Prediction of Fluid Responsiveness Based on the External Jugular Vein Distensibility Index After Changes in Volume Status in Healthy, Anesthetized, and Mechanically Ventilated Dogs

ABSTRACT

Objective

To investigate whether point‐of‐care ultrasound of the external jugular vein (EJV) can predict fluid responsiveness (FR) in healthy, anesthetized, mechanically ventilated dogs.

Design

Prospective, nonrandomized experimental study.

Setting

University‐based small animal research facility.

Animals

Six healthy Beagle dogs.

Interventions

Dogs were investigated at six time points (TPs): baseline (TP₁); 20 mL/kg of circulating blood was collected over 10 min (TP₂); half of the collected blood was autotransfused for 10 min (TP₃); remaining collected blood was autotransfused for 10 min (TP₄); 0.9% normal saline (10 mL/kg for 10 min) was administered (TP₅); and an additional dose of 0.9% normal saline (10 mL/kg for 10 min) was administered (TP₆). Hemodynamic variables, Doppler images of the left ventricular outflow tract (LVOT), and M‐mode images of the EJV were obtained at each TP. FR was evaluated during TP_3–6. FR was defined as an increase of >15% in the LVOT velocity time integral following fluid challenge, while other results were defined as fluid nonresponsiveness (FNR). The external jugular vein distensibility index (EJVDI) was calculated as follows: [(maximal EJV diameter − minimal EJV diameter)/minimal EJV diameter] × 100%. The maximal EJV diameter was measured during inspiration, and the minimal EJV diameter was measured during expiration. In addition, gray zones indicating the range of diagnostic uncertainty were proposed in various indices for predicting FR.

Measurements and Main Results

Among the 24 fluid challenges performed between TP₃ and TP₆, 11 FR and 13 FNR were identified. The area under the receiver operating characteristic curve for the EJVDI in predicting FR was 0.92, with a cut‐ff value of 22.7%, and the gray zone was identified as 22.6%–27.3%.

Conclusions

The EJVDI could be used to predict FR in healthy, anesthetized, mechanically ventilated dogs. Further studies are required before point‐of‐care ultrasound of the EJV can be applied in various clinical settings.

Abbreviations

AUROC

area under the receiver operating characteristic

CO

cardiac output

CVC

cauda vena cava

CVP

central venous pressure

EJV

external jugular vein

EJVDI

external jugular vein distensibility index

ETco ₂

end‐tidal partial pressure of carbon dioxide

FNR

fluid nonresponsiveness

FR

fluid responsiveness

HR

heart rate

IJV

internal jugular vein

IJVDI

internal jugular vein distensibility index

IVC

inferior vena cava

IVCDI

inferior vena cava distensibility index

LVOT

left ventricular outflow tract

PIP

peak inspiratory pressure

POCUS

point‐of‐care ultrasound

PPV

pulse pressure variation

SPV

systolic pressure variation

TP

time point

VTI

velocity time integral

1. Introduction

Fluid therapy is a cornerstone of treatment for critically ill patients; however, indiscriminate fluid administration is associated with various complications, poor prognosis, and increased mortality in both human and veterinary medicine [1, 2, 3]. Studies in human medicine have shown that approximately 50% of critically ill patients are not fluid responsive [4]. This indicates that evaluating a patient's volume status and fluid responsiveness (FR) prior to fluid administration is important, and fluid therapy should be based on these assessments [5, 6].

Although the gold standard for assessing FR is changes in cardiac output (CO) after a fluid challenge, FR has traditionally been assessed using static indices, such as central venous pressure (CVP) and pulmonary artery occlusion pressure. However, recent studies have reported that static indices are more appropriate for evaluating a patient's volume status, but are unreliable at predicting FR [7, 8]. Conversely, dynamic indices such as pulse pressure variation (PPV), systolic pressure variation (SPV), and stroke volume variation, which result from heart–lung interactions, are known to be sensitive and reliable predictors of FR [9, 10]. However, monitoring these indices requires the insertion of an invasive arterial line, and their reliability decreases under conditions such as intraabdominal hypertension and arrhythmia [11, 12].

Point‐of‐care ultrasound (POCUS) has been prioritized for assessing volume status and FR in human intensive care, and research on its clinical application has been reported in veterinary medicine [13, 14]. POCUS offers the advantages of noninvasiveness, rapid real‐time assessment, and the ability to simultaneously evaluate both static and dynamic indices [6, 15]. Recent human studies have demonstrated that POCUS of the inferior vena cava (IVC) can be used to evaluate volume status and FR of patients under mechanical ventilation [16, 17]. Similarly, in dogs, the diameter of the cauda vena cava (CVC) correlates with CVP, whereas the cauda vena cava distensibility index (CVCDI) reliably predicts FR [18, 19]. However, performing POCUS of the IVC or CVC can induce pain in critically ill patients and inadequate ultrasound windows may limit the examination [6, 20, 21]. Therefore, POCUS of the internal jugular vein (IJV) has been suggested in human medicine, and the internal jugular vein distensibility index (IJVDI) can assess FR as effectively as the inferior vena cava distensibility index (IVCDI). In addition, performing POCUS of the IJV is relatively fast and easy, and it is useful in cases where IVCDI application is restricted [22, 23, 24, 25].

