Preload Dependence Is Associated with Reduced Sublingual Microcirculation during Major Abdominal Surgery

Perioperative hemodynamic optimization has proven to improve postoperative outcome in high-risk surgical patients.^1,2  This allows a reduction in the length of stay, mortality, and postoperative complications.^3–5  Fluid administration is the cornerstone of perioperative hemodynamic optimization and is based on the prediction of preload responsiveness. Dynamic indices, such as pulse pressure variation, detect preload dependence and are used to predict fluid responsiveness. Preload dependence is defined as a state in which increases in right ventricular and/or left ventricular end-diastolic volume result in an increase in stroke volume.⁶  Nevertheless, while fluid resuscitation guided by dynamic indices leads to optimization of macrocirculation, it remains unclear whether it can improve microcirculation as well. Such improvement can only occur if the physiologic microvascular response is preserved. However, hemorrhagic shock, reperfusion injuries, sepsis, traumatic injuries, and inflammation are conditions known to affect the microcirculatory responses, subsequently leading to a loss of the physiologic relation between macro- and microcirculation. Therefore, in such cases, a systemic hemodynamic-driven resuscitation would not be effective in restoring an adequate microcirculation and tissue oxygenation.^7–12  In this regard, indicators of normal microcirculation should be taken into account by the physician in his evaluation of the necessity of fluid resuscitation, even in the case of preload dependence.

Microcirculation monitoring remains limited to a few easily accessible sites. Sublingual microcirculation sequences can be obtained using a handheld vital microscope (sidestream dark field imaging device). This noninvasive technique allows an assessment of microvascular variables such as microvascular flow index, percentage of perfused vessels, perfused vessel density, and heterogeneity index.¹³ 

Alteration of microvascular perfusion has been closely associated with organ dysfunction and mortality in septic and hemorrhagic shocks.^14–17  Several studies have investigated the relationships between macrovascular hemodynamic optimization and microcirculation in shock. However, very few studies have analyzed the behavior of microcirculation during major surgery.^18–21  To optimize cardiac output and oxygen delivery, hemodynamic management guided by functional assessment of fluid responsiveness is used. Dynamic predictors of fluid responsiveness, namely pulse pressure variation and stroke volume variation, have been shown to be useful in anticipating patients who will respond to a fluid load with a significant increase in stroke volume and cardiac output.²²  Therefore, dynamic predictors of fluid responsiveness are increasingly used to guide fluid therapy.  A pulse pressure variation value greater than 13% allows discrimination between responders and nonresponders with a sensitivity of 94% and a specificity of 96%.^23,24  The purpose of this study was to test the hypothesis that during abdominal surgery, microvascular perfusion is impaired during preload dependence, identified by a pulse pressure variation greater than 13%, and recovers after fluid administration.

This prospective study was conducted in the Department of Anesthesiology and Critical Care of the University Hospital of Bicêtre, Le Kremlin Bicêtre, France.

This observational study was approved by our local Institutional Review Board (Protection of Persons Committee, University Hospital of Bicêtre; 2013-A01277-38). Patients of the study were treated in accordance with the local care protocols. Examination of the sublingual microcirculation with a sidestream dark field imaging device is completely noninvasive and respects the noninterventional study design.

Selection of patients was done during anesthesia consultation. Included patients planned for an abdominal surgery of 90 min or greater and the anesthetist deemed the implementation of an arterial catheter necessary to monitor blood pressure based on their condition and the risk of hemodynamic instability during surgery. Oral and written information was given to each patient.

Patients were excluded for any of the following reasons: surgical duration of less than 90 min, laparoscopy, nonsinus cardiac rhythm, end-stage heart failure, pregnancy, age less than 18 yr, contraindication to arterial catheterization (Allen test positive or inconclusive), and impossible access to the patient’s head during surgery.

All cardiovascular, respiratory, and neurologic comorbidities, American Society of Anesthesiologists and New York Heart Association scores, usual medications, clinical abnormalities, and results of preoperative diagnostic tests were recorded on the anesthesiology report.

