Septic shock results in mortality rates as high as 40% and is a leading cause of intensive care unit (ICU) admissions.1,2,3,4 Although early and aggressive intravenous fluid administration is recommended by clinical practice guidelines as a cornerstone treatment for septic shock, the optimal volume of intravenous fluid beyond the initial resuscitation and the timing for deresuscitation (i.e., stopping fluids and initiating diuresis) remain uncertain. Given the mounting evidence that suggests iatrogenic volume overload can precipitate organ injury,5,6,7 defining these endpoints are an important research priority to advance the care of patients in septic shock.

While there are multiple potential mechanisms by which excessive fluid resuscitation can cause organ injury, high central venous pressure underpins many of these pathways. High venous pressures precipitate endothelial dysfunction, interstitial edema, and organ congestion.8,9,10 Moreover, high venous pressures impair oxygen delivery to susceptible organs by reducing the perfusion pressure across vascular beds. Although high venous pressure has been associated with acute kidney injury (AKI) in critically ill patients, heterogeneity in patients’ physiology and measurement error precludes us from using central venous pressure to adequately guide fluid administration and guide deresuscitation.10,11

Aided by its portability, safety, and ability to provide real-time anatomic and physiologic data,10,12 point-of-care ultrasound (POCUS) has emerged as a useful tool to help tailor resuscitation strategies for patients with septic shock. Recent studies have shown that POCUS can quantify venous congestion by interrogating the size of the inferior vena cava (IVC) and blood flow patterns of the hepatic vein (HV), portal vein (PV), and intrarenal veins (IRV).10,13,14 Using these findings, one can grade the severity of venous congestion using a validated score called the Venous Excess Ultrasound Score (VEXUS).10,13,14 Higher VEXUS scores are associated with organ dysfunction, particularly AKI, in a broad cohort of critically ill patients,10,14,15,16,17 although its prevalence and significance for patients with septic shock remains unknown. Determining whether venous congestion is associated with clinical outcomes in septic shock could provide clinicians with an important bedside tool to guide intravenous fluid administration.

The primary objective of this study was to evaluate the feasibility of a larger observational study that will determine the association between venous congestion quantified using POCUS and clinical outcomes in patients with septic shock. The secondary objectives were to determine the prevalence of venous congestion in septic shock, describe the natural history of venous congestion on day 1 and day 3 of ICU admission, and explore the association between venous congestion and outcomes in patients with septic shock.

Methods

We conducted a pilot prospective multicentre cohort study of patients with septic shock admitted to the ICU. The Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) statement18 provided guidance on the design and reporting of the study. The Western Health Sciences Research Ethics Board in London, ON, Canada approved the conduct of this study (REB# 120202).

Setting

We recruited patients at the Critical Care Trauma Center (CCTC) at Victoria Hospital and the Medical Surgical Intensive Care Unit (MSICU) at University Hospital in London, ON, Canada. The CCTC is a 32-bed ICU that cares for patients with trauma and general medical and surgical issues. The MSICU is a 30-bed ICU that cares for patients with medical, surgical, neurosurgical, and transplant-related issues.

Patient enrolment

We enrolled adult patients (age ≥ 18 yr) with septic shock within 12 hr of ICU admission, defined as sepsis (according to the Sepsis-3 definition) with hypotension requiring vasoactive medications, as well as any one of the following features of end-organ dysfunction at the time of eligibility assessment: 1) serum lactate greater than or equal to 2.2 mmol·L−1, 2) AKI of at least Acute Kidney Injury Network stage I on ICU admission or decreased urine output (≤ 0.5 mL·kg−1·hr−1 for more than two hours after admission), 3) altered mental status, defined as a Glasgow Coma Scale < 13, or 4) the need for invasive mechanical ventilation.

We excluded patients who had any limitations on their wishes for life support interventions (e.g., do not resuscitate order) at the time of ICU admission, had a previous liver transplant, or with pre-existing end-stage renal disease requiring renal replacement therapy (RRT).