The external jugular vein (EJV) is larger than the IJV and functions as the primary route of venous return from the heads of dogs [26]. Therefore, POCUS of the EJV has been proposed as a predictor of FR in dogs^a. Variability in the EJV diameter has been observed in a hypovolemic dog during fluid resuscitation, indicating a potential alternative for evaluating FR^a. Furthermore, normal EJV diameter in healthy dogs under spontaneous breathing has been documented, increasing the potential for clinical application^b. However, whether the external jugular vein distensibility index (EJVDI) could predict FR in mechanically ventilated dogs remains unclear.

In this context, the present study aimed to evaluate (1) whether the EJVDI could be used as a dynamic index for predicting FR and (2) whether EJV diameter is associated with the patient's volume status. Additionally, the correlations between EJVDI and PPV and SPV, as well as the association between EJV diameter and CVP, were investigated. The hypotheses were (1) EJVDI could predict FR and (2) EJV diameter could reflect volume status in healthy, anesthetized, and mechanically ventilated dogs with experimentally induced changes in volume status.

2. Materials and Methods

2.1. Animals

This prospective, nonrandomized experimental study was approved by the Ethics Committee for Experimental Animals of Seoul National University (SNU‐230725‐3). Three intact male and three intact female Beagle dogs aged 20–21 months were used as subjects in this study. All dogs were confirmed to be healthy based on physical examination, CBC, serum biochemistry panel, and thoracic and abdominal radiographic examinations.

2.2. Anesthesia and Instrumentation

All dogs were fasted for 12 h, and water was restricted for 4 h prior to anesthetic induction. A 24‐gauge IV catheter^c was placed aseptically in the cephalic vein, and anesthesia was induced with alfaxalone^d (2 mg/kg, IV) 5 min following preoxygenation using a face mask. After endotracheal intubation, the dog was placed in dorsal recumbency. Anesthesia was maintained with isoflurane^e in 100% oxygen at a flow rate of 2 L/min, and the end‐tidal concentration of isoflurane was regulated between 1.6% and 1.8%. Rocuronium^f was administered (0.4 mg/kg, IV), followed by a continuous‐rate infusion (0.4 mg/kg/h) to suppress spontaneous breathing. Volume‐controlled mechanical ventilation was initiated, and the tidal volume was set to 13 mL/kg with a positive end‐expiratory pressure of 3 cmH₂O. The inspiratory to expiratory ratio was 1:2, and the respiratory rate was controlled between 10 and 20 breaths/min to maintain the end‐tidal partial pressure of carbon dioxide (ETco ₂) at 35–45 mmHg. The end‐tidal concentration of isoflurane, peak inspiratory pressure (PIP), positive end‐expiratory pressure, ETco ₂, respiratory rate, and tidal volume were monitored using an infrared gas analyzer integrated into a multiparameter monitor^g.

A 24‐gauge IV catheter was inserted aseptically into the dorsal pedal artery to monitor invasive arterial blood pressure. A 5‐Fr double‐lumen catheter^h was placed aseptically in the left jugular vein to measure CVP and collect circulating blood. Both catheters were connected to a pressure transducer, allowing the continuous monitoring of MAP and mean CVP using a multiparameter monitor^g. After the instrumentation was completed, the dog was positioned in left lateral recumbency, and the pressure transducers were positioned at the estimated level of the right atrium and calibrated to atmospheric pressure. Catheters were periodically flushed with heparinized saline. Maintenance fluids were not administered as 20 mL/kg of fluid was additionally administered (10 mL/kg, two times) during the experimental procedure. PPV, SPV, heart rate (HR), esophageal temperature, and SpO₂ were also monitored on the same monitor. Esophageal temperature was maintained between 37.0°C and 38.0°C using a forced‐air warming blanketⁱ.

2.3. Ultrasonographic Measurement of the Left Ventricular Outflow Tract (LVOT) Velocity Time Integral (VTI) and EJVDI

Ultrasound examinations to obtain images of the LVOT and EJV were performed by two operators, with each operator consistently performing one specific examination to minimize the potential for interobserver variability. The operators have been performing emergency ultrasound examinations, including echocardiography, in the emergency room for over 4 years. Transthoracic echocardiography was performed using an ultrasound system^j equipped with a 2–9 MHz phased‐array probe with the dogs positioned in left lateral recumbency. Waveforms in the LVOT were obtained at end expiration in the left apical five‐chamber view using pulse‐wave Doppler at the level of the aortic valve. The sampling gate was placed just proximal to the aortic valve leaflets, and its appropriate placement was confirmed by observing the valve closure click [27]. Doppler images of the LVOT were saved on an ultrasound machine for subsequent analyses.

Subsequently, right EJV examination was performed using a 4–16 MHz linear array probe on the same ultrasound machine at the midpoint between the ramus and thoracic inlet. The probe was then placed perpendicular to the skin, and the shape of the EJV was identified in the transverse plane using the B‐mode at minimal depth [22, 24]. Images and cineloops were recorded in the M‐mode, with minimal pressure applied to avoid compressing the EJV during the respiratory cycle (Figure 1).

