Animal preparation
Animals were male Wistar rats aged between 12 and 16 weeks. Rats had free access to water and feeding until they were anesthetized. Inhalation anaesthesia was performed in an induction chamber using a concentration of 2–4% of isoflurane. Anaesthesia was relayed by an intraperitoneal injection of 90 mg kg−1 ketamine (Imalgène 1000; Merial, Paris, France), 5 mg kg−1 xylazine hydrochloride (Sigma, Saint-Quentin Fallavier, France) and 50 µg kg−1 buprenorphine (Centravet, Maison-Alfort, France). Anaesthesia was maintained throughout the experiment with additional injections of ketamine (a quarter of the initial dose) using a peritoneal catheter. Anaesthesia efficacy was tested every 10 min using the pedal withdrawal reflex by pinching the foot pad. If the rat withdrew its leg in response to foot pad’s stimulation, then a complementary dose of anaesthesia was administered. After anaesthesia, rats were placed in the supine position on a heating blanket warmed at 38 °C. A tracheostomy was performed, and the rats breathed room air throughout the rest of the experiment. The right carotid artery was cannulated with a polyethylene catheter (PE ED 1.16 mm) which was used for exsanguination. The right femoral artery was cannulated (PE ED 0.69 mm), and this catheter was connected to a pressure transducer (Edwards Lifescience, Guyancourt, France). The right jugular vein was cannulated (PE ED 0.92 mm) to administer fluid during resuscitation. Catheters were filled with NaCl 0.9%. The pressure transducer was linked to a data acquisition system (MP30, Biopack system, Paris, France). An additional pressure transducer (Codan France, Bischwiller, France) was necessary in the CL groups. It was connected to the carotid artery catheter and linked to the CL device.
Experimental procedure
Haemorrhagic shock was induced using a Wiggers’ model with fixed pressure [12]. Rapid exsanguination was performed for a period of 5 min to decrease mean arterial pressure (MAP) to a level of 30 mmHg. The blood withdrawn during exsanguination was stored in a heparinized syringe. MAP was maintained between 30 and 35 mmHg for 1 h by withdrawing or if necessary re-infusing blood. At the end of the exsanguination phase, rats were randomized into five groups for the resuscitation phase: three groups were resuscitated with fluid (Ringer’s Lactate solution) and two groups were resuscitated with a combined treatment of fluid and norepinephrine (Fig. 1). The resuscitation phase lasted 1 h.
Among groups resuscitated with fluid, one received standardized manual resuscitation. Fluid was administered at a fixed rate (2 mL kg−1 min−1). Infusion was started and stopped by an anaesthesiologist intensivist to target a SAP of 85 mmHg (M-F group). Two groups received automated treatment using CL based on different algorithms. The first one used a proportional–integral (PI) controller (CL-PI group) using continuous SAP acquisition to adapt infusion rate, while the second one used a fuzzy logic (FL) controller (CL-FL group) using mean SBP calculated every minute to adapt infusion rate (see Additional file 1 for a more complete description of closed-loop algorithms). To limit clot dislocation and high oscillation of arterial pressure, we limited the maximal fluid rate at 4 mL kg−1 min−1 if SAP was < 70 mmHg and at 2.5 mL kg−1 min−1 for SAP ≥ 70 mmHg.
Among groups resuscitated with combination of fluid and norepinephrine, one received manual resuscitation using a standardized protocol (M-FNE group). To achieve a progressive increase in fluid and norepinephrine, we used a protocol with alternating increase in fluid and norepinephrine. Three boluses of 10 mL kg−1 administered at 2 mL kg−1 min−1 of Ringer’s lactate were allowed until the maximum ceiling of norepinephrine. Norepinephrine was started at a rate of 0.1 µg kg−1 min−1 and was increased every 3 min if the target was not reached. If SAP exceeded 88 mmHg, norepinephrine rate was decreased at a rate equivalent to the mean value of the two last steps. This protocol was based on observed volumetric interactions used in clinical practice with adaptation to rat model which requires higher doses of norepinephrine. Infusion was started and stopped by an anaesthesiologist intensivist to target an SAP of 85 mmHg.
The algorithm used in the group treated with automated CL combining fluid and norepinephrine (CL-FNE group) used a combination of PI and FL. The CL-FNE combined a PI regulator for fluid and a FL regulator for NE. Several conditional rules were included to mimic the physician decisions. A schematic of the system set-up is presented in Fig. 2.
Arterial pressure was extracted every 200 ms. Analysis of performance during resuscitation was divided into two phases: the rising phase of arterial pressure until 80 mmHg of SAP and the stabilization phase into the target zone. During the stabilization phase, the performance was evaluated using the following variables as described elsewhere [13]: the percentage of time passed in the target zone of SAP 80–90 mmHg, the percentage of time passed under the target zone (SAP < 80 mmHg), the percentage of time passed over the target zone (SAP > 90 mmHg), the performance error (PE), the median performance error (MDPE), the median absolute value of performance error (MDAPE), the wobble (the median absolute deviation of each PE from the MDPE) and the global score (overall performance of the system) [14].
In all groups, fluid was administered using an infusion pump (Alaris GH, Carefusion, Voisins-le-Bretonneux, France). Norepinephrine was infused using an appropriate infusion pump for small animals (Harvard Apparatus Pump 33, Harvard apparatus, Les Ulis, France).
Blood gas analyses was performed (RAPILab348EX, SIEMENS Healthcare diagnostic, Saint Denis, France) and arterial lactate (THE EDGE, ApexBio, Taiwan) and haematocrit were measured at the end of the haemorrhagic phase and immediately after the end of resuscitation using 100-µL blood gas capillary tubes.
CL system
The CL system was based on a sbRIO 9626 EOM device (National Instruments) programmed with Labview connected to a WiFi module. The system was connected to a pressure transducer and to infusion pumps (Fig. 2). The state of the system was displayed on a personal computer in real time through a specific graphical user interface.
Statistical analyses
Calculating the number of animals needed for the protocol is not conventional in this type of experiment since the intergroup and intra-group variance is rarely predictable [assuming the use of analysis of variance (ANOVA)]. Given the fact, on the one hand, that only large effects are sought and, on the other hand, that these experiments are complex, six–eight individuals per group are generally used. However, since the model of haemorrhagic shock is known in the literature to have significant variability, we decided to include 10 rats per group.
Normality of distribution of variables was tested by the Shapiro–Wilk test. Since the studied variables were non-normally distributed, all data are presented as median—interquartile range.
Among the three groups resuscitated with fluid, the effect of the resuscitation mode on the performance parameters and on the volume of fluid administered during resuscitation was analyzed globally using Kruskal–Wallis test. Comparisons of CL-PI or CL-FL with the M-F group were made using Dunn’s test. Among the two groups resuscitated with combined treatment, the effects of resuscitation mode were analyzed using Mann–Whitney test.
In addition, we tested the hypotheses that systemic variables such as blood gas, lactate and haematocrit measured at the end of the exsanguination period and after resuscitation can be different between groups by nonparametric two-way ANOVA for repeated measures [15].
To assess the effect of norepinephrine, we compared the groups with best performances in the presence or absence of norepinephrine by Mann–Whitney tests.
Statistical analysis was performed with GraphPad Prism 6 (GraphPad Software, San Diego, CA) and the R software (http://cran.r-project.org/) using the nparLD package.