All patients were treated in accordance with local care protocols. The pulse pressure variation is routinely used as a marker to detect preload dependence,²⁵  and a pulse pressure variation value of 13% allows discrimination between responders and nonresponders with a sensitivity of 94% and a specificity of 96%.²³  Every 10 min, the following parameters were collected and recorded on the monitoring report: cardiac index, stroke volume index measured using the Flotrac-Vigileo device (Edwards Lifesciences, USA), pulse pressure variation, heart rate, mean arterial pressure, pulse oxygen saturation, respiratory rate, temperature measured from the Philips IntelliVue Monitoring (Philips Medical Systems, USA), Bispectral Index, inspired and expired oxygen, carbon dioxide, volatile anesthetic fractions, and tidal volume (V_T).

Videos of the sublingual microcirculation were obtained using a sidestream dark field imaging device (Microscan, MicroVisionMedical, The Netherlands), a technique derived from the orthogonal polarized spectral imaging technology.²⁶  Intraoperative sublingual video allowed direct visualization of sublingual blood flow in microvascular networks. The emitted light, corresponding to the hemoglobin wavelength absorption, displays each erythrocyte in black on a light background. Images acquisition and analysis were performed according to international guidelines.¹³  The greatest care was provided in order to avoid pressure artifacts in accordance with the published recommendations.¹³  Sequence quality was systematically evaluated using the “Microcirculation Image Quality Score” described by Massey et al.²⁷  Poor quality sequences were therefore eliminated at the bedside of the patient, and another sequence was performed to have two good quality sequences for each measurement point. Patient identity and video identification were hidden and replaced by an alphanumeric code assigned randomly, allowing a blinded analysis.

During the surgery, two 10- to 15-s video sequences were acquired every 10 min using the sidestream dark field imaging device by an investigator who was not involved in the patient’s care. No real-time on-site analysis of videos was done. No decision concerning the management of the patient was made on the basis of microcirculatory data acquired during the surgery.

According to the consensus,^13,28  a retrospective analysis of these videos was performed by a second investigator, using a dedicated software (Automated Vascular Analysis ver. 1.0; Academic Medical Center, University of Amsterdam, The Netherlands). We have considered the average values of three sites for each timepoint. Considering the beginning of the fluid challenge, four set of measurements have been considered: the video sequence before the fluid challenge, the two video sequences that preceded this one (i.e., 10 min and 20 min before), and the video sequence after the fluid challenge.

Four microcirculatory parameters have been analyzed (illustrative examples of videos of normal and impaired microcirculatory flow appear in the Supplemental Digital Content 1, https://links.lww.com/ALN/B902, and Supplemental Digital Content 2, https://links.lww.com/ALN/B903, respectively.):

1. The microvascular flow index, which is a qualitative evaluation of the microvascular flow. The image is divided into four quadrants, and the predominant type of flow in very small vessels (i.e., diameter less than 20 μm) is assessed in each quadrant using an ordinal score (0 = no flow, 1 = intermittent flow, 2 = sluggish flow, 3 = normal flow). The overall score, called microvascular flow index, is the sum of each quadrant score divided by the number of quadrants.^13,28 

2. The percentage of perfused vessels, which is calculated as follows: 100 × (total number of vessels – [no flow + intermittent flow])/total number of vessels.

3. The perfused vessel density (which can also be referred to as functional capillary density), which is calculated automatically by dividing the area of perfused vessels by the total area of interest using the Automated Vascular Analysis software.

4. The heterogeneity index, which is calculated as follows: (highest site microvascular flow index – lowest site microvascular flow index) divided by the mean of the microvascular flow index of all sublingual sites.