FIGURE 1.

Ultrasonographic measurement of external jugular vein distensibility index (EJVDI) in healthy, anesthetized, mechanically ventilated dogs. The maximal and minimal diameters of the external jugular vein were measured to obtain EJVDI. D, distance.

Because the ultrasound operators were not blinded to the experimental procedures, image analysis was conducted using the stored images following the completion of all experiments. To minimize potential bias, especially considering that the volume status changed in a nonrandomized sequence, the stored images were randomized before being analyzed by the operator who performed the respective examination.

The LVOT VTI was measured by manually tracing the Doppler flow signal using electronic calipers incorporated in the ultrasound machine software. LVOT VTI was measured through the averaging of three consecutive waveforms. EJV diameter values were measured using the inner‐to‐inner‐edge method with the same electronic calipers [27]. The EJV diameter during inspiration (EJV_d‐max) and expiration (EJV_d‐min) was obtained over three consecutive respiratory cycles, and the average values for each were calculated. Using these averages, the EJVDI was calculated as follows: [(EJV_d‐max − EJV_d‐min)/EJV_d‐min] × 100% [24]. Measurements were repeated on the stored images using the same method to obtain intraobserver variability for the LVOT VTI and EJVDI.

2.4. Experimental Design

After all instrumentation was completed, hemodynamic variables were stabilized for 10 min (Figure 2). The experimental procedures included a total of six time points (TPs): TP₁, baseline; TP₂, 20 mL/kg of circulating blood was manually collected over 10 min through a central venous catheter into 50‐mL syringes, prefilled with an appropriate amount of citrate‐phosphate‐dextrose‐adenine, and then transferred from the syringes into a blood transfusion bag; TP₃, half of the collected blood was autotransfused (10 mL/kg for 10 min) through the central venous catheter using an infusion pump^k with an 18‐µm filter^l; TP₄, remaining collected blood was autotransfused (10 mL/kg for 10 min) in the same manner; TP₅, 0.9% normal saline (10 mL/kg for 10 min) was administered through the central venous catheter using an infusion pump^m; and TP₆, an additional dose of 0.9% normal saline (10 mL/kg for 10 min) was administered in the same manner. Between each TP, a 5‐min resting period was allocated to stabilize the hemodynamic status prior to data collection. At each TP, the HR, MAP, mean CVP, PPV, SPV, SpO₂, ETco ₂, tidal volume, and PIP were recorded. Subsequently, Doppler images of the LVOT and M‐mode images of the EJV were obtained. After the experiment was completed, FR was evaluated for a total of 24 fluid challenges in six dogs from TP₃ to TP₆. FR was defined as an increase of >15% in LVOT VTI following autotransfusion or fluid administration compared to the previous TP, otherwise fluid nonresponsiveness (FNR) was defined.

FIGURE 2.

Time points (TPs) of the experimental design and data collection in healthy, anesthetized, mechanically ventilated dogs. The volume status of each dog was changed in a stepwise, nonrandomized manner as follows: TP₁, baseline; TP₂, 20 mL/kg of circulating blood was collected over 10 min; TP₃, half of the collected blood was autotransfused for 10 min; TP₄, remaining collected blood was autotransfused for 10 min; TP₅, 10 mL/kg of normal saline was administered for 10 min; and TP₆, an additional dose of 10 mL/kg of normal saline was administered for 10 min. Between each TP, a 5‐min resting period was allocated to stabilize the hemodynamic status prior to data collection. At each TP, hemodynamic variables and Doppler images of the LVOT and M‐mode images of the EJV were obtained. HR, heart rate; CVP, central venous pressure; PPV, pulse pressure variation; SPV, systolic pressure variation; ETco ₂, end‐tidal partial pressure of carbon dioxide; V_T, tidal volume; PIP, peak inspiratory pressure; LVOT, left ventricular outflow tract.

2.5. Recovery From Anesthesia

Following completion of the experimental procedures, the arterial and jugular venous catheters were removed, and meloxicamⁿ (0.2 mg/kg, SC) was administered for analgesia. Rocuronium administration was discontinued and reversed with neostigmine^o (0.08 mg/kg, IV). Simultaneously, glycopyrrolate^p (0.01 mg/kg, IV) was administered to counteract the side effects of neostigmine. The vaporizer was turned off, and recovery was initiated. Following extubation, each dog was transferred to an individual kennel. Hemodynamic variables and catheter sites were monitored periodically for 24 h.

2.6. Statistical Analysis

All analyses were performed using commercially available software^{q , r , s}. The normality of hemodynamic variable data was assessed using the Shapiro–Wilk test and visual inspection with Q–Q plots. Data are presented as the mean and SD or median and interquartile range, depending on the distribution. A linear mixed model or a generalized linear mixed model was applied depending on the normality of the data. These models were used to (1) investigate changes in hemodynamic variables during blood removal and fluid challenge, with TPs included as a fixed effect and individual dogs as a random effect, and (2) analyze differences in hemodynamic variables both within and between groups, with time (before and after the fluid challenge) and group (fluid responders and nonresponders) included as fixed effects and individual dogs as a random effect. Bonferroni post hoc tests were performed to compare differences: (1) between TP_2–6 and TP₁, as well as between TP_3–6 and TP₂; and (2) within each group over time and between groups.

