Proceedings of Réanimation 2017, the French Intensive Care Society International Congress

© The Author(s) 2017. This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. PHYSICIAN ABSTRACTS

Introduction Donation after circulatory death corresponds to the cat-Introduction Predicting the neurological outcome of patients admitted to the ICU after a cardiac arrest successfully resuscitated remains difficult [1]. The bispectral index (BIS) allows for the rapid and standardised assessment of the cortical function based upon eletroencephalogram analysis whereas the ratio of suppression (RS) is indicative of the absence of electrical activity of the brain. We aimed to evaluate the prognostic value of the BIS and the RS for predicting neurologic outcome after cardiac arrest. Patients and methods This was a prospective, single center, observational study performed in a large regional University hospital. Adult patients admitted to the ICU for cardiac arrest between March 2012 and October 2014 were included in the study. The exclusion criteria was pregnancy. The BIS and the RS were collected as soon as possible after ICU admission. The patients were not included in the analysis if they died within 24 h, if they had a low signal quality [defined as high EMG artefacts (≥30 dB)], or if the monitoring of BIS started 24 h after ICU admission. The neurological outcome of the patients was based upon the cerebral performans category (CPC) calculated at 3 months. CPC score of 1 or 2 indicated good outcome, whereas CPC score of 3-5 indicated poor outcome. Results During the study period 148 patients were admitted to our ICU for a cardiac arrest. The BIS and RS were monitored in 103 patients.17 patients were excluded (early death ≤24 h; low quality of signal; BIS and RS performed ≥24 h after ICU admission). Thus, 86 patients were enrolled in this study. The means age was 57.6 ± 16.8 years, 61 patients (70.9%) were male, the cardiac arrest was out-of-hospital in 63 patients (73.3%), hypoxia was the main cause of cardiac arrest (43%), 60 patients (70.6%) were treated with therapeutic hypothermia. At 3 month of follow-up, a total of 50 patients (58.1%) had died and 55 patients (63.9%) were classified as having a poor outcome. The mean duration from the return of spontaneous circulation (ROSC) to the BIS and RS measurements was 5.7 ± 3.0 h. The BIS values were significantly lower in patient with poor outcome compared with patients with a good outcome (5.9 ± 11.1 vs 37.1 ± 18.0, p < 0.0001). The RS values were significantly higher in patient with poor outcome group compared to those with good neurological outcome (85.9 ± 26.3 vs 18.4 ± 31.3, p < 0.0001). The BIS predicted poor outcome with a likelihood ratio of 23.8 and an area under the curve (AUC) of 0.918 [95% CI (0.839-0.966)]. The optimal sensitivity [78.4%, 95% CI (67.3-89.5)] and specificity [96.5%, 95% CI (89.8-100)] for neurological outcome prediction was obtained using a cut-off value of BIS < 5. The RS predicted poor outcome with a likehood ratio of 23.8 and an AUC of 0.936 [95% CI (0.862-0.977)]. The optimal sensitivity [78.0%, 95% CI (66.9-89.0)] and specificity [96.8%, 95% CI (90.8-100)] for neurological outcome prediction was obtained using a cut-off value of RS > 84. In multivariable logistic regression model, BIS or RS predicted poor outcome with an odds ratio of 65.0 [95% CI (6.1-689.2), p = 0.0005]. Discussion The results of this study using the EEG derived parameters BIS and RS confirm previous findings showing that they are linked to the neurological outcome of patients admitted to the ICU after a cardiac arrest [2]. In particular, a BIS < 5 and/or a RS > 84, measured at ICU admission, were both strongly associated with a poor neurological outcome at 3-months.
Conclusion BIS values may be used to predict long term neurological outcome of patients following cardiopulmonary resuscitation. The ability to accurately predict early non-recovery after cardiac arrest could facilitate discussions with families and limit use of ICU resources in futile cases. BIS and RS values measured at ICU admission might be considered as additional prognostic tools available for the intensivist.
Introduction Avoiding the use of iodinated contrast media (CM) is frequent, fearing it may contribute to AKI. The aim of this systematic review and meta-analysis was to quantify the risk of AKI attributable to CM in ICU patients. Materials and methods A systematic review until December 31st 2015, through 5 databases, searched for studies evaluating intravascular administration of iodinated CM. Only controlled studies evaluating AKI following CM exposure in ICU patients matched to unexposed patients were included in the meta-analysis. Meta-analysis was performed on patient-level data using a hierarchical Bayesian nested mixed effects multiple logistic regression model. Bayesian methodology allows evaluating how evidence-based physicians would assess the AKI risk attributable to CM according to both their a priori belief and the presentation of the controlled studies identified in the systematic review. Two meta-analyses were performed with different a priori hypotheses. An objective meta-analysis modeled a neutral state of a priori belief (Odds Ratio [OR] of 1 with impartial distribution) yielding a posteriori OR distribution representative of data collected in controlled studies. A subjective meta-analysis modeled the common belief that CM increases the AKI risk, using an a priori OR of 1.37 based on uncontrolled studies holding clinical community consensus. We determined the minimum a priori relative effective sample size (RESS, representing the a priori strength of belief ) needed to observe a significant a posteriori OR distribution: how much physicians have to be convinced a priori that CM increases AKI risk to maintain this belief after being confronted with the studies data. Results Among 5696 references, 10 compared ICU patients receiving CM with an unexposed group and 4 performed risk adjustments for baseline AKI risk. Three studies used patient matching: overall, 280 CM patients were matched with 280 control patients. The resulting a posteriori OR did not reach statistical significance: with no prior assumption, there is no evidence that CM increases the risk of AKI in the ICU. Using an a priori OR of 1.37 (subjective meta-analysis), the a posteriori distribution of the OR did not reach statistical significance except when modelling a very high a priori belief that CM causes AKI (minimum a priori RESS 4.8-folds higher than the RESS of the objective meta-analysis and 70-folds higher than a neutral objective a priori hypothesis). Conclusion This systematic review and meta-analysis did not enlighten a risk of AKI attributable to iodinated CM in ICU patients. Introduction Acute kidney injury is particularly common in septicshock patients and is associated with high mortality. The putative effect of renal replacement therapy (RRT) on the prognosis of such patients is highly debated; some advocating that outcome might improve owing to modulation of inflammation. We aimed to compare outcomes of septic-shock patients with severe acute kidney injury (stage 3 of KDIGO classification) treated with an early RRT strategy (all patients immediately received RRT) with those treated with a delayed RRT strategy (patients received late RRT or no RRT at all).

