The main results of our study can be summarized as follows. A change of 20% of S-wave peak velocity or 25% of S-wave VTI may identify preload responsiveness during a PLR maneuver in sedated and mechanically ventilated critically ill patients. Analysis of D-wave velocities had worse predictive capacity.

The hemodynamic effects of a PLR maneuver have been estimated to mimic a fluid challenge of approximately 300 ml [21]. Current expert recommendations suggest performing the test from a semi-recumbent rather than from supine position, to induce shifts from two vascular beds (the legs and the splanchnic region) and enhance diagnostic accuracy [21]. This was hypothesized in the study by Jabot et al., in which sequential postural changes from semirecumbent to supine and from supine to leg elevation had similar net effect to that of the entire maneuver performed at once [22]. The authors inferred that the scaffolded approach isolated the relative contributions of both vascular territories (splanchnic and legs), but unfortunately, it wasn’t measured [22]. The current study provides direct assessment on the impact of PLR on splanchnic flow through Doppler assessment, potentially confirming this hypothesis. In this sense, the fact that significant changes in flow-related measurements of the splanchnic region accurately predicted changes in CO, support that the contribution of this venous reservoir is relevant, at least in this clinical context. Future research endeavors should further investigate the relative contribution of each territory in this and other clinical contexts.

Contemporary research has assessed the accuracy of splanchnic Doppler signals to predict preload responsiveness. Cheong et al. analyzed the utility of portal vein pulsatility index (PVPi), finding that a value of PVPi > 32% predicted fluid unresponsiveness with a sensitivity 30.8% (17–47.6%) and specificity of 100% (85.8–100) [29]. On the other hand, Wu et al. analyzed absolute values of S and D waves of HVD, finding higher delta S-wave velocity on patients where cardiac output increased after volume expansion as compared to those where it did not (30.1 cm/s ± 10.2 to 37.1 cm/s ± 12.5; p < 0.003) [30], moreover, no diagnostic accuracy was analyzed. Our study further expands the results obtained by these previous researchers, since they aim specifically at interrogating Doppler waves that represent forward flow to the atrium and are near inferior vena cava drainage, as compared to portal vein. Thus, we believe it has a stronger physiological background and could interrogate more directly changes in VR, explaining their increased accuracy.

Multiple tests to predict preload responsiveness have focused on measuring arterial flow-related variables to identify markers of CO increase [31], such as pulse pressure or stroke volume. Only two proposed preload responsiveness tests have focused on interrogating the venous side of circulation, namely SVC and IVC diameter variability [32, 33]. However, both present technical and conceptual challenges. The former requires transesophageal ultrasound (US) measurements [34], hindering its’ widespread applicability, while the latter has multiple technical drawbacks related to insonation angle, IVC sphericity and venous compliance, limiting its’ diagnostic accuracy [35]. The novel approach proposed on this study, which focuses on blood flow variations measured through pulsed Doppler rather than anatomical variations, such as diameter, could help overcome these issues, and approach indirectly to VR changes. Due to the angulation required, obtaining flow patterns from the IVC is not feasible, but the HV provides an adequate US assessment site that drains directly into the IVC without major valves or obstructions, becoming a valuable and emerging window for hemodynamic assessment.

Recently, the visual inspection of HVD has also been studied as a marker for venous congestion or fluid intolerance, being a pivotal component of the VExUS score [36]. S-wave reversal, which represents retrograde flow, has been proposed as a marker of hepatosplanchnic congestion. Altered VexUS has correlated to key hemodynamic parameters such as high CVP values [37], and could predict the risk of acute kidney injury in different contexts. HVD has also been used as a marker of venous congestion in heart failure context with prognostic implications as well [38]. As proposed by Kenny et al., the dynamic assessment of waveform patterns (i.e., increase in D wave or reversion of S wave) through a PLR maneuver or fluid challenge could aid in identifying potential congestion [39, 40]. Thus, HVD could be a fairly unique tool able to assess not only fluid responsiveness as seen this manuscript, but also assess fluid tolerance, which is of growing clinical importance [41, 42]. Moreover, future research efforts should answer these areas of uncertainty.