Anesthesia was performed according to our local protocol. Propofol was used as the induction agent, and then sevoflurane or desflurane were used for maintenance. Intravenous sufentanil was used for intraoperative analgesia. Atracurium or succinylcholine was used for intubation. Atracurium boluses were administered thereafter during the surgery. The minimum sevoflurane or desflurane inspired concentration was used according to the Bispectral Index. Sevoflurane or desflurane inspired concentration was left unchanged during fluid challenges. Intraoperative fluid management consisted of a basal infusion of lactated Ringer’s solution at a rate of 4 ml · kg^–1 · h^–1 to cover evaporative losses and basal fluid requirements. The local hemodynamic protocol was based on pulse pressure variation, cardiac index, and mean arterial pressure^29,30  (Supplemental Digital Content 3; https://links.lww.com/ALN/B901). Preload was optimized by fluid administration until a value of pulse pressure variation less than 13% was obtained. Fluid challenge was performed using a volume of 500 ml of isotonic saline solution (0.9%) in less than 10 min. If the cardiac index value was less than 2.5 l · min^–1 · m^–2 without any improvement after fluid challenge, dobutamine was administered to reach this minimum cardiac index value. If pulse pressure variation and cardiac output were within the target range but mean arterial pressure was still less than 65 mmHg, norepinephrine was administered. Concerning the ventilator settings, the V_T was set at 8 ml/kg⁻¹ of ideal body weight and the respiratory rate were set to keep end-tidal carbon dioxide between 36 and 41 mmHg. The following pulse pressure variation validity criteria were used: regular cardiac rhythm (no arrhythmia or extrasystoles), controlled mechanical ventilation in the absence of spontaneous breathing and with V_T values between 8 and 10 ml/kg⁻¹ of ideal body weight and absence of acute cor pulmonale.

A sample size equal to n = 17 patients was chosen to provide an 80% power to detect by a Wilcoxon test any change in microvascular parameter associated to preload dependence with a standardized effect size equal to 1 with a nominal P value at 0.0166. The methods of Shieh et al.³¹  were used for regarding power calculation. Considering the pilot nature of the current study, we did not adjust the four parameters tested for multiple comparisons, but for each parameter, Bonferonni correction for multiple comparisons was used to test difference between times, and a nominal P value less than 0.0166 was required to claim statistical significance.

Changes in micro- and macrovascular parameters were analyzed 20 min and 10 min before, at the time of the preload dependence, and after fluid challenge completion. For those patients with more than one preload dependence episode, these episodes were averaged. The variations with time of the microvascular parameters (microvascular flow index, percentage of perfused vessels, perfused vessel density, and heterogeneity index) and macrovascular parameters (mean arterial pressure, pulse pressure variation, heart rate, and stroke volume index) were analyzed by comparing the values corresponding to the time of preload dependence with the three other times recorded using t test for matched pairs for macrovascular parameters with Gaussian distribution (macrovascular parameters) or Wilcoxon signed-rank test for microvascular parameters that have non-Gaussian distribution. To estimate correlation between microvascular parameters and pulse pressure variation or stroke volume index, we used a mixed model for assessing correlation in the presence of replication³²  after rank transformation since data were collected at several times for the same patients and since microvascular parameters were non-Gaussian. All results are shown as mean ± SD for variables with Gaussian distributions and median (interquartile range) for variables with non-Gaussian distributions. Tests were performed using SAS version 9.4 (SAS Institute, USA).

Seventeen consecutive patients were included in the study over a 3-month period (flow chart in Supplemental Digital Content 3, https://links.lww.com/ALN/B901). Every patient agreed to participate in the study after oral and written information. Among the 17 patients, one patient did not receive any fluid challenge and was therefore excluded.

Mean age was 59 ± 21 yr, mean American Society of Anesthesiologists score was 2 ± 1 and mean New York Heart Association score was 2 (±1). Patients’ demographic data are reported in Supplemental Digital Content 3, Table 1, https://links.lww.com/ALN/B901.

Preload dependence episodes occurred 32 times. Thirty-two fluid challenges were consequently performed. The median number of fluid challenges per patient was 2 ± 1 (distribution of preload dependence episodes in the Supplemental Digital Content 3, https://links.lww.com/ALN/B901). The average total volume of fluid received per patient was 1,678 ± 1,027 ml. No patient received norepinephrine or erythrocytes during the surgery. Occurrence of preload dependence was concomitant with a significant reduction of the mean arterial pressure from 83 ± 15 mmHg 20 min before fluid challenge to 72 ± 9 mmHg at the time of fluid challenge (P = 0.016) and stroke volume index 43 ± 8 ml/m² 20 min before fluid challenge to 36 ± 8 ml/m² at the time of fluid challenge (P < 0.001) with no significant change in heart rate (figs. 1 and 3). Fluid challenge significantly decreased pulse pressure variation with an increase in stroke volume index and had no effects on mean arterial pressure and heart rate (fig. 1).