Patients and methods
We did a post hoc subgroup analysis in a subset of septic-shock patients with severe acute kidney injury (stage 3 of KDIGO classification) from a multicenter randomized controlled trial. In the trial, patients from 31 intensive care were randomly assigned (1:1) to either an early or a delayed RRT initiation strategy. With the early strategy, RRT was initiated within 6 h after inclusion criteria were met. With the delayed strategy, RRT was started if at least one of the following criteria was met: severe hyperkalemia, metabolic acidosis, pulmonary edema, serum urea concentration greater than 40 mmol/l, or oliguria for more than 72 h after randomization. The primary outcome was overall survival at day 60.

Results
Of the 413 septic-shock patients (on a total of 620 patients), 209 were managed with early strategy and 204 with delayed strategy. The Kaplan-Meier estimates of mortality at day 60 did not differ significantly between the early and delayed strategies; 101 deaths occurred among 209 patients in the early-strategy group (48.5%; 95% confidence interval [CI] 41.3-54.9), and 99 deaths occurred among 204 patients in the delayed-strategy group (48.5%, 95% CI 41.2-55.0; P = 0.97) (Fig. 3). A total of 97 patients (47.5%) in the delayed-strategy group did not receive renal-replacement therapy. The number of days RRT-free days was significantly higher in the delayed strategy group (21 [5-29] vs 17 , p < 0.001). Median length of stay in hospital did not differ significantly between groups (20  vs 19  days, p = 0.9). Conclusion The timing of RRT in septic-shock patients with severe acute kidney injury did not significantly influence mortality. However, a conservative strategy avoided many unnecessary RRT sessions.
Introduction Doppler-based renal resistive index (RI) measurement may hold promise in differentiating transient from persistent AKI in selected critically ill patients. Although several studies have suggested adequate performance in predicting short-term reversibility of AKI, most of these studies were performed in limited patient samples [1]. Additionally, a recent study has identified discrepant results regarding its diagnostic performance [2] suggesting confirmatory studies to be required.
The main objective of this study was to assess diagnostic performance of RI in predicting persistent AKI in critically ill patients. Secondary objectives were to assess diagnostic performance of semi-quantitative assessment of renal perfusion (SQP) using color-Doppler in predicting persistent AKI and performance of both tests in predicting needs for renal replacement therapy (RRT).
Patients and methods Prospective multicenter study performed in eight ICUs from December 2013 to April 2016. This study was declared to Clinicaltrial.gov. Adult patients requiring mechanical ventilation were included in this study. Patients with mild to severe chronic kidney diseases, arrhythmia, or obstructive renal dysfunction were excluded from this study. Patients with hospital stay shorter than 72 h and in whom renal reversibility could not be assessed were secondarily excluded. Acute kidney injury (AKI) was defined according to the KDIGO definition. Transient AKI was defined as AKI with recovery within the first 3 days following inclusion. Intra-renal RI was calculated as (peak systolic velocity − end-diastolic velocity)/peak systolic velocity. SQP was assessed using a scale ranging from 0 (absence of renal perfusion) to 3 (renal vessels identifiable until the arcuate arteries in the entire field of view Introduction Tuberculous meningitis (TBM) is frequently associated with neurological complications requiring admission in intensive care unit (ICU). Adjunctive steroids reduce mortality, but may have no effect on disabling neurologic deficits in survivors. Moreover, the benefit of steroids is controversial in immunocompromised patients and in patients with severely altered mental status at hospital presentation. We aimed to identify indicators of poor functional outcome in adult patients with severe forms of TBM requiring ICU admission.
In particular, we investigated the effect of adjunctive steroids on functional outcomes and 1-year mortality.

