approach is a method used to assess the nonconclusive range of clinical measurements (25,27). In the first step, bootstrap resampling was performed for ΔSV-V T C and ΔVTI-V T C, and the "best threshold" was estimated from each bootstrap sample (obtained by ROC analysis); then, the standard deviation of those 1,000 estimated thresholds was used to estimate the standard error (and 95% confidence interval) of the original (on original data set) best threshold estimate. This calculation was performed using MedCalc (MedCalc Software, Mariakerke, Belgium). In the second step, the inconclusive range for each variable was determined to assess the responsiveness, which was calculated as the cutoff value with sensitivity of <90% or specificity of <90%. If the range (95% CI of the best cutoff threshold) calculated from the first step was larger than that calculated from the second step, the values from the first step were selected as the grayzone values.

Sample size estimation
The sample size was calculated using PASS (version 15, NCSS, LLC, Kaysville, Utah, United States). Since there are no previous studies related to the application of VTC in patients during OLV, we expect an area under the ROC curve (AUC) for VTC-induced hemodynamic changes to be set at of at least 0.75. This value was compared with the null hypothesis (AUC = 0.50; sample size for negative/positive group = 1). A minimum of 36 patients were needed (type I error of 0.05 and type II error of 0.2).

Patient characteristics
Of the 80 patients initially enrolled in the study, 58 completed the study; four patients were excluded because of high airway Study protocol. T1, with tidal volume of 5 mL/kg ideal body weight; T2, after application of VTC by increasing the tidal volume to 8 mL/kg ideal body weight for 2 min; T3, after reducingtidal volume back to 5 mL/kg ideal body weight for 2 min; and T4, after volume expansion (by intravenous infusion of 250 mL normal saline over 10-15 min). ΔSV-VTC indicates relative change of SV after VTC; ΔVTI-VTC indicates relative change of VTI after VTC, ΔMAP-VTC indicates relative change of MAP after VTC, ΔSV-VE indicates relative change of SV after volume expansion.
Frontiers in Medicine 04 frontiersin.org pressure after V T C, five patients were excluded because of difficulty in image acquisition, eleven patients were excluded because of hemodynamic instability during the studyand two were excluded because of operative method change ( Figure 2). Table 1 summarizes the characteristics of the patients. Among the 58 patients, 32 (55%) were classified as responders and 26 (45%) as nonresponders. There were no significant differences in characteristics between the two groups (p > 0.05). Table 2 shows the hemodynamic variables in responders and nonresponders. In both groups, MAP, SV, and VTI decreased significantly after V T C and increased significantly after volume expansion. Figure 3 depicts the evolution of SV and VTI in responders and nonresponders during the four-step study period. Figure 4 shows the systolic blood flow spectrum at the aortic valve orifice, as seen through the long axis section of gastric fundus, at different time points. The peak systolic velocity at the aortic valve orifice decreased significantly at T 2 , recovered at T 3 , and then again increased after volume expansion at T 4 .

Hemodynamic data
Relationship between hemodynamic changes induced by V T C and hemodynamic changes induced by volume expansion Table 3 and Figure 5 show the relationship between ΔSV-V T C and ΔSV-VE, ΔVTI-V T C, and ΔSV-VE. Figure 6 shows the predictive values of ΔSV-VTC, ΔVTI-VTC, and ΔMAP-VTC for fluid responsiveness during OLV. A 16.1% decrease in ΔSV-V T C predicted fluid responsiveness with sensitivity of 78.1% and specificity of 84.6%; the area under the curve (AUC) for ΔSV-V T C to discriminate responders was 0.81 (95% CI, 0.68-0.90). A 14.5% decrease in ΔVTI-V T C predicted fluid responsiveness with sensitivity of 78.1% and specificity of 80.8%; the AUC for ΔVTI-V T C to discriminate responders was 0.79 (95% CI, 0.66-0.89). A 5.1% decrease in ΔMAP-V T C predicted fluid responsiveness with sensitivity of 53.1% and specificity of 65.4%; the AUC for ΔMAP-V T C to discriminate responders was 0.56 (95% CI, 0.42-0.69).

