Hemodynamic profiling by critical care echocardiography could be more accurate than invasive techniques and help identify targets for treatment

In this prospective observational study, non-invasive critical care echocardiography (CCE) was used to obtain quantitative hemodynamic parameters in 107 intensive care unit (ICU) patients; the parameters were then visualized in a novel web graph approach to increase the understanding and impact of CCE abnormalities, as an alternative to thermodilution techniques. Visualizing the CCE hemodynamic data in six-dimensional web graph plots was feasible in almost all ICU patients. In 23.1% of patients, significant tricuspid regurgitation prevented correlation between thermodilution techniques and echocardiographic hemodynamics. Two parameters of longitudinal right ventricular function (TAPSE and S’) did not correlate in ICU patients. Clinical surrogate parameters of hemodynamic compromise did not correlate with measured hemodynamics. 26.2% of the patients with mean arterial pressures above 60 mmHg had cardiac indices (CI) below 2.5 L min−1·m−2. A CI below 2.2 L·min−1·m−2 was associated with a significant ICU survival disadvantage. CCE was feasible in addition or as an alternative to thermodilution techniques for the hemodynamic evaluation of ICU patients. Six-dimensional web graph plots visualized the hemodynamic states and were especially useful in conditions in which thermodilution methods were not reliable. Hemodynamic CCE identified patients with previously unknown low CI, which correlated with a higher ICU mortality.

The invention of the pulmonary artery catheter (PAC) by Swan and Ganz in 1970 made it possible to measure extended individual hemodynamics in the intensive care unit (ICU) 1 . Following its introduction, the PAC was increasingly employed in the hemodynamic assessment of critically ill patients until studies revealed that patients in whom a PAC was used were at risk of serious adverse events and even adverse outcomes, such as an increased 30-day mortality rate 2 . There is a multitude of reasons for these adverse effects, which include interpretation problems 3 , data acquisition problems 4 , and the difficulty in translating the data into treatment decisions 5 . Despite a number of disadvantages, the PAC is still in use today 6 . However, in many European countries, transpulmonary thermodilution techniques and pulse contour analysis, such as that used by the PiCCO system, have replaced the PAC for a variety of purposes 7 . All currently available techniques and systems for assessing hemodynamics have a number of relevant limitations and contraindications. PAC measurements are, for example, inaccurate in patients with significant tricuspid or pulmonary valve regurgitation 8,9 . Pulse contour analysis is inaccurate in patients with significant tricuspid valve regurgitation, arrhythmias or pulmonary edema 10 . To overcome these disadvantages, especially in critically ill patients, there is need for alternative methods that would ideally be non-invasive, measure continuously and be feasible in situations where current thermodilution techniques have limitations and contraindications.
Critical care echocardiography (CCE) meets two of these requirements. However, this technique has previously been used only descriptively to identify cardiac abnormalities such as the presence or degree of aortic stenosis. Its ability and potential for obtaining hemodynamic measurements in the critical care environment has neither been routinely assessed nor quantified precisely. www.nature.com/scientificreports/ As early as 1983, Huntsman et al. had demonstrated a close correlation between cardiac output (CO) determined by echocardiography and PAC measurements 11 . Since then, many more echocardiographic techniques for assessing hemodynamic values have been evaluated in non-critical care settings. It has been shown that in cardiac intensive care patients, non-invasive two-dimensional and Doppler echocardiographic parameters improve risk stratification for hospital mortality 12 . In detail, Jentzer and colleagues presented work suggesting that echocardiographic hemodynamic assessment in cardiac ICU patients allows not only a more accurate categorization of the clinical shock phenotype but also potentially allows a more individualized therapeutic approach 13 . In another recent study, in the majority of non-cardiologic surgical ICU patients, significant or critical cardiac abnormalities were detected that are associated with a higher ICU mortality 14 .
In this study, our first aim was to combine validated techniques of non-invasive echocardiographic measurement of hemodynamic values and evaluate the feasibility of a hemodynamic CCE protocol without the need for invasive hemodynamic techniques.
The second aim of our study was to develop and evaluate hemodynamic profiles visualized as six-dimensional web graph plots that easily allow identification of individual therapeutic targets in order to specifically improve deteriorating hemodynamic states in a particular patient.
Furthermore, the third aim of our study was to assess the impact of descriptive structural echocardiographic parameters on quantitative hemodynamic states.