Patients and methods
We conducted a retrospective cohort study (January 1st 2004 to June 15th 2016) in consecutive TBM cases admitted to the medical ICUs of 12 university-affiliated hospitals, located in the Paris area, France. We used multivariate logistic regression analysis to identify factors associated with a poor functional outcome. A poor functional outcome was defined as a score of 3-6 on the modified Rankin scale (mRS) 90 days after ICU admission. In a secondary analysis, we used Cox proportional hazards modeling to evaluate the risks of death at 1 year. Data are presented in median (interquartile range) or numbers (percentage). Results A total of 95 patients were studied, including 44 (46%) immunocompromised patients. At ICU admission, the Glasgow Coma Scale score was 11 (8)(9)(10)(11)(12)(13)(14), 64 (67%) patients had Medical Research Council (MRC) grade 3 illness, and 66 (69%) required invasive mechanical ventilation. Brain MRI was performed in 79/95 (83%) cases. Antituberculous therapy was administered to 90/95 (95%) patients (5 died before treatment could be started), and 76/90 (84%) patients received adjunctive steroids. The duration of ICU stay was 10 (4-24) days.
The following factors were independently associated with mortality: CSF protein levels >1.9 g/L [adjusted relative risk (aRR): 2.47, 95% CI 1.17-5.25, p = 0.018], hydrocephalus on MRI (aRR: 3.45, 95% CI 1.50-7.91, p = 0.003), brain infarction on MRI (aRR: 2.40, 95% CI 0.99-5.81, p = 0.051). The use of adjunctive steroids had a protective effect on 1-year mortality (aRR = 0.16, 95% CI 0.05-0.45, p = 0.0006). Conclusion Despite antituberculous therapy and supportive care, severe forms of TBM are characterized by a poor outcome in more than 50% of cases. Elevated CSF protein levels, hydrocephalus and brain infarction on MRI at ICU admission represent major indicators of poor outcome. Our data suggest that use of adjunctive steroids is associated with reduced disability and mortality, irrespective of immune status and severity of illness. We conclude that adjunctive steroids may benefit to all patients with a suspicion of TBM admitted to the ICU. Introduction Early Standard Rehabilitation (ESR), first passive and then passive/active, is recommended for critically ill patients in whom it reduces the duration of mechanical ventilation (MV), improves functional status, muscle strength and quality of life after hospital discharge. The early addition of leg bicycling on a cyclo-ergometer and of electrical muscle stimulation (EMS) is now part of common practice in the ICU. Whether it can preserve or improve muscle strength and further increase the beneficial effects of ESR is little known. Patients and methods Single-centre, randomised study comparing the effects of the combination of early and daily leg bicycling + EMS of the quadriceps + ESR (intervention group) versus ESR alone (usual care group) on the global muscle strength assessed by the MRC score at ICU discharge by a physiotherapist blinded to the randomization group (NCT02185989). All consecutive patients were potentially eligible if they were deemed to need more than 72 h of care in ICU. Main non-inclusion criteria were resuscitated cardiac arrest, presence of pacemaker or implantable defibrillator, acute cerebral disease requiring deep sedation for at least 72 h, known neuromuscular disease, and amputation of a lower limb. Randomization was stratified by sex, MV or not at study entry, and day of admission (Thursday/ Friday vs other days). The interventions were applied right from Day 1 (within 72 h of admission), 5 days/week. Protocoled ESR consisted of daily multistep program (from 10 passive mobilisations of each joint in comatose patients to passive/active muscle work, transfer to chair, standing and walking, depending on patient's level of wakefulness/ cooperation). In the intervention group, 30 min passive/active leg bicycling (even in bed-ridden patients) and 54 min EMS of the quadriceps were performed 5 days/week in addition to ESR, according to pre-established programs. Conclusion Although safe and resulting in lower decline in muscle thickness as observed on echography (not blinded assessment), the addition of daily leg bicycling and EMS to ESR did not result in higher global muscle strength as assessed by the MRC score (blinded assessment) at ICU discharge in a mixed and heterogeneous population of critically ill patients. Exploratory subgroup analyses are underway and perhaps will help to identify subsets of patients in whom the studied intervention might be beneficial and might deserve further investigations.

Competing interests
None.

O16
Post-intensive care syndrome: a population-based observational study of healthcare use Youenn Jouan 1 , Noémie Teixera 2 , Claire Hassen-Khodja 3 , Antoine Guillon 1 , Christophe Gaborit 3 , Leslie Grammatico-Guillon 3 , Stephan Ehrmann 1 Introduction Consequences of hospitalization in critically ill patients have been recognized for several years with physical, cognitive and psychological consequences published under the denomination of the post intensive care syndrome which will become a new challenge for intensivists. The impact of social, personal and financial consequences has been less reported. The primary objective of this pilot study is to report the social, financial, psychological, physical and emotional consequences in a group of critically ill patients compared to a group of patients never hospitalized in ICU. The second objective is to investigate patient's perceptions to better understand their memories of their hospitalization through a qualitative approach.

Patients and methods
We designed a case control study in three ICUs belonging to the Outcomerea research group (July 2014-May 2015). Case patients were adult patients ventilated for more than 48 h. We excluded patients not speaking or understanding French and patients who denied participation. Clinical and demographics characteristics of the cases were extracted from the Outcomerea database. They were interviewed 6-months after ICU discharge. Control patients, matched on age and sex and never hospitalized in ICU, were interviewed faceto-face during an hospital consultation. All patients completed the same questionnaires in a random order, exploring emotional and post traumatic-stress syndrome (Impact of Event Scale-revised, cut-off >22), self-sufficiency in daily activities (activity of daily Living, ADL), quality of life (first question of the SF-36), and questions about their place of living, of working and financial conditions, need of psychological help and marital status. We used a phenomenological approach to report patient's perceptions.