Gray-zone approach for ΔSV-V T C and ΔVTI-V T C
The inconclusive region of hemodynamic changes to predict fluid responsiveness was depicted using gray-zone approach. The gray-zone of ΔSV-V T C was between −20.1 and −10.21% with 28 numbers of patients (14 responders and 14 nonresponders) ( Figure 7A). The grayzone of ΔVTI-V T C was between −22.9 and −9.6% with 29 numbers of patients (17 responders and 12 nonresponders) ( Figure 7B).

Discussion
This study evaluated the value of V T C-induced hemodynamic changes for predicting fluid responsiveness in patients receiving FIGURE 2 Flow diagram of the study.
Frontiers in Medicine 05 frontiersin.org OLV and found significant changes in hemodynamics after V T C in both responders and nonresponders; in both groups, MAP, SV, and VTI increased significantly after volume expansion. The V T Cinduced changes in SV and VTI showed significant correlation with the increases in SV after subsequent volume expansion, but the correlation of MAP was poor. ROC curve analysis showed that the V T C-induced changes in SV and VTI could predict fluid responsiveness of OLV patients, while the changes in MAP could not.
Increase in intrathoracic pressure leads to decrease in venous return and right ventricular preload. Meanwhile, increase in transpulmonary pressure leads to increase in right ventricular afterload and decrease in right ventricular ejection, which finally results in decreased cardiac output (28). Previous study found thatPEEP induces a decrease in SV, especially in patients with insufficient blood volume (29). This explains why SV, VTI, and MAP decreased to a certain extent when the tidal volume of patients on OLV was increased from 5 mL/kg to 8 mL/kg. When tidal volume, was  Data are expressed as mean ± standard deviation or median (interquartile range). T1, with tidal volume of 5 mL/kg ideal body weight; T2, after application of VTC by increasing tidal volume up to 8 mL/kg ideal body weight for 2 min; T3, after reducing tidal volume back to 5 mL/kg ideal body weight and stabilize for 2 min; T4, after volume expansion (with intravenous infusion of 250 mL normal saline over 10-15 min).
Frontiers in Medicine 06 frontiersin.org Changes in (A) heart rate, (B) mean arterial pressure, (C) stroke volume, and (D) velocity time integral among responders and nonresponders at different time points. T1, with tidal volume of 5 mL/kg ideal body weight; T2, after application of VTC by increasing the tidal volume to 8 mL/kg ideal body weight for 2 min; T3, after reducingtidal volume back to 5 mL/kg ideal body weight for 2 min; and T4, after volume expansion (by intravenous infusion of 250 mL normal saline over 10-15 min). Frontiers in Medicine 07 frontiersin.org decreased, the hemodynamics recovered rapidly. These findings suggest that tidal volume loading test could be a reliable functional hemodynamic test to predict fluid responsiveness in patients receiving OLV. At present, the use of dynamic indicators such as stroke volume variation (SVV) and pulse pressure variation (PPV)for evaluation of fluid responsiveness during OLV is controversial (19, 30). Although there is a mechanism of hypoxic pulmonary vasoconstriction in the process of OLV, part of the blood flow persists in the nonventilated side but is not involved in the respiratory-related changes of SV. Meanwhile, because of pleural cavity opening and other factors, there are no periodic changes related to mechanical ventilation in the nonventilated lung. A previous study reported that when the tidal volume during OLV was changed from 6 mL/kg to 8 mL/kg, the area under the ROC curve improved from 0.648 to 0.776, indicating that the ability of dynamic indicators to predict fluid responsiveness in OLV patients is related to the tidal volume, and that accuracy can be achieved only when the tidal volume is ≥8 mL/kg (20). Lungprotective ventilation is becoming standard intraoperative management for improving postoperative outcomes during OLV (15,   Frontiers in Medicine 08 frontiersin.org 31,32). In patients receiving lung-protective ventilation, dynamic indices are not very useful. Therefore, a new approach is needed to assess the volume status of patients during OLV.
