Rationale Injury of the right ventricle (RV) is common in patients admitted to intensive care and impacts management and prognosis. Echocardiography has become the cornerstone of diagnosis and management of such an impairment. However, the many studies using echocardiography already published use different approaches to characterize and define "RV injury" making it difficult to accurately assess its incidence and prognostic impact. Briefly, 3 approaches are proposed. The first is based on a more cardiological approach that consists of using parameters of the RV systolic function, and defines the injury as these parameters estimates are below certain threshold. The most commonly used parameters are TAPSE, fractional area change, and Doppler velocity of the systolic (S') wave at the tricuspid annulus. The second approach is based on the pattern of cor pulmonale, defining RV injury as the presence of RV dilatation and paradoxical septal motion. The third approach is to define RV injury according to its consequences on upstream congestion (kidney, liver for example). This is probably the most physiological approach. It is thus defined as the association of an RV dilation with an abnormally high central venous pressure (CVP).
The hypothesis is that the best definition of RV injury is the one that will be best associated with hemodynamic status (e.g. the lack of response to volume expansion or its equivalent such as passive leg raising, PLR), ventilation and oxygenation parameters and with clinical outcomes, including mortality (ICU or in-hospital), duration of invasive ventilation (in ventilated patients), ICU length of stay and duration of catecholamine infusion.
All patients will receive a transthoracic (TTE) or a transesophageal (TEE) echocardiography according to the usual practice of the center and the patient's situation. A standard TTE or TEE echocardiography will be performed within 48 hours of inclusion (Echo-1), and another TTE or TEE (Echo-2) for fluid responsiveness study immediately after Echo-1.
The following information will be recorded
1. Patients characteristics (at admission and/or at inclusion) Date of ICU admission, date of inclusion, age, gender, past history of hypertension, diabetes mellitus, ischemic heart disease, cardiomyopathy of any cause, chronic kidney disease, COPD (chronic obstructive pulmonar disease), chronic PH (pulmonary hypertension), height, size (BMI), SAPS II, "primary" inclusion (within 2 days following admission) or "secondary" inclusion (during ICU stay), admission diagnosis: sepsis, septic shock, acute exacerbation of COPD, pulmonary embolism, acute asthma, drug poisoning, brain injury, myocardial infarction, pneumonia, COVID-19, miscellaneous.
2. Hemodynamic status (at the time of echo) Heart rate, sinus rhythm (Y/N), blood pressure (systolic, diastolic, mean), pulse pressure variations, PPV (if available), CVP (or RAP if PAC), serum lactate, base deficit, type of catecholamine and dose (in µg/kg/min), VA ECMO (yes/no).
3. Renal, liver function and biomarkers (at the time of echo or the closest to the study) Creatinine, ASAT, ALAT, bilirubin, platelets, troponin, BNP (if available)
4. Blood gas analysis (at the time of echo or the closest to the study) PaO2, PaO2/FiO2, PaCO2, pH, SaO2.
5. Ventilatory settings and oxygen delivery (at the time of echo)
* Nasal cannula (O2 L/min), High flow nasal cannula (FiO2), NIV, invasive ventilation
* For NIV (non invasive ventilation): FiO2, PEEP, pressure support
* For invasive ventilation (volume or pressure controlled mode): FiO2, tidal volume, respiratory rate, total PEEP, external PEEP, plateau pressure, peak pressure.
* VV ECMO (yes/no).
* Prone position yes/no
* Berlin's definition of ARDS yes/no
6. Status
* Discharged form ICU: alive/dead and date.
* Discharged from hospital: alive/dead and date.
* Length of catecholamine infusion (days).
* Length of invasive mechanical ventilation (days).
* Renal replacement therapy Y/N
7. ECHO 1 (route and date): Has to be done in a patient in semirecumbent position in case a PLR is done. All measurements will be done at end-expiration, except for respiratory variations. Some parameters will be mandatory and others optional.
LV systolic function: LVEDV, LVESV, LVEF (4-chambers); LVFAC (short axis); LVOT VTI (5-chambers) or RVOT VTI in case of poor acoustic window, LVOT diameter, MAPSE, S' mitral annulus (TDI).
LV diastolic function: E, A, E' (lateral), A' (lateral), LA volume (4 chambers), Vmax tricuspid regurgitation.
