This prospective, observational, one-centre study was carried out in the 25-bed medical intensive care unit of a university hospital. It has been approved by our institutional review board (Comité pour la protection des personnes Ile-de-France VII). All patients or their relatives gave informed consent.
The screening criteria were age ≥18 years, a transpulmonary thermodilution device in place (PiCCO2, Pulsion Medical Systems, Feldkirchen, Germany), mechanical ventilation in the volume assist control mode with a Vt of 6 mL/kg of predicted body weight, and the decision taken by the clinicians in charge to perform volume expansion.
Patients were excluded if the PLR manoeuvre was contra-indicated (intracranial hypertension) or possibly unreliable (venous compression stocking, intra-abdominal hypertension ). Other exclusion criteria were spontaneously triggered cycles on the airway pressure waveform, cardiac arrhythmias, impossibility to obtain haemodynamic stability (defined by no change in the norepinephrine dose and no change in systolic arterial pressure <10% within 5 minutes before the inclusion), poor echogenicity impeding the measurement of the IVC diameter and of the velocity time integral (VTI) in the left ventricular outflow tract.
IVC sonography was performed by one 4-year experienced intensivist (TT), who holds a university degree in echocardiography. With the 3.5-MHz cardiovascular ultrasound probe of a Philips CX 50 device (Philips ultrasound system, Philips Healthcare, DA Best, The Netherlands), the IVC was examined in the subcostal window in longitudinal section. Its diameter was measured in M-mode coupled with two-dimensional mode and zoom; 2 cm upstream of the origin of the hepatic veins. Measurements were validated when the M-mode tracing was exactly perpendicular to the IVC.
The distensibility index of the IVC, which reflects the increase in its diameter on insufflation, was calculated as IVCV = (maximum diameter on inspiration - minimum diameter on expiration)/mean of maximum and minimum diameters .
The VTI was measured at end-expiration in the left ventricular outflow tract on the apical five-chamber window. On the apical four-chamber view, the left ventricular ejection fraction was calculated by the biplane method of disks summation (modified Simpson’s rule). The average of three consecutive cardiac cycles was used for all ultrasound measurements in case of sinus rhythm and a representative cardiac cycle was chosen in case of atrial fibrillation . Endocardial contours and VTI envelope were hand drawn. Measurements were analysed offline, blinded to the patient’s response to the PLR test.
All patients had a central venous catheter in the superior vena cava territory and a thermistor-tipped catheter inserted through the femoral artery, as required by the PICCO2 device. Transpulmonary thermodilution measurements were performed by the injection of 15-mL boluses of cold normal saline (<8°C) through the central venous catheter. The average result from three consecutive 15-mL injections was recorded at each time point  and was used to obtain CI, the global end-diastolic volume (marker of cardiac preload) and the cardiac function index (estimate of the left ventricular ejection fraction) . Pulse contour analysis allowed the continuous and real-time calculation of CI after an initial calibration by thermodilution .
The intra-abdominal pressure was estimated from the bladder pressure, which was measured after the instillation of 25 mL of normal saline. The measurement was performed at end-expiration in the absence of abdominal muscle contractions. The transducer was zeroed and placed at the pubic symphysis .
At baseline, all patients were in the 45° semi recumbent position (Supplementary figure 2). A first set of thermodilution and echocardiographic measurements was performed, including CI (measured by thermodilution), PPV, SVV and IVCV. Then, we performed a PLR test as previously described . Pulse contour analysis–derived CI, PPV, stroke volume variation (SVV) and IVCV were recorded at the maximal effect of PLR on CI, which occurs within one minute  (Figure 2). A third set of measurements (CI (pulse contour analysis), PPV, SVV and IVCV) was performed once patients were returned to the semi-recumbent position and a steady state was obtained again.
A “Vt challenge” was then performed by increasing Vt from 6 to 8 mL/kg of predicted body weight for one minute . A fourth set of measurements (CI (pulse contour analysis), PPV, SVV and IVCV) was recorded once CI remained stable. Vt was then decreased back to 6 mL/kg of predicted body weight and another set of measurements was performed after a new stable state was reached, including CI (thermodilution), PPV, SVV and IVCV. Finally, in preload responsive patients, 500 mL of normal saline were infused over 10 minutes. In these patients, a last set of measurements was recorded after the end of fluid infusion (CI (thermodilution), PPV, SVV and IVCV).
Except Vt, ventilatory settings and treatments were unchanged during the study period. The intrabdominal pressure and the central venous pressure were measured at each study step. The CI measured by transpulmonary thermodilution and pulse contour analysis was continuously recorded by the PiCCO Win 4.0 software (Pulsion Medical Systems). The intravascular, intra-abdominal and airway pressure signals were continuously recorded by using a data acquisition software (HEM 4.2, Notocord, Croissy-sur-Seine, France).
Patients in whom PLR, performed at Vt = 6 mL/kg, induced an increase in CI (measured by pulse contour analysis) by more than 10%, were defined as preload responders. Variables were summarised as mean ± SD, median and interquartile range or counts and percentages. Variables before and after fluid administration were compared by a paired Student t-test or a Wilcoxon test. Variables between preload responders and non-responders were compared using a two-sample Student t-test, a Mann-Whitney U-test, a Chi-2 test or a Fisher exact test, as indicated.
Receiver operating characteristic (ROC) curves (with 95% confidence interval, CI) were generated for quantifying the ability of the following variables to detect preload responsiveness: 1) IVCV, PPV and SVV at baseline (Vt of 6 mL/Kg) 2) Changes in IVCV (ΔIVCVVt), in PPV (ΔPPVVt) and in SVV (ΔSVVVt) induced by the Vt challenge, expressed either as the change in absolute value (value during Vt challenge – value at baseline) or as the percent relative change from the baseline value ((value during Vt challenge – value at baseline)/ value at baseline x 100), 4) Changes in IVCV (ΔIVCVPLR), in PPV (ΔPPVPLR) and in SVV (ΔSVVPLR) induced by the PLR test, expressed either as the change in absolute value (value during PLR – value at baseline). The areas under ROC curves (AUROC) were compared by the Hanley-McNeil test. The best diagnostic threshold was determined as the one providing the best Youden index (sensitivity + specificity – 1).
The least significant change of IVCV was obtained from six successive measurements of IVCV performed during haemodynamic stability at Vt=6 mL/kg, by the same operator, removing the probe from the patient’s skin for each measurement, as previously described .
In order to demonstrate a significant difference between groups of ∆IVCVVt, assuming a variability of the IVC measurement of 12% [19-20] with an α risk of 5% and a β risk of 20%, we planned to include 15 preload responders and 15 preload non-responders. Statistical analysis was performed with MedCalc 11.6.0 software (MedCalc, Mariakerke, Belgium).