To overcome the tidal volume-related limitations of dynamic indicators, some researchers proposed the use of V T C (tidal volume increase from 6 mL/kg to 8 mL/kg during two-lung ventilation) to evaluate fluid responsiveness in patients receiving low tidal volume ventilation (24,33). The consensus of Chinese Experts on Perioperative Lung Protection in Thoracic Surgeryrecommends a tidal volume of 4-6 mL/kg for perioperative lung-protective ventilation. Therefore, in this study, the baseline tidal volume was set at 5 mL/kg, and then increased to 8 mL/kg. Volume state decides the hemodynamic response to of V T C. V T C will increase intrathoracic pressure and the resulting decrease in SV and VTI will make the Frank-Starling curve move to the right; the impact will be more obvious in patients with hypovolemia. However, whether V T C can be used to evaluate fluid responsiveness in patients with OLV is unknown. In this study, the tidal volume of OLV patients was increased from 5 mL/kg to 8 mL/kg. The AUCs of V T C-induced relative changes of SV and VTI were 0.81 and 0.79, respectively. The sensitivity and specificity were 78.1 and 84.6%, respectively, for SV and 78.1 and 80.8%, respectively, for VTI. A previous study that assessed the usefulness of change in SV for predicting fluid responsiveness during OLV reported an AUC of 0.84 and sensitivity and specificity of 76.5 and 84.6%, respectively (34). A similar study found that the relative change in SVV and PPV induced by PEEP upregulation from 0 cm H 2 O to 10 cm H 2 O can predicted fluid responsiveness in patients during OLV. The AUCs of change in SVV and change in PPV for evaluating fluid responsiveness during OLV were 0.90 and 0.88, respectively; the sensitivity and specificity were 88 and 82%, respectively for change in SVV and 83 and 72%, respectively, for change in PPV (35). These findings suggested that functional hemodynamic tests can accurately predict the volume reactivity of patients with OLV.
Although most studies use ROC analysis to evaluate diagnostic efficiency and determine the cutoff value, sensitivity, and specificity, the method may not accurately reflect the clinical situation (36). ROC analysis can only provide a single cutoff value to separate patients into two types (e.g., responder or nonresponder in this study). However, in our study, "responder" does not necessarily mean that the patient is in a state of hypovolemia. A variety of other factors, such as increased venous volume and decreased venous wall tension, will also lead to the patient being classified as a "responder. " The gray-zone method provides two cutoff values, which is of more practical use (37). This method clearly provides the highest and lowest cutoff values, which allows the clinician to make more rational decisions regarding rehydration in the operating room. When ΔSV-V T C or ΔVTI-V T C is in the gray-zone between the two cutoff values, there is uncertainty. In our study, 48% of patients were in the gray zone of ΔSV-V T C for predicting fluid responsiveness and 50% were in the gray zone of ΔVTI-V T C for predicting fluid responsiveness. For patients in the gray zone, a mini fluid challenge can be carried out to observe the changes in hemodynamics. Using gray-zone values instead of a single cutoff value for goal-directed fluid therapy can optimize fluid management and free the clinician from the binary "black-white" decision making of the ROC method.
There are several limitations in the present study. First, the total number of patients may not be sufficient to generalize our result to all patients. Further studies with larger number of patients calculated with precise parameter are needed to describe the physiologic mechanism of V T C on dynamic parameters during OLV. Second, the results of this study cannot prove the effectiveness of V T C in predicting fluid responsiveness in patients with decreased left ventricular function. Patients with decreased cardiac reserve function are more vulnerable to volume load and need more refined fluid management strategies. In addition, due to the patient's own condition, ultrasound measurements could not be obtained in few patients in this study, and the use of transesophageal ultrasound may be impractical in many cases due to the limited availability of ultrasound machines or in shortterm surgeries.

Conclusion
In conclusion, the relative changes in SV and VTI induced by V T C can accurately predict fluid responsiveness during OLV, but the relative change in MAP induced by V T C cannot.

Data availability statement
The datasets presented in this study can be found in online repositories. The names of the repository/repositories and accession number(s) can be found in the article/supplementary material.