Venous congestion ultrasound

Trained ultrasound operators performed venous congestion ultrasound within 24 hr of ICU admission (day 1) and on ICU admission day 3. The venous congestion ultrasound evaluates four vessels: the IVC, HV, PV, and IRV. Measurements from these four vessels informed the validated VEXUS score.10,13,14 The VEXUS score ranges from 0 to 3, and scores of 2 or 3 represent severe congestion. The Electronic Supplementary Material eAppendix provides an overview of image acquisition and interpretation procedures for the VEXUS score.

Ultrasound operators used a curvilinear probe with an abdominal imaging preset. If the curvilinear probe did not produce adequate images, a phased array probe was used. We obtained images on a Mindray M9 machine (Mindray Bio-Medical Electronics Co., Ltd., Shenzen, China) and archived them in QPath (Telexy Healthcare Inc., Vancouver, BC, Canada), a privacy-compliant cloud-based archiving system. Ultrasound operators used electrocardiogram gating to correlate the Doppler tracings to the cardiac cycle and obtained measurements during end-expiration.

Ultrasound operators included fellows or attending physicians with experience performing Doppler ultrasound as part of their routine POCUS practice. Each operator underwent a two-hour didactic training session, and their first scan was directly supervised by one of the investigators (R. P.). Two experts (R. P. and J. B.) reviewed all POCUS scans independently to ensure that the acquisition was correct. We used the interpretation of the VEXUS operator in our final analysis. We quantified the interrater agreement between the two experts using an unweighted kappa.

Blinding practices

In keeping with previous recommendations for blinding practices during POCUS research,19 ultrasound operators performing venous congestion measurements were not blinded to the patient's clinical information. Nevertheless, the clinical team caring for the patient was blinded to the venous congestion ultrasound results to not influence the natural history of venous congestion, and occurrence of secondary outcomes that may be influenced by clinicians’ management decisions (e.g., initiation of RRT). The outcome assessors were blinded to the VEXUS results.

Data collection

We extracted data from the electronic medical record, paper chart, and the Critical Care Information System (CCIS). Baseline characteristics and patient demographics included age, sex, body mass index, admission multiorgan dysfunction score, pre-existing comorbidities, source of sepsis, creatinine, lactate, white blood cell, hemoglobin, and platelet count at the closest time of ICU admission. We also collected data on the cumulative fluid balance on day 1 and day 3 (at the time the VEXUS scan was performed), highest norepinephrine equivalent requirements on day 1, and the need for invasive or noninvasive mechanical ventilation at the time the VEXUS scan was performed. For cumulative fluid balance on day 1 and day 3, we determined the pre-ICU fluid balance in several different ways depending on the location of the patient before ICU admission. For patients from the emergency department (ED), we included the entire ED fluid balance. For patients from the ward, we looked at 72 hr of pre-ICU cumulative fluid balance. For patients from the operating room, we looked at the intraoperative fluid balance. We did not include fluid balance at outside hospitals for direct transfers.

At both participating sites, critical care practitioners routinely perform focused bedside echocardiography on most patients admitted to the ICU with shock to assess for biventricular function, valvulopathy, and to quantify cardiac output. If a cardiac POCUS or echocardiogram was performed within 12 hr of the VEXUS measurements, we collected data on left ventricular (LV) function, right ventricular (RV) size and function, moderate or severe tricuspid regurgitation, the tricuspid annular plane systolic excursion (TAPSE), RV S’ velocity on tissue Doppler imaging (TV S’) (cm·sec−1), and stroke volume determined through LV outflow tract velocity time integral and cardiac output (L·min−1). We defined LV dysfunction as a visual assessment of ejection fraction of < 55%. Right ventricular dysfunction was determined by qualitative assessment, or a TV S’ < 11 cm·sec−1 or a TAPSE < 17 mm.