Results
Of the 96 eligible patients, 20 (20.8%) died at 6 months, 39 (40.6%) were excluded and 37 were entered in the analysis and compared to 37 control patients. Characteristics of the case patients were: age (median: 65, range: 47-73), 64% male, SAPS II (51, 37-64), ICU stay (12 days, 8-19), hospital stay (29 days, 22-41). The median IES-R score was significantly higher in cases (14, vs control (6, 3-10), p < 0.01. IES-R > 22 was found in 13 (35.1%) cases patients versus 1 (2.7%) of control patients (p < 0.01). Activities of daily living without help were significantly most often performed in control versus cases for bathing (n = 37, 100% vs n = 33, 89%, p = 0.04) and continence (n = 37, 100% vs n = 33, 89% p = 0.04). Perception of the quality of life was not significantly different between cases and controls but increasing quality of life was much important in cases (n = 15, 40.6%) versus control (3,8.1%), p < 0.01. See Table 4. Three themes were found in the qualitative analysis: the ICU stay seen as a traumatized period, a period without memories and support from families and friends. Their representations in the verbatims were 13 (35.1%), 13 (35.1%) and 3 (8.1%). Conclusion This pilot study reported substantial neuropsychological and functional alterations related to the ICU stay and emphasized the need for better estimating these modifications in a multicenter study. Addressing these consequences adds to the role of intensivists for elaborating prevention programs and promoting post intensive care syndrome to non ICU practitioners to collaborate together for the best future of ICU patients.

Introduction
In the ICU context, in the case of organ donation, patients' relatives are at the centre of the decision process: within a limited time frame, they will be told that the patient is brain dead and will be asked to consider organ donation. Qualitative studies have put forward that understanding brain death facilitates decisionmaking and impacts on the final decision (donation vs non donation). However the impact of understanding brain death on relatives' grieving process has never been evaluated. In this study, we searched for correlation between semi-quantitative answers to questions related to understanding of brain death and experience of the process in a questionnaire completed by relatives 1 month after the patient's death and post-traumatic related symptoms (PTSD) and complicated grief. Patients and methods This is an ancillary study of a larger prospective, observational study in 28 ICUs in France that aimed to compare grief symptoms of relatives of donor patients versus relatives of non-donor patients. For each brain dead patient, the relative who served as the surrogate was included at time of organ donation discussion. Relatives were assessed at 3 time points during a telephone interview: at 1 month, to complete a questionnaire regarding their experience in the ICU and description of the organ donation request and procedure, including understanding of brain death; at 3 months to complete the Hospital Anxiety and Depression Scale (HADS) and the Impact of Event Scale-Revised (IES-R) for PTSD symptoms; at 9 months, to complete the IES-R and the Inventory of Complicated Grief (ICG).
Results 202 relatives were included in the study. At 1 month after the patient's death, 79.2% of relatives completed the questionnaire, at 3 months 70.3% completed the HADS and the IES-R and at 9 months 61.4% completed the IES-R and the ICG. One month after the death, 35% of relatives declared having difficulties in understanding brain death and 32% experienced decision-making as difficult. Results show that experience of the decision making process impacts on relatives' well-being. At 3 months, compared to relatives who did not find the decision difficult, those who did find it difficult more often presented significant PTSD symptoms (40.54 vs 65.88%, p = 0.016). At 9 months, compared to relatives who understood brain death, those who did not understand brain death had higher global ICG score [23 (12.5-36.5) vs 36 (28-43.75), p = 0.010] and more often presented complicated grief symptoms (46.15 vs 75%, p = 0.026). There was a trend in increased prevalence of PTSD related symptoms with 60% in the group of relatives who did not understand brain death versus 47.2% in the group that did, but this was not significant (p = 0.33).
Discussion Results show that difficulty experienced during organ donation discussion and decision impacts on relatives' well-being in the months that follow the patient's death. Support to relatives should be proposed in this context. Interestingly, understanding of brain is a key component of relatives' experience: on top of possibly impacting on the decision itself, it significantly impacts on relatives' grieving process 9 months after the patient's death. Promoting better understanding of brain death, proposing clearer explanations, by using various media, may improve both relatives' understanding and well-being. Conclusion Our study shows that understanding of brain death is a key component of relatives' experience that significantly impacts on the grieving process. Efforts should be made to improve relatives' understanding of brain death.

Competing interests
None. is recognized as a priority. Nevertheless, specific data are needed to understand what specific aspects of ICU care affect the relatives' grieving process. This study aims at providing information on potential links between anxiety/depression, PTSD related symptoms, complicated grief and components of ICU end-of-life care in order to specify ICU practices that may affect the risk of developing these symptoms. Patients and methods This is an ancillary study of the CAESAR study-a prospective, observational study in 41 ICUs in France. Eligible patients were adults who died after at least 48 h in the ICU. For each patient, the relative who served as the surrogate was included at time of death. Relatives were assessed 21 days then 3, 6, and 12 months after the death during a telephone interview. At 21 days they completed the CAESAR scale; at 3 months, they completed the Hospital Anxiety and Depression Scale (HADS) and the Impact of Event Scale-Revised (IES-R) and at 6 and 12 months they completed the IES-R and the Inventory of Complicated Grief. In this study, we searched for correlation between semi-quantitative answers to CAESAR questions (and not scores) and outcomes.

Impact of ICU end-of-life care on relatives' grief symptoms
Results 475 patients and their relatives were included. Response rates were 90.5, 81.3, 59.4 and 45.2%, at day-21, 3, 6 months and at 1 year, respectively. 5 domains are associated with significant increased risk of developing ICU burden (p ≤ 0.05 for each variable).
1. Quality of care and symptom control. Perception that pain was not under control and that the patient had difficulties in breathing is associated with increased risk of developing anxiety and depression at 3 months, PTSD related symptoms at 6 and 12 months, complicated grief at 6 months. Dissatisfaction with quality of care is associated with increased risk of developing anxiety and depression at 3 months, PTSD related symptoms at 3 months, complicated grief at 12 months.