Since the same dogs could be classified as both fluid responders and nonresponders across repeated fluid challenges, a generalized linear mixed model was applied to estimate predicted probabilities for each variable (EJVDI, PPV, SPV, mean CVP, EJV_d‐max, and EJV_d‐min) in predicting FR. Group (fluid responders and nonresponders) was included as a fixed effect, while individual fluid challenges were modeled as a random effect. Receiver operating characteristic curves were generated using the predicted probabilities to assess the ability of the dynamic (EJVDI, PPV, and SPV) and static (mean CVP, EJV_d‐max, and EJV_d‐min) indices to discriminate between FR and FNR. The area under the receiver operating characteristic (AUROC) curves were compared using the De Long method [28]. The optimal cutoff value of each variable was estimated by maximizing the Youden index [sensitivity + (specificity − 1)]. The gray zones were obtained as follows: (1) 95% confidence intervals of the Youden index were calculated from a 1000 bootstrap population; (2) the range of cutoff values corresponding to <90% sensitivity and <90% specificity was calculated (zone of poor diagnostic accuracy); and the largest of these intervals was chosen as the gray zone [29].

The correlation between variables was evaluated using Pearson's correlation coefficient or Spearman's rank correlation depending on the distribution. Bland–Altman analysis was used to obtain intraobserver variability of the LVOT VTI and EJVDI. Mean differences and 95% limits of agreement were calculated.

Since the AUROC for IJVDI in predicting FR is reported as 0.88 in human medicine [22], we established a threshold of 0.85 for EJVDI. Given the repeated fluid challenges in healthy dogs, we expected a similar number of fluid responders and nonresponders [14, 30]. Based on this assumption, a priori sample size calculations indicated that a minimum of 18 fluid challenges was required to obtain a significant AUROC curve for EJVDI in predicting FR with a discriminatory accuracy (>0.85, alpha 0.05, power 0.8). Statistical significance was set at p < 0.05.

3. Results

During the experimental procedure, the RR was maintained at 12 breaths per min, with V_T between 134 and 143 mL, PIP between 13 and 15 cmH₂O, SpO₂ between 98% and 100%, and ETco ₂ between 38 and 43 mmHg (Table 1). Among the 24 fluid challenges performed between TP₃ and TP₆, 11 FR and 13 FNR were identified (TP₃: 6 FR; TP₄: 4 FR and 2 FNR; TP₅: 1 FR and 5 FNR; TP₆: 6 FNR).

TABLE 1.

Hemodynamic variables collected in healthy, anesthetized, mechanically ventilated dogs at six time points (TPs) in a stepwise, nonrandomized manner: TP₁, baseline; TP₂, 20 mL/kg of circulating blood was collected over 10 min; TP₃, half of the collected blood was autotransfused for 10 min; TP₄, remaining collected blood was autotransfused for 10 min; TP₅, 10 mL/kg of normal saline was administered for 10 min; and TP₆, an additional dose of 10 mL/kg of normal saline was administered for 10 min. Data are presented as means ± SD or as median (IQR).

Variables	TP1	TP2	TP3	TP4	TP5	TP6
LVOT VTI (cm)	9.63 ± 0.89	5.66 ± 1.10 a	9.46 ± 0.76 b	10.80 ± 1.30 b	11.70 ± 1.59 a , b	12.40 ± 1.41 a , b
EJVDI (%)	19.2 ± 7.2	36.6 ± 9.4 a	26.0 ± 11.0 b	22.3 ± 6.5 b	15.4 ± 4.0 b	12.0 ± 3.4 b
PPV (%)	9 (7–11)	28 (21–31) a	11 (9–15) b	7 (6–9) b	7 (6–9) b	7 (6–7) b
SPV (mmHg)	5 ± 1	9 ± 2 a	6 ± 1 b	5 ± 1 b	5 ± 1 b	5 ± 1 b
EJVd‐max (mm)	4.51 ± 0.41	3.73 ± 0.71	4.35 ± 0.77	4.62 ± 0.72	4.90 ± 0.46 b	5.26 ± 0.37 b
EJVd‐min (mm)	3.79 ± 0.29	2.75 ± 0.62 a	3.47 ± 0.69	3.80 ± 0.71 b	4.25 ± 0.37 b	4.70 ± 0.34 a , b
Mean CVP (mmHg)	4.5 ± 1.2	1.7 ± 1.2 a	4.0 ± 0.9 b	5.2 ± 1.6 b	6.2 ± 1.7 a , b	6.5 ± 1.5 a , b
MAP (mmHg)	62 ± 6	47 ± 6 a	61 ± 5 b	69 ± 8 b	70 ± 11 b	71 ± 12 b
VT (mL/breath)	134 ± 10	136 ± 8	135 ± 7	143 ± 12	140 ± 12	139 ± 14
PIP (cmH2O)	13 ± 1	14 ± 1	14 ± 1	14 ± 1 a	14 ± 1 a	14 ± 1 a

Abbreviations: CVP, central venous pressure; EJVDI, external jugular vein distensibility index; EJV_d‐max, maximal diameter of the external jugular vein; EJV_d‐min, minimal diameter of the external jugular vein; IQR, interquartile range; LVOT VTI, left ventricular outflow tract velocity time integral; PIP, peak inspiratory pressure; PPV, pulse pressure variation; SPV, systolic pressure variation; V_T, tidal volume.