Outcomes

The primary outcome was the successful completion of VEXUS scans on day 1 and day 3 for patients still in hospital (i.e., not discharged or deceased). We defined successful completion as adequate image acquisition to interpret the VEXUS score for the patient, as determined by a VEXUS expert (R. P.). Other feasibility outcomes included recruitment rate (number of patients recruited per month) and consent rate (total number of patients consented divided by the number of patients approached for consent). We set feasibility targets of recruitment rate of one patient per site per month, more than 80% of VEXUS scans completed, and less than 10% of patients declined consent. We obtained secondary outcome data from the electronic medical record and the CCIS database. The secondary outcomes included major adverse kidney events at 30 days (MAKE − 30, composite of death, use of RRT, or Creatinine ≥ 200% baseline at 30 days), use of RRT in the first 30 days, duration of vasoactive medications (vasopressors or inotropes) and mechanical ventilation, ICU length of stay, and 30-day mortality.

Sample size and statistical analyses

The sample size was based on achieving adequate precision around estimates for feasibility.20 We would deem a larger observational study feasible if the lower bound of the 95% confidence interval (CI) for completeness of venous congestion scans was 80%. Assuming a preliminary estimate of 90% and a margin of error of 10%, 75 patients would afford us 95% confidence that the venous congestion scan completion rate was at least 80%.20

We used IBM SPSS for Windows version 26 (IBM Corp., Armonk, NY, USA) and R version 4.2.3 (R Foundation for Statistical Computing, Vienna, Austria) software for data analysis. We present results of statistical tests as two-sided and considered a P value < 0.05 statistically significant. For continuous variables, we used mean and standard deviation to summarize data with normal distributions, and median and interquartile range [IQR] for nonnormally distributed data. Counts and percentages represented categorical variables. We classified patients with severe venous congestion (VEXUS score of 2 or 3) on day 1 of ICU admission as congested, while those with VEXUS scores of 0 or 1 were classified as not congested. We compared the baseline characteristics and outcomes of patients who were congested to those who were not congested using the Student’s t test, Mann–Whitney U test (for nonparametric variables), or Chi square test. We excluded missing data from any analyses without imputation. We had a complete data set for feasibility metrics, for which 95% CIs were calculated.

We used the Fine and Gray model for an unadjusted estimate of the hazard ratio (HR) of RRT, and then adjusted for the competing risk of death. We present the cumulative incidence function for the cumulative incidence renal replacement from the day of ICU admission to day 30. For this time-to-event analysis, we excluded four patients who had been started on intermittent RRT prior to admission to the ICU and completion of the VEXUS scan.

Results

From January 2022 to January 2023, we screened 340 patients and enrolled 75 patients in this feasibility study. Figure 1 presents an overview of screening and enrolment procedures. The overall completion rate for VEXUS scans was 95% (95% CI, 90 to 98). Portal veins, HV, and IRV Doppler could not be obtained in three, two, and two patients respectively (in these cases, the VEXUS score was calculated from the remaining Dopplers). The kappa between the two experts for VEXUS interpretation was excellent at 0.95 (95% CI, 0.90 to 1.0). Table 1 provides an overview of the feasibility outcomes.

Fig. 1
figure 1

Study flow diagram

Full size image
Table 1 Feasibility outcomes
Full size table

On day 1 of ICU admission, 19% (14/75) of patients showed venous congestion, while 16% (10/61) of patients showed venous congestion on day 3. Figure 2 provides the natural history of venous congestion. The distributions of VEXUS scores on day 1 and day 3 are provided in Fig. 3. We recruited 6.25 patients per month between the two sites. The consent rate for eligible patients was 100% (95% CI, 94 to 100).

Fig. 2
figure 2

Alluvian graph showing the natural history of venous congestion between day 1 and day 3

Full size image
Fig. 3
figure 3

VEXUS scores on day 1 and day 3. On day 1, 33 patients had a VEXUS score of 0, 28 patients had a VEXUS score of 1, eight patients had a VEXUS score of 2, and six patients had a VEXUS score of 3. On day 3, 30 patients had a VEXUS score of 0, 21 patients had a VEXUS score of 1, four patients had a VEXUS score of 2, six patients had a VEXUS score of 3, and 14 patients did not have a score because no repeat scan was completed (i.e., deceased, unable to complete in timeframe).