Of note, in this cohort, preload responsive patients had significantly higher baseline TAPSE readings as compared to non-responders (although median values were within normal range). Thus, it could be argued that the difference in preload responsiveness status was determined because patients had better right ventricular function, determining a steeper cardiac function curve. Recent reports such as Zhang et al. have shown that TAPSE correlates with HVD-derived indexes such as the systolic filling fraction [S-wave velocity/(S-wave + D-wave velocities)] [20]. Our results show that delta S-wave velocity correlates with TAPSE (Additional File 3). Both phenomena could be explained by the fact that during ventricular systole the tricuspid annulus moves inward to the cardiac apex, which causes a further anterograde flow during systole (thus reflected in the S wave of HVD) [14]. Moreover, this study was not designed to address this question, and clinical interpretation should be made with caution considering other key echocardiographic results obtained, but certainly it deserves further exploration.

This study has several strengths. First, it’s bicentric nature. Second, delta S-wave velocity presented an adequate AUC–ROC (> 0.8), with less than 40% of measurements in the gray zone of diagnostic tolerance. This becomes particularly relevant when considering the easiness of image acquisition, as compared to LVOT–VTI. In fact, Spiegel et al. only reported 7.9% of inadequate HVD window in a cohort of patients admitted to ICU [43], similar rate to that shown by Prager et al. in a cohort of septic shock patients [15]. Even though our study was not designed to assess image acquisition feasibility of HVD, we excluded only 4% of patients for this reason out of the eligible patients screened (Additional File 1).This contrasts significantly to the prevalence of inadequate windows to obtain LVOT–VTI, which can amount up to 22% [44, 45]. Added to the non-invasive nature of the technique, no requirement of CO monitor and its’ easier learning curve, HVD could emerge as a valuable tool at the bedside assessment of critically ill patients.

This study has several limitations. First, as all ultrasound-based assessments, the technique is operator dependent and intra or inter-observer variability could occur. Second, we decided to assess changes in CO through a PLR rather than the more classic design of administrating a fluid challenge [25] to assess fluid responsiveness. Even though the fluid challenge has been praised as a gold-standard technique, both the physiological determinants of PLR (± 300 ml of autotransfusion) and the diagnostic accuracy of its’ derived measurements (sensitivities and specificities > 85%) [24], provide a comparable alternative. This approach has already been used in the contemporary literature of diagnostic accuracy of preload responsiveness tests [46, 47], and has the added benefit of avoiding potential deleterious fluids [48, 49] to patients which are preload unresponsive. Another potential criticism could be the use of non-calibrated CO monitors such as those used in this study. Moreover, the technical determinants (i.e., pulse contour analysis) for tracking relative changes of CO are the same as those used in transpulmonary thermodilution (TPTD), and the least significant changes of CO variability has been estimated around 1%, significantly under our detection threshold of PR of 10% [50]. The multibeat pulse contour monitoring used has a short refresh rate (20 s), and presents an adequate ability to capture quick and relative changes such as those happening during a PLR, as shown in recent studies [51, 52]. There was a predominance of septic patients, which could present with higher rates of preload responsiveness [53]. Thus, this could preclude extrapolation to other diagnostic groups of circulatory failure. Since our study was the first assessment of the diagnostic accuracy of this novel technique, we decided to avoid potential physiological confounders, such as right sided valvular diseases, light sedation or patients with respiratory efforts, which could introduce bias or error and preclude a correct interpretation of results.

Future studies should confirm these results and test the diagnostic capacity of HVD in broader contexts of critical illness, such as patients with spontaneous breathing, light sedation, arrythmias or valvular diseases. Aswell, the usefulness of the dynamic assessment of fluid intolerance through identification of flow patterns of S and D waves in concomitance with preload responsiveness could aid clinicians on identifying the best balance between risk/benefit ratio on fluid administration. Finally, future studies could compare, through hand-tracking devices, the difference on number of movements and distance travelled between HVD and LVOT–VTI during a PLR maneuver [54].

In conclusion, in sedated and mechanically ventilated critically ill patients, dynamic changes of S-wave on HVD after a PLR maneuver had a suitable predictable capacity to identify preload responsiveness. This technique could become valuable in scenarios of basic hemodynamic monitoring and when cardiac US is not feasible. Future studies should confirm these results and their relationship with dynamic venous congestion assessment.