^a Significant difference compared to TP₁ (p < 0.05).

^b Significant difference compared to TP₂ (p < 0.05).

3.1. Changes in Hemodynamic Variables During Blood Removal and Fluid Challenge

LVOT VTI significantly decreased at TP₂ (Table 1; p < 0.001) following blood removal. Similarly, MAP and mean CVP at TP₂ were significantly lower than at TP₁ (p = 0.003 and p < 0.001, respectively). At TP₂, EJV_d‐min was significantly lower than that at TP₁ (p = 0.007). EJVDI was significantly higher at TP₂ compared to TP₁ (p < 0.001). Similarly, PPV and SPV were significantly higher at TP₂ (both p < 0.001). From TP₃ to TP₆, the fluid challenge resulted in significant increases in LVOT VTI, EJV_d‐max, EJV_d‐min, MAP, and mean CVP compared to TP₂. In contrast, EJVDI, PPV, and SPV significantly decreased during this period. No significant differences in hemodynamic variables were observed between TP₄ and TP₁. At TP₆, LVOT VTI and mean CVP were significantly higher than those at TP₁ (p < 0.001 and p = 0.009, respectively). Similarly, EJV_d‐min at TP₆ was significantly higher than that at TP₁ (p = 0.023). However, no significant changes were observed in the EJVDI, PPV, or SPV between TP₆ and TP₁.

3.2. Comparison of Hemodynamic Variables Between FR and FNR Groups

LVOT VTI increased significantly by approximately 39% (from 7.68 to 10.70 cm) in the FR group and 5% (from 10.90 to 11.40 cm) in the FNR group following the fluid challenge (p < 0.001 and p = 0.018, respectively). A significant difference in LVOT VTI between the FR and FNR groups was observed before the fluid challenge (p < 0.001); however, this difference was not observed after the fluid challenge (See Table 2)

TABLE 2.

Effects of a fluid challenge on hemodynamic variables collected in healthy, anesthetized, mechanically ventilated dogs: comparisons were made within group (before vs. after fluid challenge) and between group ([FR, n = 11] vs. [FNR, n = 13]). Data are presented as means ± SD or as median (IQR).

Variable	FR group		FNR group
	Before fluid challenge	After fluid challenge	Before fluid challenge	After fluid challenge
LVOT VTI (cm)	7.68 ± 2.58 b	10.70 ± 1.70 a	10.90 ± 1.59	11.40 ± 1.57 a
EJVDI (%)	33.4 ± 10.4 b	24.5 ± 9.1 a , b	18.0 ± 4.7	14.2 ± 4.3 a
PPV (%)	19 (10–28) b	9 (7–12) a	7 (6–10)	7 (6–8)
SPV (mmHg)	7 (6–8) b	5 (5–6) a	5 (4–6)	5 (4–5)
EJVd‐max (mm)	4.01 ± 0.76 b	4.55 ± 0.69 a	4.73 ± 0.63	4.98 ± 0.58
EJVd‐min (mm)	3.03 ± 0.70 b	3.68 ± 0.69 a , b	4.01 ± 0.58	4.37 ± 0.55 a
Mean CVP (mmHg)	2.9 ± 1.8 b	4.6 ± 1.5 a , b	5.4 ± 1.8	6.2 ± 1.6 a
MAP (mmHg)	55 ± 11 b	65 ± 7 a	68 ± 10	70 ± 12

Abbreviations: CVP, central venous pressure; EJVDI, external jugular vein distensibility index; EJV_d‐max, maximal diameter of the external jugular vein; EJV_d‐min, minimal diameter of the external jugular vein; FNR, fluid nonresponsiveness; FR, fluid responsiveness; IQR, interquartile range;LVOT VTI, left ventricular outflow tract velocity time integral; PPV, pulse pressure variation; SPV, systolic pressure variation.

^a Significant difference from the time point recorded before the fluid challenge (p < 0.05).

^b Significant difference between FR and FNR group at the same time point (p < 0.05).

The EJVDI decreased significantly in both the FR and FNR groups following fluid challenge (p < 0.001 and p = 0.017, respectively). A significant difference in EJVDI between the FR and FNR groups was observed before the fluid challenge (p < 0.001) and was maintained after the fluid challenge (p = 0.002). PPV and SPV decreased significantly following fluid challenge only in the FR group (both p < 0.001). A significant difference in PPV and SPV between the FR and FNR groups was observed before the fluid challenge (both p < 0.001); however, this difference was not observed after the fluid challenge.