Full size image

We found no statistically significant differences in age, sex, basal metabolic index, or multiorgan dysfunction score between patients with and without congestion. Patients with venous congestion were more likely to have a prior history of coronary artery disease (43% vs 7%; P = 0.02) and atrial fibrillation (50% vs 10%; P = 0.002), but we found no differences in the rate of chronic kidney disease, diabetes, chronic obstructive pulmonary disease, congestive heart failure, stroke, or cirrhosis. Similarly, we found no important differences between groups in the source of sepsis, admission serum lactate, hematology serum blood tests on ICU admission day 1, and need for mechanical ventilation. Table 2 provides an overview of the demographics and baseline characteristics for patients with and without venous congestion.

Table 2 Baseline characteristics of patients based on venous congestion status
Full size table

With respect to hemodynamic variables, we found no important differences in the highest dose of norepinephrine equivalents between patients with and without venous congestion. Patients with venous congestion were more likely to receive inotropic agents on day 1 (36% vs 11%; P = 0.03). Patients without congestion had a higher median positive fluid balance on day 1 (2,514 mL vs 1,296 mL; P = 0.09) and day 3 (5,510 mL vs 279 mL; P = 0.09) of ICU admission. Patients who were not congested on day 1 but became congested on day 3 had a higher median positive fluid balance compared with patients who were congested on day 1 but became noncongested on day 3 (9,200 mL vs − 368 mL; P = 0.13). Table 3 provides an overview of hemodynamic and resuscitation data.

Table 3 Clinical variables of patients based on congestion status
Full size table

Patients with venous congestion on day 1 of ICU admission were more likely to exhibit LV systolic dysfunction (43% vs 13%; P = 0.01), RV systolic dysfunction (57% vs 10%; P < 0.001), RV dilation (64% vs 21%; P = 0.006), and lower TV S’ (8.6 vs 14.0; P = 0.009). The rates of moderate or severe tricuspid regurgitation, stroke volume, cardiac output, or TAPSE were similar between groups. Table 4 provides a comparison of the echocardiographic data based on venous congestion.

Table 4 Critical care echocardiography data
Full size table

We did not find any important differences in MAKE-30, use of RRT at 30 days, vasopressor days, duration of mechanical ventilation, ICU length of stay, or 30-day mortality between patients with and without congestion. Patients with venous congestion may show an increased risk for requiring RRT (unadjusted HR, 3.35; 95% CI, 0.94 to 11.88; P = 0.06), but this finding was limited by CIs that crossed no effect (Fig. 4). Table 5 presents the secondary outcomes of patients with and without venous congestion.

Fig. 4
figure 4

Cumulative incidence function for renal replacement therapy at 30 days based on venous congestion

Full size image
Table 5 Patient outcomes based on venous congestion status
Full size table

Discussion

This pilot multicentre prospective cohort study takes the first step to understanding the prevalence, natural history, and associated outcomes in patients with septic shock and venous congestion. Based on the VEXUS scan completion rate, recruitment rate, and consent rate, it is feasible to undertake a larger observational study to address important knowledge gaps to determine whether venous congestion is an important physiologic marker to guide intravenous fluid administration. We also hypothesize that venous congestion may be associated with a higher risk of requiring RRT, but we require larger prospective studies to fully understand this relationship.

The derivation and validation of VEXUS occurred in a cardiac surgical population, where venous congestion was strongly associated with the development of AKI on day 7.10,15 Furthermore, markers of venous congestion are associated with longer duration of life support therapies and complications in cardiac surgical populations.21 Despite these findings, the biological thresholds for harm have not yet been established for venous congestion in patients with septic shock. It is conceivable that the threshold for venous congestion-induced organ injury may be lower in septic shock patients with endothelial dysfunction and capillary leak syndromes; however, this has yet to be determined. Future research should focus on determining the thresholds at which venous congestion is harmful for patients with septic shock.