As shown in figures 2 and 3, most of the microvascular parameters decreased significantly during the preload dependence episodes. Microvascular flow index decreased from a median (interquartile range) of 2.8 (2.6, 2.9) 20 min before fluid challenge to 2.3 (1.8, 2.8) at the time of fluid challenge (P = 0.009), and perfused vessel density from 16.9 mm/mm² (14.7, 21.4) to 14.9 mm/mm² (12.0, 16.4; P = 0.009), while heterogeneity index was increased from 0.2 (0.2, 0.4) to 0.5 (0.4, 0.7; P = 0.001). Percentage of perfused vessels changed from 91.6% (86.1, 97.8) to 76.5% (66.1, 94.0) and did not reach the significance level adjusted for multiplicity. After fluid challenge, all microvascular parameters changed significantly: the microvascular flow index increased to 2.8 (2.6, 3.0; P = 0.009), the percentage of perfused vessels increased to 92.9% (89.8, 99.4; P < 0.002), and the perfused vessel density increased to 18.6 mm/mm² (16.2, 21.8; P < 0.001), while the heterogeneity index significantly decreased to 0.2 (0.1, 0.4; P< 0.001). Note that when considering values 10 min before fluid challenge, before fluid challenge, and after fluid challenge, microvascular flow index, perfused vessel density, and heterogeneity index were significantly correlated to pulse pressure variation values (r = –0.253, P = 0.041; r = –0.310, P = 0.013; r = 0.317, P = 0.001, respectively). There was no significant correlation between mean arterial pressure or stroke volume index and microvascular parameters.

Our findings suggest that sublingual microcirculation was reduced when preload dependence (pulse pressure variation greater than 13%) occurred during major abdominal surgery. Indeed, sublingual microcirculatory variables such as microvascular flow index and perfused vessel density were significantly decreased when preload dependence occurred. Fluid challenge decreased pulse pressure variation and restored microvascular variables but did not restore mean arterial pressure.

In the current study, only sublingual microcirculation was observed, but several studies demonstrated that sublingual microcirculatory changes were correlated to gut^33–37  and renal^38,39  microcirculatory changes. In a large animal model of sepsis, the severity and the time course of microcirculatory changes were similar in sublingual and intestinal microcirculation.³³  Similar results were reported in a sheep model of hemorrhagic shock.³⁶  de Bruin et al. have reported in patients undergoing elective gastrointestinal surgery that gut serosal microvascular imaging with a sidestream dark field device is similar to sublingual assessment.³⁷  Recently, Lima et al.,³⁹  in a lipopolysaccharide-induced shock model, demonstrated that sublingual microcirculation could reflect renal microvascular alterations detected by contrast-enhanced ultrasonography during shock and fluid resuscitation. These results support our belief that the sublingual microcirculation is representative of microcirculation of other organs and especially of vital organs.

The current study assessed the relationship between preload dependence, which refers to the ability of the heart to change its stroke volume in response to a change in its preload, and the sublingual microcirculation during major abdominal surgery. The goal of fluid resuscitation is to restore or maintain oxygen delivery to the tissues. The patterns of microcirculation are difficult to predict solely on the basis of changes in macrocirculatory variables. Indeed, in the context of decreased cardiac output, the physiologic microvascular response primarily aims to preserve tissue perfusion. However, it has been demonstrated that microcirculation could be still altered despite optimization of macrocirculation in shock.⁴⁰  This is due to specific microcirculation alterations observed during shock such as viscosity changes, endothelial dysfunction, and glycocalyx or erythrocytes alterations.⁴¹  The degree of microcirculatory alterations may vary with the intensity of shock and of inflammation and may also vary over time. In this regard, in the early phase of septic shock, improvement of microcirculation was observed after fluid infusion, whereas in the late phase of septic shock, optimization of macrocirculation was not always associated with improvement in microcirculation.⁴²  Persistence of sublingual microcirculatory disorders over time was shown to be associated with a poor outcome.^14–16,43 