Quality of communication. Dissatisfaction with communication with
either doctors or nurses is associated with increased risk of developing anxiety and depression at 3 months, PTSD related symptoms at 3 and 6 months, complicated grief at 6 months. 3. Kindness. Perception that the team was not kind enough is associated with increased risk of developing anxiety and depression at 3 months, PTSD related symptoms at 3, 6 and 12 months. 4. Preparation for death. Relatives who were not informed that the patient was dying, who were unable to express important things or to say goodbye were more at risk of developing anxiety and depression at 3 months, PTSD related symptoms at 3, 6 and 12 months and complicated grief at 6 and 12 months. 5. Presence at time of death is associated with increased risk of developing PTSD related symptoms and complicated grief at 6 and 12 months.

Discussion
Relatives are sensitive to interaction between the ICU team and themselves as well as between the team and the patient. Quality of communication (both verbal and non verbal) and support, as well as preparation for the death, are key components of relatives' experience that impact on grief symptoms in the months that follow the patient's death.
Conclusion Quality of care and support during the dying process are at the heart of the relatives' experience. This study puts forward practices that may be improved in order to promote both palliative care and family centered care in the ICU and, in fine, decrease grief symptoms in bereaved relatives.

Competing interests
None.

O20
Complicated grief after organ donation of brain-dead patients: evaluation of the donors'relatives 6 months after the death Hélène Foulgoc 1 , Julien Rogier 1 , Elsa Delobbe 1 Introduction In 2015, 1824 organ donations were performed in France. We already know that complicated grief after the death of a relative in the intensive care unit is frequent but the population of organ donors'relatives hasn't been studied. Currently, there doesn't exist any sector to orientate donors' families. The purpose of our study is to investigate the complicated grief donors' relatives, its associated diseases and risk factors. Patients and methods Prospective, single-center, observational study. During the first meeting with the transplantation coordination team of our hospital, it was proposed to the organ donors' relatives to participate. The relatives who had accepted received the ICG, IDS-SR and PCLS questionnaires at 1, 3 and 6 months after the death. The primary endpoint was the presence of complicated grief, defined by an ICG > 25, among the relatives of organ donors after 6 months from the death. The secondary end points were the presence of major depressive disorders (IDS-SR ≥ 15) and post-traumatic stress disorders (PTSD) (PCLS ≥ 44) at 6 months and the analysis of complicated grief's risk factors: about the deceased person (age, sex, cause of the death, place of the death, time between the hospital's entrance and the death) and about the relative ( Discussion Even if only 36% of the relatives completed the scales at 6 months, this study shows a high risk of complicated grief among donor's relatives. But we can see that the prevalence of complicated grief in this population is closed to the prevalence find after the death of a relative in the intensive care unit. Because of the lake of responses at 6 months, we can't conclude on the risk factors. Conclusion 55.2% of donor's relatives developed a complicated grief 6 months after the death. The establishment of dedicated channels to orientate these bereaved persons seems important. Further studies should be conducted to define better the complicated grief's risk factors following an organ donation. Introduction Long after the death of a loved one, end-of-life decisions can remain with the living and have been implicated in post ICU burden. In the case of organ donation, the relatives are at the centre of the decision process: within a limited time frame, they will be told that the patient is brain dead and will be asked to consider organ donation. Attention has been focused on how relatives make the decision to donate or not to donate the patient's organs but only very few studies have focused on the impact of organ donation decision making on their psychological well-being during the months that follow the patient's death. The goal of this study was to describe the grieving process of relatives who were approached about organ donation in the context of brain death and to compare grief symptoms of relatives of donor patients (DPs) versus relatives of non-donor patients (NDPs).