Methods
Institutional approval, patient population and study setting. This prospective observational study was conducted in 107 consecutive patients admitted to two anesthesiology-supervised ICUs in a tertiary care university hospital. Written informed consent was obtained from all patients. Only patients with a non-cardiological, non-cardiothoracic admission diagnosis, who had not undergone recent cardiac surgery, were eligible for the study. The patient population mainly consisted of postoperative patients of any discipline, trauma patients, and patients with sepsis or acute respiratory distress syndrome. The echocardiographic assessment was conducted on the third day of the ICU stay in order to exclude postoperative patients who were admitted simply for observational reasons. The study protocol was reviewed and approved by the local ethics committee (ethics proposal Universitaetsmedizin Goettingen [UMG 11/12/13], 18/02/2014). All methods were performed in accordance with the relevant guidelines and regulations and in accordance with the amended Declaration of Helsinki; the protocol and the primary and secondary aims of the study were registered in the German Clinical Trials Register (DRKS00010208, 25/04/2016).
Echocardiographic technique, calculation of hemodynamic parameters and hemodynamic profiling. Transthoracic echocardiography was conducted according to the recommendations and guidelines of the American Society of Echocardiography [15][16][17][18][19] . If these were not applicable, guidelines of the European Association of Cardiovascular Imaging were used 20,21 . In the event that neither guideline was applicable, or if threshold values were not defined, the echocardiographic techniques or values are explicitly described in the text below.
Hemodynamic evaluation was performed according to established 19 and separately evaluated techniques as follows: Mean arterial pressure (MAP) was measured invasively via an indwelling arterial catheter. If none was in place, blood pressure was measured with an oscillometric device. Central venous pressure (CVP) was measured with a central venous catheter, or if not available, approximated from the respiratory-induced changes in inferior vena cava diameter 16 . Stroke volume (SV) and CO were determined as described by Huntsman et al. 11 (SV = π·(LVOT / 2) 2 ·VTI LVOT ; CO = (SV·HR) / 1000; LVOT: left ventricular outflow tract; VTI: velocity-time integral; HR: heart rate). Cardiac index (CI) was calculated by the CO divided by the body surface area (BSA); BSA was determined by the Du Bois formula (BSA = 0.007184·W 0.425 ·Ht 0.725 ; W: weight; Ht: height). Systemic vascular resistance (SVR) was calculated using the formula SVR = (80·[MAP -CVP]) / CO. In the calculation of the systemic vascular resistance index (SVRI), CI replaced CO in the formula. Left atrial pressure (LAP) was estimated by evaluating diastolic function parameters according to previously published guidelines 18 . Mean pulmonary capillary wedge pressure (PCWP) was estimated using the Nagueh formula 22 (PCWP = 1.24·[E / average e´] + 1.9). Pulmonary vascular resistance (PVR) was calculated according to the Abbas equation (PVR = TRV / VTI RVOT ·10 + 0.16; TRV: tricuspid regurgitation velocity; VTI RVOT : velocity-time integral of the right ventricular outflow tract) 23 . Wood Units (WU) were converted to dyn·s·cm -5 by multiplying by 80. Systolic pulmonary artery pressure (sPAP) was approximated according to the formula sPAP = (4·TRV 2 ) + RAP (RAP: right atrial pressure) 24 . In the majority of cases, RAP was measured through an indwelling central venous catheter or estimated based on the size of the inferior vena cava and its collapsibility 16,25 .
Stroke volume, CI, PVR, sPAP, SVRI and PCWP were plotted in a six-dimensional web chart to enable a rapid understanding of the hemodynamic situation. Various hemodynamic states can be easily recognized by using this visualization technique. We refer to this combined visualization and recognition technique as hemodynamic profiling.
The same expert CCE examiner performed all examinations. This examiner is trained in cardiology, anesthesiology and intensive care medicine, and was therefore particularly qualified to directly integrate the findings into changes in therapeutic management.
A General Electric (GE) Healthcare Vivid S5 machine equipped with a phased array adult 1.5-3.6 MHz sector scanner was used for the study. All necessary Doppler features (CD, PW, CW and TDI), imaging modalities (2D and M-mode) and software features (proximal isovelocity surface area (PISA) etc.) were available. Images were stored digitally and analyzed immediately after each examination. Kaplan-Meier curves were used for survival analysis, and groups were compared by the log-rank test. Gaussian distribution was assessed by the Shapiro-Wilk test; associations and correlations were evaluated by linear regression analysis and reported with the Pearson correlation coefficient (Pearson's r), the coefficient of determination (r 2 ) and adjusted r-squared values (adj. r 2 ). All regression equations are reported within the relevant figures. Spearman's correlations and 95% confidence intervals were performed between all indices. Two-tailed p-values are reported; a p-value below 0.05 was considered statistically significant.