The EJV_d‐max increased significantly only in the FR group (p = 0.01), whereas the EJV_d‐min increased significantly in both the FR and FNR groups (p < 0.001 and p = 0.026, respectively) following fluid challenge. Before the fluid challenge, significant differences were found between the FR and FNR groups for EJV_d‐max and EJV_d‐min (p = 0.012 and p < 0.001, respectively); however, only EJV_d‐min remained significantly different after the fluid challenge (p = 0.011). The mean CVP increased significantly in both the FR and FNR groups following fluid challenge (p < 0.001 and p = 0.009, respectively). A significant difference in the mean CVP between the FR and FNR groups was observed before the challenge (p = 0.001) and was maintained after the fluid challenge (p = 0.035). MAP increased significantly only in the FR group (p < 0.001), and no significant difference in MAP was observed between the FR and FNR groups after fluid challenge.

3.3. AUROC Curve Analysis for Dynamic (EJVDI, PPV, and SPV) and Static (Mean CVP, EJV_d‐max, and EJV_d‐min) Indices

Both the dynamic and static indices showed a predictive ability for FR; however, despite the differences in the AUROC curves, statistical significance was not observed. The AUROC curves for the EJVDI, PPV, and SPV were 0.92, 0.89, and 0.88, respectively. The cutoff values for distinguishing FR from FNR were 22.7% (sensitivity, 90.9%; specificity, 92.3%) for EJVDI, 9% (sensitivity, 81.8%; specificity, 76.9%) for PPV, and 6 mmHg (sensitivity, 81.8%; specificity, 92.3%) for SPV (Figure 3; Table 3).

FIGURE 3.

Comparison of the areas under the receiver operating characteristic curves for (A) dynamic (EJVDI, PPV, and SPV) and (B) static (mean CVP, EJV_d‐max, and EJV_d‐min) indices at predicting fluid responsiveness in healthy, anesthetized, mechanically ventilated dogs. EJVDI, external jugular vein distensibility index; PPV, pulse pressure variation; SPV, systolic pressure variation; CVP, central venous pressure; EJV_d‐max, maximal diameter of the external jugular vein; EJV_d‐min, minimal diameter of the external jugular vein.

TABLE 3.

Diagnostic performance of the dynamic (EJVDI, PPV, and SPV) and static (mean CVP, EJV_d‐max, and EJV_d‐min) indices at discriminating FR (n = 11) and FNR (n = 13).

Variables	AUROC (95% CI)	Cutoff value	Gray zone	Sensitivity (95% CI)	Specificity (95% CI)	p‐value
Dynamic indices
EJVDI (%)	0.92 (0.74–0.99)	22.7	22.6–27.3	90.9 (58.7–99.8)	92.3 (64.0–99.8)	<0.001
PPV (%)	0.89 (0.69–0.98)	9	6–17	81.8 (48.2–97.7)	76.9 (46.2–95.0)	<0.001
SPV (mmHg)	0.88 (0.68–0.98)	6	5–6	81.8 (48.2–97.7)	92.3 (64.0–99.8)	<0.001
Static indices
Mean CVP (mmHg)	0.83 (0.63–0.95)	4	3–6	63.6 (30.8–89.1)	84.6 (54.6–98.1)	<0.001
EJVd‐max (mm)	0.85 (0.64–0.96)	4.27	3.60–5.07	81.8 (48.2–97.7)	76.9 (46.2–95.0)	<0.001
EJVd‐min (mm)	0.84 (0.63–0.96)	3.81	3.05–4.43	81.8 (48.2–97.7)	76.9 (46.2–95.0)	<0.001

Abbreviations: AUROC, area under the receiver operating characteristic; CI, confidence interval; CVP, central venous pressure; EJVDI, external jugular vein distensibility index; EJV_d‐max, maximal diameter of the external jugular vein; EJV_d‐min, minimal diameter of the external jugular vein; FNR, fluid nonresponsiveness; FR, fluid responsiveness; PPV, pulse pressure variation; SPV, systolic pressure variation.

The AUROC curves for the mean CVP, EJV_d‐max, and EJV_d‐min were 0.83, 0.85, and 0.84, respectively. The cutoff values differentiating FR from FNR were 4 mmHg (sensitivity, 63.6%; specificity, 84.6%) for mean CVP, 4.27 mm (sensitivity, 81.8%; specificity, 76.9%) for EJV_d‐max, and 3.81 mm (sensitivity, 81.8%; specificity, 76.9%) for EJV_d‐min.

The gray zone for EJVDI was 22.6%–27.3%, including five of the 24 fluid challenges (Figure 4; Table 3). The gray zones for PPV, SPV, mean CVP, EJV_d‐max, and EJV_d‐min were 6%–17%, 5–6 mmHg, 3–6 mmHg, 3.60–5.07 mm, and 3.05–4.43 mm, respectively. Of the 24 fluid challenges, 18 were in the PPV gray zone, 13 in the SPV gray zone, 17 in the mean CVP gray zone, 15 in the EJV_d‐max gray zone, and 16 in the EJV_d‐min gray zone. Based on the gray zone approach, when the EJVDI is below 22.6%, PPV is less than 6%, and SPV is below 5 mmHg, as well as when CVP exceeds 6 mmHg, EJV_d‐max is greater than 5.07 mm, and EJV_d‐min exceeds 4.43 mm, the likelihood of FNR is high.