Patients and methods
We conducted a prospective, observational study in 28 ICUs in France. For each brain dead patient, the relative who served as the surrogate was included at time of organ donation discussion. Relatives were assessed at 3 time points during a telephone interview: At 1 month, to complete a questionnaire regarding their experience in the ICU; at 3 months to complete the Hospital Anxiety and Depression Scale (HADS) and the Impact of Event Scale-Revised (IES-R) for PTSD symptoms; at 9 months, to complete the IES-R and the Inventory of Complicated Grief (ICG). The primary outcome measure was the IES-R (at 3 and 9 months).
Results 202 relatives were included in the study among which 158 were relatives of DPs and 44 of NDPs. At 1 month after the patient's death, 79.2% of relatives completed the questionnaire, at 3 months 70.3% completed the HADS and the IES-R and at 9 months 61.4% completed the IES-R and the ICG. More than the decision itself, quality of the organ donation process impacts on relatives' grief symptoms. Conclusion The decision (donation vs no donation) has no impact on grief symptoms in the months following the patient's death. However, experience of the request and of the decision itself, as well as quality of communication and support, are elements that effect relatives' experience and grieving process.
Introduction In developed countries, the application of the "Surviving Sepsis Campaign" (SSC) protocols in the management of sepsis and septic shock has been associated with increased survival [1].
In the context of a fragile health system, the application of these protocols may be expected to prove to be difficult to undertake and the impact may also be expected to vary [2]. Patients and methods We conducted a prospective, quasiexperimental before and after study in the university hospital of Kinshasa (DRC) comparing 33 consecutive patients in septic shock treated according to usual care versus 39 patients treated according to a protocol type EGDT from 1st February 2014 to 31st July 2014 (pre-protocol phase) and 1st September 2014 to 28th February 2015 (post-protocol phase). Between the 2 phases, we have drawn and implemented a local protocol based on SSC recommendations. A kit consisting of central venous catheters, devices capable of measuring central venous pressure, flow regulator (Dosiflow R) and lactate reader was available.
In the absence of electrically driven syringe pumps, catecholamines were administered by continuous perfusion using a flow regulating device (Dosiflow). We drew up calculating tables based on a standard dilution and a variable administration rate controlled by the flow regulating device, thus permitting precise and constant dose administration in millilitres/hour (ml/h). The main outcome measures were: the rate of compliance with 6-h sepsis care bundles and the hospital mortality until J30. The Student's and U Mann-Whitney tests were used to compare quantitative variables, and the chi-square test to compare qualitative variables.
Kaplan-Meier survival curves were used and the differences between the two curves were analyzed using the log rank test. This study received the approval of ethics committee and was registered 23th January 2014 under number ESP/CE/053/14. Results Baseline characteristics of patients were similar in both two groups. Lung infection was the main source of septic shock and antibiotics administration delay was less than 3 h in the two groups. In contrast, blood cultures and blood lactate were performed only in patients of the EGDT group.
Patients of the EGDT group received more IV fluids (+1226 ml on average), more catecholamines (+24, 9%) and were more often transfused (+13, 7%) than the patients treated in the habitual fashion. Protocol compliance was substantially improved, passing from 0 to 50%. The absolute risk reduction of mortality was 17% (100 vs 83%; p = 0.0037) when all the therapeutic measures had been employed.
The Kaplan-Meier survival curves showed that the patients in the preprotocol group had a significantly greater risk of dying compared with those in the post-protocol group. The average length of stay was 2.2j vs 5.2j p = 0.0037. Discussion Our results in terms of compliance with the sepsis bundle were satisfying given that even in highly developed countries the compliances rates are often below 50% [3]. Ignorance of the very existence of these recommendations and logistical factors are often quoted as the reasons why compliance is so poor The positive impact that we observed on mortality was found in similar studies [4]. Other more recent randomized controlled studies have not found significant differences but this may be explained by the similarity of treatment protocols in the two groups [5].
Our study is coherent according to current literature but there exist certain limits which might be expected in this kind of study.
Our research was implemented in only one centre and these results cannot be extrapolated to all sub-Saharan Africa, nor even to our own country. As in all before/after studies, comparison with a historic control group is liable to introduce a bias based on confusion. We sought to minimize this bias however by choosing a short study period (12 months) with the same therapeutic team during the entire study.
Conclusion Despite a difficult socio-economic context, the implementation of a local protocol based on the recommendations of the SSC was associated with improved outcome in septic shock patients in our hospital compared to usual care. We thus intend to continue and to widen the application of this protocol in an effort to evaluate if this positive effect will be observed in a greater number of patients.

Introduction
Despite being a fully preventable infectious disease, tetanus is still responsible for about 500,000 death worldwide. In developed countries, the incidence is strongly associated with a lack of vaccination coverage in the elderly. In the USA or in France, people over 65 year-old have a twice to tenfold increase in annual incidence of tetanus versus younger patients [1]. Introduction To date, there is no consensus on which anesthetic protocol should be used before neonatal intubation. As a result, awake intubation, although strongly discouraged by current recommendations, is still common in neonates, especially in France. Many experts consider Propofol, a short-acting anesthetic agent, an appropriate drug for premedication before neonatal intubation. Propofol is known to decrease systemic vascular resistance leading to low arterial blood pressure, which consequently raises concerns about its hemodynamic tolerance. However in neonates, low mean arterial blood pressure (MABP) is poorly correlated with low systemic or cerebral blood flow. NIRS (Near InfraRed Spectroscopy) allows cerebral tissue oxygenation monitoring, reflecting cerebral blood flow. The aim of our study was to compare cerebral oxygenation profiles between a combination of a synthetic opioid plus a muscle-blocker and propofol, used as premedication prior to neonatal endotracheal intubation.
Patients and methods Observational prospective study, conducted in 2 of the 8 centers participating in a randomized, controlled, doubleblind, multicenter trial (PRETTINEO study, ClinicalTrial.gov identifier NCT01490580). Patients were randomly assigned (1:1) between "atropine-propofol" and "atropine-atracurium-sufentanil" before elective or semi-urgent intubation in the neonatal intensive care unit. Randomization was stratified on weight and center. Exclusion criteria included low blood pressure defined as a MABP (in mmHG) below gestational age (in weeks). Physiological parameters, including pulse oxymetry (SpO 2 ) and regional cerebral oxygen saturation (rScO 2 ), were collected 1 min prior to induction of anesthesia, and then up to 60 min after. To investigate the balance between oxygen delivery and consumption, cerebral fractional tissue oxygen extraction (FTOE) was calculated as FTOE = (SpO 2 − rScO 2 )/SpO 2 . The analyzed parameters included changes in rScO 2  Only one studied propofol before intubation and suggested that low rScO 2 could be attributable to low SpO 2 during the procedure, but not to low MABP (Smits et al., J Pediatr, 2016). None of them compared propofol to another anesthetic protocol. Our study is thus the first to compare cerebral oxygenation between two currently acceptable regimens of anesthesia for premedication before neonatal intubation. Conclusion This observational study is expected to provide useful information about cerebral oxygenation in neonates during intubation and to disentangle the effects of the drugs from the effects of the procedure. We also expect this study will en courage Neonatologists to avoid awake intubation.
Introduction Cellulitis (CEL) and Necrotizing Fasciitis (NF) are lifethreatening skin infections much more frequent in adults than in children. Therefore, despite major differences in epidemiology, microbiology, and outcome, clinical guidelines for management are directly adapted from adult ICU experience. We designed a prospective multicentric cohort study to describe the clinical course of CEL/NF requiring admission to PICU. Discussion This study reports the largest prospective multicentric international cohort of pediatric skin infections needing PICU admission to date. Surgery was the first risk factor reported in our cohort, before varicella. Severe circulatory and respiratory failures on admission in a context of deep biological inflammatory syndrome required aggressive treatments during the first 3 days. Lower part of the body was more frequently involved than reported before in childhood. The diagnosis was quickly suspected and confirmed mainly by ultrasound and CT scan, with the use of MRI still very rare. More than a half of patients have been operated after classical treatments associating immunoglobulins and antitoxin-antibiotics. Infection appeared mainly monobacterial as usually described in children but gram-negative strains are emerging. Mortality was still low and sequella were rare. Conclusion Cellulitis (CEL) and Necrotizing Fasciitis (NF) are still rare in childhood, with low mortality and few sequella if aggressive treatment including surgery is emergently performed. However, it seems that epidemiology has changed during the last decade in high-income countries. Statistical analysis of this database is ongoing to identify predictors of surgical requirements and to describe more precisely bacterial strains involved. Ann. Intensive Care 2017, 7(Suppl 1):8