RV function: RVEDA/LVEDA, paradoxical septal motion (Y/N), TAPSE (tricuspid annular plane systolic excursion) , S' tricuspid annulus, RVFAC (4 chambers, RVEDA-RVESA/RVEDA), IVC diameter, SVC collapsibility index (if TEE; max-min/max), pulmonary acceleration time and its pattern (biphasic versus monophasic), respiratory variations of RV ejection flow (end expi -end inspi/end expi, whatever the condition of ventilation, i.e. a negative value in a spontaneously breathing patient indicates a "normal" response, a "positive" value some RV injury, for instance acute asthma).
Hepatic vein flow pattern: PWD profile in a supra hepatic vein.
8. ECHO 2 LVOT VTI (or RVOT VTI in case of poor acoustic window, but taking the same than in ECHO 1) 1 minute after PLR or after a fluid challenge (mL to be noted).
A responder to fluids is defined if increased VTI is more than 10%. At the time of ECHO 2, only reports blood pressure, heart rate, PPV (if initially recorded), CVP, sinus rhythm Y/N.
Storing and handling of data Data will be collected prospectively on an e-CRF; patients are anonymized using a coding system on a secure online health data platform (Research Electronic Data Capture, RedCap, by the University of Sydney (https://redcap.sydney.edu.au/), which is a member of the RedCap consortium (https://projectredcap.org).
Statistical analysis
The primary objectives are (1) to report the classification performance of hemodynamics parameters, ventilatory parameters and biomarkers in classifying the 3 types of RV injuries, and (2) to report the association of the 3 types of RV injuries on clinical outcomes.
The association between each type of RV injury and the parameters in each domain is examined using logistic regression for binary outcomes, survival analysis for mortality and linear regression after log-transformation for duration data. The observations will be randomly split into two datasets, the train set and test set, in a 3:2 ratio. The train set will be used to build the prediction model, and validation will be carried out in the test set. Confusion matrix will be used to extract prediction performance and accuracy, and a performance matrix will be constructed. The performance matrix contains one or more of the following metrics where appropriate: R2, the accuracy of predictions, AUCs of ROCs, and F1 scores. Agreements and consistencies of these scores between the three definitions will be compared using intraclass correlation coefficient (ICC). The mean performance score in each domain for each type of RV injury will be compared to map the characteristics of each RV injury type. The type of RV injury that has the highest overall performance scores in each domain will be deemed as the best performer in that domain. Regression coefficients, such as OR and hazard ratio, will also be reported to show if the RV injury type is statistically significantly associated with the parameter of interest.
Unless otherwise specified, descriptive statistics will be presented as median \[IQR\], and categorical data as count (%). Test summary will be reported as mean difference, OR and hazard ratio with 95% CI interval.
Power and sample size Comparisons of the predictive performance between the types of RV failure for the four domains will be carried out using ANOVA. Assuming a small to moderate effect size defined by Cohen (i.e. 0.2) in the worst performing domain, for a significant level of 0.015 and a power of 0.9 (to minimize for false positive rate), the required sample size for the train set is 240 patients. Since another 160 patients are required for the test set, a total sample size of 440 is required if a 10% cases are invalid due to missing data and entry error. With twenty-two centres, the investigators anticipate to recruit 440 patients in about 6 months' time (from January 1, 2023 to June 30, 2023).
The database will be managed by Stephan Huang (Nepean Hospital, Sydney) as well as the statistical analysis. The help of a statistician if necessary may be requested from time to time (Dr. Daulasim, PhD public health, ICU, Ambroise Paré hospital).
Ethical and regulatory considerations The data will be collected prospectively and patients will be anonymized using a coding system on a secure online health data platform (Research Electronic Data Capture, REDCap, by the University of Sydney (https://redcap.sydney.edu.au/) REDCap Consortium ( https://projectredcap.org). The Chief Investigator Professor Antoine Vieillard-Baron owns the data of this project. Only project investigators have access to REDCap, and are responsible to enter their own patients' health data. Apart from the investigator himself/herself and the database administrator (SH), other investigators have no access to data from other centres, except after the conclusion of the study. All downloaded data will be de-identified. REDCap is capable of compliance with just about any standard - for example, HIPAA, Part-11, and FISMA standards (low, moderate, or high). Each of those standards has been used across various consortium sites, as well as other standards (including similar international regulations, like GDPR).
Echocardiography has been a common practice in the ICUs participating in the study for many years and each investigator is recognized as an international expert in the field.
Each investigator undertakes to respect patients' rights and the health data research regulations in force in their country of practice.
Financing and Legal Lead Research for funding in progress. There is no conflict of interest. The project leader is Prof. Vieillard-Baron, the co-project leader is Stephan Huang.