In this study, nearly 20% of patients showed venous congestion on day 1 of ICU admission, which is slightly lower than the 40% seen in cardiac surgical populations.10,15,16 We also observed that venous congestion is a dynamic process; a number patients who were congested on day 1 were no longer congested on day 3 and vice versa. Interestingly, patients with congestion on day 1 had a lower fluid balance on day 1 and day 3. One explanation may be that certain patients show a lower threshold to develop venous congestion because of cardiac dysfunction. In our cohort, patients with venous congestion were also more likely to have biventricular failure. Another explanation is that patients with congestion had other markers of fluid overload that led to clinicians giving less fluid. Nevertheless, patients who transitioned from being congested on day 1 to not congested on day 3 had a more negative fluid balance on day 3, while patients who were not congested on day 1 but became congested on day 3 had a more positive fluid balance on day 3. Although this study was not powered to statistically ascertain these differences, the results raise the possibility that venous congestion may be dynamic and modifiable through control of intravascular volume. Previous findings show a large absolute difference, but statistical significance was not reached probably because the number of patients was small. These findings may also challenge the common practice of initiating or titrating diuretics based on 24-hr fluid balance alone.

While some clinicians may view venous congestion as an independent physiologic entity, some of our findings challenge this notion. Patients with venous congestion were more likely to exhibit biventricular dysfunction. Is venous congestion independently associated with worsening organ dysfunction or is it a downstream consequence of RV pathology? We hypothesize that multiple factors contribute to venous congestion. Patient-related factors, such as pre-existing comorbidities (e.g., diastolic dysfunction, pulmonary hypertension), disease-related factors (e.g., acute biventricular dysfunction, acidosis, hypoxemia), and treatment-related factors (e.g., intravenous fluid administration, mechanical ventilation, positive end expiratory pressure) may all converge to precipitate venous congestion in susceptible patients. Future studies should explore the interplay between cardiovascular dysfunction and venous congestion to determine which physiologic entities are the best candidates for targeted resuscitation strategies. We hypothesize that the management of venous congestion requires more than just the administration of diuretics and the withholding of intravenous fluids. Rather, clinicians need to employ a multimodal approach to treat RV and/or LV dysfunction, pulmonary hypertension, and intrabdominal hypertension that could be resulting in organ-level congestion.

This study has several important limitations. First, given the feasibility objectives, it was not powered to explore clinical outcomes. Therefore, all outcome data should be considered exploratory. Furthermore, the study was not powered to perform multivariable regression to determine whether venous congestion was independently associated with clinical outcomes. Second, missing venous congestion measurements on day 3 limit conclusions evaluating the evolution of venous congestion during the resuscitation phase. Some of these data were missing because of logistics, as day 3 fell on a Saturday or Sunday so providers were not available to perform the repeat VEXUS scan. With increasing depth of VEXUS experience and dedicated knowledge translation efforts, we hope to improve this to ensure feasibility of a definitive trial. Third, VEXUS was not initially derived nor validated in the septic shock population. Although VEXUS has construct validity, the thresholds for severity of venous congestion may not optimally discriminate which patients will experience harm.

Conclusion

The results of this study indicate that a larger prospective multicentre cohort study to determine whether venous congestion is independently associated with clinical outcomes in septic shock is feasible. Furthermore, this study raises the notion that venous congestion may be a modifiable risk factor as it appears that differences in fluid balances between day 1 and day 3 are associated with changes in congestion status for patients. Finally, venous congestion was associated with biventricular dysfunction on echocardiography and pre-existing cardiovascular disease. These findings suggest that venous congestion results from an interplay between cardiovascular status, respiratory status, and fluid balance, and is likely not an independent physiologic entity. Given the increasing organic adoption of VEXUS in clinical practice, there is an increasing sense of urgency to understand this complex physiologic process to optimize its safe and effective implementation into practice.