Annals of Intensive Care 2017, 7(Suppl 1):O28
Introduction Immunosuppression induced by sepsis is well described in adults. Therapeutic trials with immunomodulatory treatments are already underway, using HLA-DR expression on monocytes (mHLA-DR) or lymphopenia as biomarkers of immunosuppression. Pediatric patients with Septic Shock (SS) have been much less studied. Thus, the main objective of this study was to explore post-sepsis immunoparalysis in a pediatric cohort study. Both sides of cellular immunity were assessed: innate immunity (mHLA-DR) and adaptive immunity (lymphocyte subsets). We also wanted to obtain normal values of mHLA-DR in healthy children, according to age. Conclusion For the first time, we reported that healthy children presented mHLA-DR values with no variation according to age, and similar to those seen in adults. Innate immunity was similarly altered in children with Septic Shock than in adult patients. Adaptive immunity seemed to be less impaired by the shock than in adults. Further analyses will be conducted to complete these findings as functional tests on T cells. As secondary infections considerably increase the length of stay at the hospital, this report paves the way for a multicenter therapeutic trial, in order to evaluate immunostimulatory treatment in children with Septic Shock. Conclusion We managed one of the larger neonatal study for VAP in critically ill neonates. These pneumonias remained frequent in critically ill neonate and are associated with a higher mortality and morbidity. Patients with a birth weight less than 1000 g seem to be particularly vulnerable.

Competing interests
None.

Competing interests
None.

O32
Driving pressure is a significant predictor of mortality in the acurasys and proseva randomized controlled trials in ARDS patients Claude Guérin 1 , Laurent Papazian 2 , Reignier Jean 3 , Louis Ayzac 4 , Anderson Loundou 5 , Jean-Marie Forel 6 Introduction Driving pressure (ΔP) across the respiratory system has been suggested as the strongest predictor of hospital mortality in ARDS patients. We wonder whether this result may be due to the wide range of tidal volume (VT) and PEEP used across the trials included and whether a strict control of them would minimize the role of ΔP as predictor. Therefore, we investigated ΔP in two trials in which lung protective mechanical ventilation was applied to ARDS patients. Our working hypothesis was that ΔP was a risk factor just like compliance (Crs) or Plateau pressure (Pplat) of the respiratory system. Patients and methods ARDS patients included in the Acurasys and Proseva trials previously reported were used. Both had near inclusion criteria (notably PaO 2 /FIO 2 < 150 mm Hg and PEEP ≥ 5 cm H 2 O) and similar lung protective mechanical ventilation (in particular VT 6 ml/ kg predicted body weight and PEEP/FIO 2 table). Both found survival benefit in the experimental group. SOFA, continuous neuromuscular blocking agent (NMBA) infusion, prone position, combined use of NMBA and prone position, pH, PaCO 2 , PaO 2 /FIO 2 , lactate, breathing frequency, VT, PEEP, Pplat, Crs and ΔP were recorded at day 1 after inclusion together with gender, age and SAPSII at the time of admission and compared between survivors and nonsurvivors at day 90. Cox proportional hazard models were used with covariates significantly different between survivors and non survivors at the threshold of 0.20 and mortality at day 90 as dependent variable. Due to the obvious colinearity between ΔP, Crs and Pplat we performed the following analyses. First we made a specific Cox model for each of them. Second, we developed three Cox models in which we used the above variables by couples (Pplat-ΔP; Crs-ΔP and Crs-Pplat). We made the following assumptions: if both variables in the couple lacked significance in the second model, the same information was carried by each component of the couple; if one or both variables kept significant correlation each brought significant and distinct information; if significant correlation was kept for one of the variables in the couple and lost for the other the former would be more informative than the latter. Results Both trials enrolled 805 patients of who 787 had data available at day 1 of who 533 survived and 254 did not. In the univariate analysis, ΔP averaged 13.7 ± 3.7 and 12.8 ± 3.7 cmH 2 O (P = 0.002) in nonsurvivors and survivors, respectively. In each single Cox model, Hazard ratios (HR) were 1.05 (1.02-1.08) (P = 0.005), 1.04 (1.01-1.08) and 0.985 (0.972-0.999) (P = 0.023) for ΔP, Pplat and Crs, respectively. PEEP and VT were not significant risk factors in any model. In the model with ΔP and Pplat used together, each of them kept significance [HR 1.31 (1.07-1.61) and 1.13 (1.02-1.26)]. In the model with ΔP and Crs both lost significance and the same was true in the model using Pplat and Crs. Discussion ΔP was a significant predictor with a 5% increase in mortality per each cmH 2 O increment of ΔP, a result similar to that found by Amato et al. However, Pplat conveys the same information. This is likely due to the fact that VT and PEEP were similar in each group of both trials and not different between survivors and non survivors. Conclusion When strict lung protective mechanical ventilation is applied to ARDS patients ΔP, Crs and Pplat were risk factors of mortality. ΔP and Pplat bring distinct and significant information.