In our study, sublingual microcirculatory variables were assessed in patients without prior hemodynamic instability and microvascular impairment. It is interesting to note that sublingual microcirculation was impaired when episodes of preload dependence occurred, which were associated with significant decreases in stroke volume index and mean arterial pressure. With the observed alterations of sublingual microcirculation being contemporaneous with increases in pulse pressure variation (greater than 13%), we can link them to changes in preload. These changes in preload could be due to hypovolemia and/or to a decrease in venous tone with venous dilation. The fact that fluid challenge was able to restore microcirculatory alterations pleads for hypovolemia. Simultaneous changes in stroke volume index and sublingual microcirculatory variables during episodes of preload dependence as well as during fluid challenges suggest there was no dissociation between microcirculation and microcirculation at the level of the sublingual territory. Whether this is related to the use of anesthetic drugs cannot be elucidated by our study.

Our study suggests that episodes of preload dependence have an impact on sublingual microcirculation (and maybe on other sites) during major abdominal surgery. It is important to realize that the presence of fluid responsiveness is not an absolute indication to give fluids. The decision to administer fluid therapy must be supported by the need for hemodynamic improvement. The fact that microvascular alterations were concomitant with episodes of preload dependence is an additional argument to guide fluid resuscitation using dynamic variables (especially pulse pressure variation)³⁰  and suggests that it is important to correct preload dependence occurring during surgery to avoid microvascular alterations. However, when pulse pressure variation cannot be used (arrhythmia, V_T values less than 8 ml/kg⁻¹ of ideal body weight, spontaneous breathing, or acute cor pulmonale) or when pulse pressure variation value is in the gray zone (between 9% and 13%), microvascular sublingual parameters could serve as additional indicators when deciding to administer fluids. Carbon dioxide monitoring appears to be another potential approach. Ospina-Tascón et al. have reported that venous-to-arterial carbon dioxide difference could reflect the adequacy of microvascular blood flow.⁴⁴  Use of sublingual carbon dioxide monitoring is still to be validated.⁴⁵ 

The fact that we observed no change in mean arterial pressure after fluid challenge in spite of correction of stroke volume index suggests that a change in arterial tone occurred during fluid challenge. In critically ill patients with septic shock, fluid administration significantly reduced arterial load (i.e., the external opposition that must be overcome by the ventricles during ejection), even when increasing cardiac output.⁴⁵  This explains the increase in cardiac output after fluid administration with no improvement in blood pressure for some septic patients. Monge García et al. reported on septic shock that in a preload responder group, only 44% of the patients were pressure responders.⁴⁶  This fluid-induced reduction in arterial load may have resulted from (1) vasoplegia induced by anesthetic drugs, (2) change in arterial tone due to inflammation, (3) reduction of arterial tone via flow-mediated vascular relaxation induced by fluid administration (increase of nitric oxide production and endothelial shear stress), and (4) recruitment of previously closed vessels with an increase of the effective diameter of arterial system and, consequently, reduction in arterial resistance.

Nevertheless, our observational study has several limits. The retrospective microvascular analysis was conducted by a blinded investigator, but video acquisition was performed by a nonblinded investigator. Moreover, it can be argued that the number of patients included in this study is too small to make any conclusion. In addition, it should be stressed that since this is a pilot study, the absence of significant differences for some study parameters may be due to the small sample size. Most patients received β blockers or calcium blockers (which reflects real life), which can have an impact on the observed results. However, we observed no differences between patients treated by each of these drugs and those not receiving them. But we compared small numbers of patients. As angiotensin-converting enzyme inhibitors were stopped 24 h before surgery, their impact can be minimal. We cannot assess with our data the impact of β blockers and calcium blockers on the observed response.

Preload dependence was associated with reduced sublingual microcirculation during major abdominal surgery. Fluid administration successfully restored microvascular perfusion. This finding suggests immediate correction of preload dependence to avoid reduced microcirculation. In the future, microvascular sublingual parameters could serve as additional indicators when deciding to administer fluids.

Support was provided solely from institutional and/or departmental sources.

The authors declare no competing interests.