FIGURE 4.

Gray zone (gray‐shaded region) of the EJVDI for discriminating between FR and FNR status in healthy, anesthetized, mechanically ventilated dogs. Plots of sensitivity (green line) and specificity (blue line) at various cutoff values are presented. EJVDI, external jugular vein distensibility index.

3.4. Correlations Between Dynamic (EJVDI, PPV, and SPV) and Static (Mean CVP, EJV_d‐max, and EJV_d‐min) Indices

Significant correlations were observed between the EJVDI and PPV (r = 0.73, p < 0.001) and SPV (r = 0.64, p < 0.001). The mean CVP significantly correlated with both EJV_d‐max (r = 0.47, p = 0.004) and EJV_d‐min (r = 0.51, p = 0.002). In addition, significant correlations were found between PPV and SPV (r = 0.77, p < 0.001) and between EJV_d‐max and EJV_d‐min (r = 0.95, p < 0.001) (Supplementary Figure S1).

3.5. Intraobserver Variability in LVOT VTI and EJVDI

The Bland–Altman analysis of intraobserver variability revealed a bias of LVOT VTI of −0.05 cm, with 95% limits of agreement ranging from −0.67 to 0.57 cm. The bias of the EJVDI was 0.08%, with 95% limits of agreement ranging from −6.93% to 7.10% (Supplementary Figure S2).

4. Discussion

This study demonstrates that EJVDI could be useful in predicting FR, whereas EJV diameter could assist in assessing volume status in healthy, anesthetized, mechanically ventilated dogs with experimentally induced changes in volume status. Moreover, significant correlations were identified between PPV, SPV, and EJVDI as well as between CVP and EJV diameter. Lastly, reliable intraobserver variability was identified in both the LVOT VTI and the EJVDI.

Fluid therapy is a key element in the management of critically ill patients; however, excessive fluid administration is associated with an increased risk of complications and mortality. Thus, it is essential to assess the FR prior to fluid administration [5, 6]. However, the presence of FR does not automatically indicate the need for fluid administration. Euvolemic or normotensive patients may also show FR as a normal physiological response to volume expansion [31, 32]. However, fluid administration is not indicated in these patients. Therefore, effective fluid management should consider dynamic indices that predict FR and static indices that assess volume status. Recently, POCUS has been used to guide fluid therapy in critical care settings in humans and veterinary medicine, as it can evaluate both dynamic and static indices across various vessels [6, 15].

Changes in intrathoracic pressure during the respiratory cycle are transmitted to the extrathoracic vessels, which serves as a mechanism to predict FR through changes in the diameter of the jugular vein [33, 34]. In the present study, changes in the EJV diameter during the respiratory cycle were found to predict FR, with the EJVDI cutoff value for predicting FR being 22.7% and an AUROC curve of 0.92. This is consistent with findings from previous human studies, where IJVDI was shown to predict FR. In these studies, the cutoff value for predicting FR ranged from 12.99% to 18.92%, with AUROC curves ranging from 0.88 to 0.951 in mechanically ventilated patients [22, 23, 24]. The AUROC curve for predicting FR in this study was comparable to those reported in human studies, but it showed a higher cutoff value. This could be due to the following factors: (1) this study was performed on healthy dogs, whereas human studies involved critically ill patients; (2) low tidal volumes (V_T < 8 mL/kg) were used to prevent lung injury in human studies, whereas the V_T used in this study was 13 mL/kg: this higher V_T may augment the changes in vessel diameter induced by the respiratory cycle [23, 35]; (3) differences in the definitions of FR, as well as variations in the type, volume, and rate of fluid administration during fluid challenges, may affect the results across studies [6]; and (4) body position could affect the diameter of the jugular vein [24]; in humans, assessments were performed in the supine position, whereas in dogs, evaluations were conducted in lateral recumbency, as the supine position is not physiological in veterinary practice. However, due to various conditions that may affect the jugular vein diameter and the interspecies differences between dogs and humans, direct comparison of the findings from studies in both species may not be appropriate. Therefore, further studies under various conditions in dogs are necessary.

PPV and SPV are well‐established dynamic indices for evaluating FR in human medicine [5, 6]. However, the reported cutoff values and AUROC curves for these indices vary across studies in veterinary medicine [36, 37, 38, 39, 40]. In this study, the cutoff values for PPV and SPV were 9% and 6 mmHg, respectively, with AUROC curves of 0.89 and 0.88, respectively. Generally, AUROC curves above 0.8 are considered clinically useful [41], and SPV showed similar predictive accuracy to PPV, which is consistent with a previous study in dogs [36].

