Patients
This prospective, bicentre study was approved by the Institutional Review Board (Comité de Protection des Personnes Sud-Ouest et Outre Mer III, Bordeaux, France N°DC2016/125) and was registered at French National Commission for Data Protection and Liberties (CNIL N°1980317).
Fifty-six non-consecutive patients were included after oral informed consent. (Written informed consent was waived by the Institutional Review Board.) Inclusion criteria were patients scheduled for neurosurgery or elective abdominal surgery, older than 18 years, equipped with radial arterial catheter and esophageal Doppler (CardioQ ODM+, Deltex Medical, Chichester, UK) for cardiac output monitoring. Non-inclusion criteria were preoperative lung disease, intracranial hypertension, left ventricular ejection fraction below 50%, arrhythmia, suspected right ventricular dysfunction, extreme body weight (BMI > 40 or < 15 kg/m2).
Perioperative management
Standard monitoring included noninvasive blood pressure, heart rate, peripheral oxygen saturation and continuous electrocardiography. After preoxygenation, anesthesia was induced using propofol and remifentanil or sufentanil. Propofol or sevoflurane and remifentanil or sufentanil were used for maintenance of anesthesia. Following tracheal intubation, patient’s lungs were ventilated with a mixture of air/oxygen using volume control mode. Tidal volume was set between 6 and 8 ml/kg of ideal body weight, and positive end-expiratory pressure was set between 6 and 10 cmH2O (Primus, Dräger, Lübeck, Germany, or Avance, General Electric Healthcare, Helsinki, Finland). Peripheral oxygen saturation was maintained above 96%, and the respiratory rate was adjusted to maintain end-tidal carbon dioxide concentration between 30 and 35 mmHg. The inspiratory-to-expiratory ratio was set to 1/2.
Hemodynamic monitoring
All patients were equipped with a radial arterial catheter inserted just after the induction of anesthesia (Vygon, Ecouen, France). The catheter was connected to a bedside monitor (Spacelabs Healthcare Company Headquarters, Issaquah, WA, USA, or IntelliVue MP70, Philips Healthcare, Andover, MA, USA) for mean arterial pressure (MAP) and pulse pressure variation (PPV) monitoring. After tracheal intubation, the probe was inserted into the esophagus via the nasal route and the good quality of the signal was confirmed as previously described.
PPV was derived from the bedside monitor by manual calculation of the difference between systolic and diastolic blood pressure. The maximal (Pulse Pressure max) and minimal (Pulse Pressure min) differences were determined during three consecutive respiratory cycles. The mean values of the three measurements were used to calculate arterial pulse pressure variability: PPV = (Pulse Pressure max − Pulse Pressure min)/[(Pulse Pressure max + Pulse Pressure min)/2] × 100, as previously described [19]. PPV was measured directly on the monitor using a screenshot.
Stroke volume, cardiac output, stroke volume variation (SVV), corrected flow time and peak velocity of aortic blood flow were assessed using esophageal Doppler (CardioQ ODM+, Deltex Medical, Gamida, Eaubonne France). SVV was calculated automatically as follows: (Stroke Volume max − Stroke Volume min)/[(Stroke Volume max + Stroke Volume min)/2] over one respiratory cycle. The SVV value was averaged over five respiratory cycles [20]. SVV value was recorded immediately after the measurement of PPV in order to record two values that covered the same time period.
Eadyn was calculated as the ratio between PPV and SVV. Net arterial compliance (C) was calculated as the ratio between stroke volume and pulse pressure [21]. Arterial resistance (R) was calculated as the ratio between mean arterial pressure and cardiac output. Arterial elastance was calculated using these two formulas: EaSAP = (systolic arterial pressure × 0.9)/stroke volume and EaMAP = mean arterial pressure/stroke volume [22, 23].
Study design
Measurements were performed in the operating room, between the end of induction of anesthesia and the end of surgery. Volume expansions and neosynephrine infusions were performed according to the routine care of the patients. Each patient received volume expansion before neosynephrine infusion. If several volume expansions or neosynephrine infusions were done in one patient, only the first volume expansion or the first neosynephrine infusion was recorded.
Volume expansion was performed if MAP ≤ 65 mmHg and SVV > 10%, and neosynephrine was infused if MAP ≤ 65 mmHg regardless of the SVV value, on the discretion of the physician in charge.
Volume expansion was done with 250 ml 0.9% saline over 10 min. One set of measurements was performed immediately before volume expansion, and the second set was done 2–3 min after the end of fluid administration.
Vasopressor infusion consisted of a fixed dose of 50 mcg of neosynephrine. One set of measurements was performed immediately before vasopressor infusion, and the second set was performed 2 to 3 min after the infusion when MAP was stabilized (MAP variation below 5% during 1 min).
Patients with hemodynamic instability requiring a decrease (or an increase) in anesthesia drug dosage, fluid infusion or administration of vasopressors other than in the protocol were excluded. Another exclusion criterion was any change in ventilatory setting by the physician in charge of the patient.
Statistical analysis
Data are expressed as median [percentile, 25–75] or mean ± SD where appropriate. Normality of the distribution was tested using D’Agostino–Pearson test. Pressure response to volume expansion was defined as an increase in MAP ≥ 10% [13, 14]. The effects of neosynephrine and volume expansion on hemodynamic parameters were analyzed using Wilcoxon rank sum test. Mann–Whitney test was used to compare hemodynamic variables before fluid challenge or neosynephrine infusion in pressure non-responder and responder patients. The relationship between Eadyn and changes in MAP induced by volume expansion was tested using Spearman rank test.
The receiver-operating characteristic (ROC) curves were generated for Eadyn and MAP to test their abilities to predict pressure response to volume expansion. Area under the receiver-operating characteristic curves were compared using De Long test [24]. The best threshold values were identified using the Youden Index (specificity + sensitivity − 1). The gray zone was determined as follows: The low cutoff value was defined to exclude positive fluid challenge in 90% of patients, whereas the high cutoff value was defined to predict positive fluid challenge in 90% of cases [25]. A diagnostic test is considered to have good accuracy when its area under the ROC curve is ≥ 0.75 [26]. Fifty-six patients were needed to demonstrate the ability of Eadyn to predict pressure responsiveness with good accuracy, i.e., area under the ROC curve > 0.75 (ratio of pressure responders = 1/3, null hypothesis = 0.50, type I error of 5% and type II error of 10%).
Random effects models were estimated in order to study the effect of several covariates on the temporal evolution of Eadyn, SVV and PPV during fluid challenge and neosynephrine infusion.
For each hemodynamic maneuver (i.e., fluid challenge and neosynephrine infusion), two different models were estimated for Eadyn, PPV and SVV. The first model included covariates related to arterial load and the second cardiac covariates. Each model was a fully adjusted model including all covariates of the study. Only baseline values (before hemodynamic maneuver) were included. To study the effect of covariates on temporal evolution of Eadyn, PPV and SVV, random effects models were estimated with subjects being considered as the random factors. For each random effects model, because only two repeated measurements were performed, we considered time function as linear. Hence, we estimated the fixed intercept, fixed effects, random intercept and time interaction between time and each covariate. The contribution of each covariate on temporal evolution was tested by a Wald test on the time–covariate interaction term [27].
Statistical analysis was performed using Medcalc (software 11.6; Mariakerke, Belgium) and R Development Core Team ([2008]. R: A language and environment for statistical computing; R Foundation for Statistical Computing, Vienna, Austria. ISBN 3-900051-07-0, URL).