Results
Patient characteristics. Patient characteristics and details on catecholamine therapy are given in Table 1.
Non-invasive hemodynamics. Determination of left ventricular ejection fraction (LVEF) was successful in 70.1% of patients using the Simpson method and in 97.2% using visual assessment techniques. Determination of right ventricular function was successful in 88.8% using tricuspid annular plane systolic excursion (TAPSE) and in 79.4% using tricuspid annular peak systolic velocity (S'). We were able to obtain valid CO and CI data in 96.3%, PCWP in 93.5%, PVR in 88.4%, sPAP in 89.7%, SVR and SVRI in 96.3%, LAP in 93.5%, and ratio of mitral inflow and the average early diastolic velocity of the mitral annulus obtained by tissue Doppler (E/e') in 93.5% of patients. Quantitative hemodynamic data are plotted as web graphs (Fig. 1). The superimposed plots of all examined patients are shown in Fig. 2.

Correlation between clinical parameters and echocardiographic non-invasive hemodynamics.
In order to identify a surrogate parameter for hemodynamic compromise, the parameters determined by CCE and non-invasive hemodynamics (TAPSE, CO and CI), MAP, systolic blood pressure (sBP) and LVEF were analyzed. No clinically useful correlations were found (  Table S1). Value of routine hemodynamic CCE analysis. Our study was not designed with survival as a primary endpoint. However, our study patients with a CI greater than 2.2 L·min −1 ·m −2 had a predicted mortality of 14.6% (individual mortality prediction according to individual SAPS II scores) and an actual ICU mortality of 18.7%.   (Fig. 4). Kaplan-Meier curves of patients with a cardiac index (CI) above and below 2.2 L·min −1 ·m −2 . A statistically significant difference in ICU survival times between both groups was determined by log-rank analysis (p = 0.030). Dots show right-censoring.