Previous studies have shown that static indices are less reliable than dynamic indices for discriminating between fluid responders and nonresponders [5, 6]. The AUROC curves for the IJV diameter and CVP were significantly lower than those for the IJVDI and stroke volume variation in mechanically ventilated patients [22]. However, this study found no statistically significant differences between the AUROC curves of mean CVP, EJV_d‐max, and EJV_d‐min and those of EJVDI, PPV, and SPV, which may be attributed to the consistent control of factors affecting stroke volume, such as anesthetic depth and ventilation settings, as well as the use of an experimentally induced hypovolemia model. During hypovolemia, increases in stroke volume are more dependent on preload than on vasodilatory or cardiogenic shock [42]; as such, static indices may accurately predict the FR. Therefore, further studies are needed to determine whether a statistically significant difference exists between the AUROC curves of EJVDI and EJV diameters in dogs with vasodilatory or cardiogenic shock.

Similar to human studies, this study revealed significant correlations among the EJVDI, PPV, and SPV, where the IJVDI was significantly correlated with stroke volume variation [22]. Moreover, the significant correlation between the EJV diameter and CVP aligns with findings from a human study, which demonstrated that blood donation significantly reduced the IJV diameter [43]. These findings indicate that the EJV diameter could be used to assess volume status. However, further research is required, as the normal EJV diameter may vary among different dog breeds. The abstract of the European Journal of Veterinary Emergency and Critical Care, which presented normal EJV diameter values measured by ultrasound in healthy, spontaneously breathing dogs, identified a significant correlation between EJV diameter and body surface area, with a mean body weight, EJV_d‐max, and EJV_d‐min of 13.1 kg, 4.84 mm, and 4.50 mm, respectively^b. In this study, the mean body weight of the dogs was 10.5 kg, and at TP₁, the EJV_d‐max and EJV_d‐min were 4.51 and 3.79 mm, respectively. Mechanical ventilation and anesthesia may have increased the differences between the EJV_d‐max and EJV_d‐min.

Unlike the standard receiver operating characteristic curve approach, which uses a single cutoff value to classify patients as fluid responders or nonresponders in a binary manner, the gray zone approach offers two cutoff values, facilitating clinical decision‐making in various diagnostic assessments. In the gray zone, FR predictions are less accurate and require the use of additional evaluation methods. However, outside the gray zone, FR and FNR can be distinguished with higher certainty [29]. This study reported the gray zones for EJVDI, PPV, SPV, CVP, and EJV diameter. The number of cases in the gray zone was lowest for EJVDI (21%, 5/24 fluid challenges), followed by SPV (54%, 13/24 fluid challenges) and PPV (75%, 18/24 fluid challenges). These findings indicate that EJVDI may be more useful than SPV or PPV for decision‐making regarding fluid administration.

The intraobserver variability bias for LVOT VTI and EJV diameter was close to zero, with narrow 95% limits of agreement, similar to one study evaluating the changes in CVC diameter and VTI in spontaneously breathing dogs for predicting FR [44]. Given that the interobserver variability was higher than the intraobserver variability in changes in CVC diameter, further research is warranted to investigate the interobserver variability of the EJVDI.

This study had several limitations. First, the minimum number of dogs required to achieve statistical significance was used due to ethical considerations. Second, the study was conducted under controlled conditions, that is, using healthy dogs of a single breed, with standardized anesthesia, mechanical ventilation, and experimentally induced hypovolemia, which limits the extrapolation of the findings to clinical settings. In clinical scenarios, different results may be observed, particularly when respiratory variations in intrathoracic pressure are decreased, which could limit the predictive ability of the EJVDI for FR [23]. Third, since all procedures were conducted in lateral recumbency, the results may differ in other positions. In humans, differences in IJV diameters have been noted between the supine position and 30° head elevation [24]. Fourth, FR was assessed using LVOT VTI rather than CO measurement through thermodilution. Although thermodilution is considered the gold standard for CO measurement, pulmonary artery catheterization is highly invasive and associated with various complications. Furthermore, as the accuracy of echocardiography compared with thermodilution has been demonstrated, the clinical use of thermodilution for measuring CO is decreasing [6]. Finally, the jugular vein is highly compliant, making it susceptible to minor external compression, a well‐known limitation when applying the IJVDI in human studies [22]. Therefore, a large amount of gel and minimal pressure were applied to maintain the transverse image of the jugular vein as intact as possible. In addition, pulsations from the carotid artery may have been transmitted to the jugular vein, potentially affecting the EJV diameter measurement [44].

In conclusion, the present study showed that the EJVDI can be used to predict FR, while the EJV diameter can assist in evaluating volume status in healthy, anesthetized, and mechanically ventilated dogs. POCUS of the EJV may be useful in situations where POCUS of the CVC or placement of an invasive line is limited, thus assisting in patient assessment and fluid therapy. Further studies are required to apply POCUS of the EJV in a clinical setting.

Conflicts of Interest

The authors declare no conflicts of interest.

Ethics Statement

This experimental study was approved by the Ethics Committee for Experimental Animals of Seoul National University (SNU‐230725‐3).

Supporting information

Supporting Information