Discussion
Although echocardiography-based hemodynamic assessments are mentioned in guidelines and recommendations [26][27][28][29][30] , CCE has not yet found its way into routine clinical practice 7 , and hemodynamic parameters are mostly determined by qualitative visual evaluation rather than by objective, quantitative methods.
Despite the known limitations of CCE image quality in ICU patients, it can be employed in the vast majority of ICU patients when performed by an expert examiner 14 . Limitations apply to patients with absent or very restricted transthoracic imaging windows.
The descriptive and comparative results of our prospective observational study have closed the gap between the classical descriptive echocardiographic and the hemodynamic assessment of the individual patient by extending CCE to include quantitative hemodynamic parameters that were previously the domain of invasive hemodynamic monitoring techniques.  www.nature.com/scientificreports/ This offers the opportunity to directly evaluate the cause of cardiovascular deterioration, for example a drop in sBP or MAP. Therefore, if a decrease in CI were detected by non-invasive hemodynamic CCE, it could now be further attributed for example to high PVR and/or diastolic dysfunction without the need for invasive hemodynamic measurements. The detected underlying pathology could thus then be subjected to specific treatment instead of simply reporting deteriorating hemodynamic values. Furthermore, quantitative hemodynamic parameters such as PCWP, CI and SVR can be used in real time and individually to evaluate the true hemodynamic impact of standard descriptive echocardiographic parameters such as a decreased LVEF or valvular disease.
Visualizing the hemodynamic results in six-dimensional web graph plots (hemodynamic profiling) might help intensivists to identify targets for hemodynamic improvement and specific treatment (Fig. 1). Considering all filled areas below the middle line of the plot as hemodynamically destabilizing and potentially harmful, targets for hemodynamic improvement can now be identified even by less experienced intensivists. Visualizing data for better interpretation results is an accepted and self-evident principle [31][32][33] . The effects of altering treatment can likewise easily be extracted from the area above the middle line of the plot, the length of whose base should not exceed that of the upper diagonal (Fig. 1c,d).
Direct hemodynamic assessments are essential in critically ill patients, because MAP, sBP and other clinical parameters used as surrogate markers cannot be used to determine CI in most critical care scenarios 34 ( Table 2). Although MAP was above the threshold value of 60 mmHg in 96.1% of all patients, CI was less than 2.5 L·min −1 ·m −2 in 26.2% and less than 2.2 L·min −1 ·m −2 in 10.7%. These findings highlight the gap between clinically perceived and actual individual hemodynamics in ICU patients.
Predicted (based on individual SAPS II score mortality estimation) and actual ICU mortality rates in patients with a CI above 2.2 L·min −1 ·m −2 did not significantly differ. In our institution, even outside the current study cohort, SAPS II predicted mortality rates and actual ICU mortality rates are very close 35 . However, in our study patients with a low CI, actual ICU mortality rates were more than two-fold higher than the mortality rate predicted by the SAPS II scores. This identifies CI as a potential risk factor for mortality in the ICU. Kaplan-Meier and log-rank analysis likewise revealed statistically significant lower survival rates in patients with a low CI while in the ICU (Fig. 4). We interpret the lower probability of survival for patients with a lower CI as a further indication for the need for routine hemodynamic assessments with implementation of appropriate therapy adjustments, although hard evidence that this improves clinical outcome is still lacking 14,36 .
The current study was not designed as a comparison study between invasive and non-invasive hemodynamic evaluation methods. However, the need for routine non-invasive hemodynamic CCE or at least CCE alone was even more evident in our patient cohort, because 23.1% of PAC or PiCCO measurements were obtained in patients with significant TR, potentially resulting in invalid data 8,9 . Without CCE, these data would have There are also other conditions that can negatively affect the accuracy of hemodynamic measurements with PACs and thermodilution techniques (e.g., arrhythmias, pulmonary edema) [8][9][10] , and these might further increase the risk of invalid measurements to greater than the 23.1% estimated above. Systematic echocardiographic examinations and non-invasive hemodynamic CCE assessments additionally revealed that the confirmed high correlation TAPSE and S' in non-ICU patients as parameters of longitudinal right ventricular function could not be reproduced in ICU patients (Fig. 3, Supplementary Table S2). There is already some evidence that the correlations of parameters found in diastolic dysfunction in non-ICU patients are not valid in ICU patients 18 . The low correlation between the longitudinal right ventricular function parameters TAPSE and S' in this study highlights the fact that many echocardiographic parameters and classifications have not been evaluated in ICU patients; correlating the echocardiographic findings with hemodynamic parameters in critically ill patients would help to assess their true impact on cardiac function in this subgroup.
Limitations of the study and caveats. Several limitations and caveats apply to our prospective observational study, which yielded a moderate amount of descriptive data. Compared to invasive hemodynamic monitoring, assessment using hemodynamic CCE requires extensive knowledge and training in echocardiography, which is currently difficult to provide to a sufficient number of non-cardiologist intensivists, and having skilled practitioners available on a 24/7 basis is even more difficult. Non-invasive hemodynamic CCE is non-continuous compared to a number of invasive methods that provide continuous data. In previous studies, non-invasive hemodynamic parameters correlated sufficiently well with invasive measurements to allow the former to be used clinically, and for absolute values to be reported. For example, the correlation coefficient between invasive and non-invasive measurements of CO is 0.94 with a standard deviation of 8.6% 14 , and for PCWP the correlation coefficient is 0.87 20 with a difference between Doppler and catheter measurements of 0.1 ± 3.8 mmHg 20 . However, SV estimation with echocardiography can be affected and limited by several geometric assumptions regarding the LVOT assessment. There is conflicting evidence that some hemodynamic parameters are affected by underlying cardiac abnormalities and under certain conditions, measurements may be of limited value. For example, the PCWP estimated by the Nagueh formula 22 is probably not well correlated with the invasively meas- www.nature.com/scientificreports/ ured PCWP in patients with normal LVEFs, left bundle branch block or mitral regurgitation 37 . Therefore, single measurements should always be interpreted with care, and multiple readings over time are preferred, especially in ICU patients, until further evidence becomes available. Survival was not a primary endpoint parameter in this study. Therefore, the observed survival benefit must be interpreted with care. Without a control group, the study results are only descriptive. In addition, specific limitations for the use of echocardiographic assessments apply to critically ill patients and those with multivalvular disease. For example, the presence of mitral regurgitation influences tricuspid regurgitation with a likely increase in PAP and a consequent impact on the regurgitant volume. Severe regurgitant lesions, such as severe TR, may also impact on sPAP approximations or functional measurements such as TAPSE and S' 38 . However, the impact of these inaccuracies is not always quantifiable. Therefore, multiple evaluations with different approaches should be performed and single readings interpreted with care.
Hemodynamic measurements are not solely the domain of thermodilution techniques and echocardiography, and can also be acquired by other techniques (e.g., bioimpedance, waveform analysis). These techniques are not discussed here but can be valuable additions or alternatives in specific situations 39 .

Conclusions
The mainly descriptive and partly comparative results of the current study indicate that hemodynamic CCE could be used in the vast majority of ICU patients to determine the hemodynamic impact of the structural changes detected by standard echocardiography and to identify dedicated targets for improving deteriorating hemodynamic states. Hemodynamic profiling could aid this process by visualizing the current hemodynamic state and simplifying target identification for specific target-directed treatments. Hemodynamic CCE could be especially advantageous when conditions are present in which thermodilution methods might produce invalid results. By the use of CCE, a subset of patients with formerly unknown low cardiac indices that have a higher mortality while in the ICU could be identified.

Data availability
The datasets generated during and/or analyzed during the current study are available from the corresponding author on reasonable request. www.nature.com/scientificreports/ Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http:// creat iveco mmons. org/ licen ses/ by/4. 0/.