Competing interests
None.

Impact of PEEP and body inclination in the supine and prone positions on esophageal pressure in ARDS patients
Mehdi Mezidi 1 , Mylène Aublanc 1 , Sophie Perinel-Ragey 1 , Floriane Lissonde 1 , Aurore Louf-Durier 1 , Romain Tapponnier 1 , Hodane Yonis 1 , Zakaria Riad 1 , Jean-Christophe Richard 1 , Claude Guérin 1 Introduction Prone positioning (PP) for long sessions in association with use of low tidal volume and cisatracurium infusion have been shown survival benefit in ARDS patients. Little is known about evaluation of PEEP setting based on end-expiratory transpulmonary pressure at zero flow (PEEPtotL) in PP. We hypothesized that esophageal pressure (Pes) was lower in PP than in supine position due the relief of the weight of the mediastinum. Hence, the PEEP level to achieve positive PEEPtotL should be lower in PP. Furthermore, the inclination of the body should also influence the measurements of Pes. The main objective of the study was to evaluate the variation of endexpiratory esophageal pressure at zero flow (PEEPtotes) in PP and in supine position, each at two body inclinations, and the subsequent effects on PEEPtotL and transpulmonary driving pressure (ΔPL). Materials and methods A prospective interventional physiologic study was performed in patients with severe ARDS (PaO 2 /FIO 2 < 150) and requiring PP. Pes was measured with an esophageal balloon. Transpulmonary pressure (Ptp) was computed as the difference between airway pressure and Pes. DPL was computed as the difference in Ptp at the end of inspiration and at the end of expiration at zero flow. Chest wall elastance (Est,cw) was calculated as the ratio of end-inspiratory Pes minus PEEPtotes divided by tidal volume. Endexpiratory lung volume (EELV) was measured by nitrogen wash-out method. Thorax angulation was 30° and 0° in supine position (SP) and 0° and 15° in PP. From PEEP set according to the low PEEP/FIO 2 table of the ARMA trial (1), PEEP level was further adjusted to achieve PEEPtotL close to 3 cmH 2 O. Measurements were done in supine position and after 1 h in PP. Results Twenty patients were included. PEEPtotes did not vary significantly between SP(30°) and PP(0° or 15°). However, opposite variations were found according to thorax angulation in PP: a rise of 1.9 (SD 1.7) cmH 2 O in PP(0°), p = 0.005 and a drop in PP(15°) of 2.2 (SD 3.2) cmH 2 O, p = 0.06. In a complementary analysis, PEEPtotes was studied in four positions [SP(30°), SP(0°), PP(0°) and PP(15°)] and followed an inversed U-shape pattern with mean values of 8.6 (SD 3.2), 12.8 (SD 2.7), 10.2 (SD 3.3) and 2.6 (SD 2.1) cmH 2 O, respectively. These differences were statistically significant (Holm adjusted p value for multiple comparisons <0.05) except between SP(30°) and PP(15°).
As a consequence, PEEPtotL rose of 2.6 cmH 2 O between SP(0°) and PP(0°). With postural variations, EELV was significantly altered. Hence, PEEPtotL at PP(0°) was computed at the EELV in SP(0°), allowing to eliminate the impact of the change of EELV between postures: 77% of PEEPtotL changes were due to change of posture per se [+3.7 (SD 2.6) cmH 2 O, p < 0.05]. Between SP(30°) and PP(0°), posture effect on PEEPtotL variation was +2.7 (SD 4.5) cmH 2 O, p = 0.055. A linear mixed model disclosed that rise of PEEPtotL was explained by PP for an amount of 2.7 (SD 0.44) cmH 2 O and by thorax angulation for 0.13 (SD 0.02) cmH 2 O per degree (p < 10 −6 ). PP was associated with lower ΔPL [−1.0 cmH 2 0 (SD 1.3), P = 0.004]. Chest wall elastance did not change between 30° supine and 0°PP [10.4 (SD 3) vs. 8.5 (SD 3.2) L/cmH 2 O, P = 0.85]. Discussion Variation of thorax angulation significantly alters Pes and Ptp values. Splitting the variation of Pes into those due to change in either EELV or posture allows a better understanding of the impact of PP on Pes. Conclusion PP(0°) was associated with Pes rise as compared with SP(30°) and Pes drop when compared with SP(0°). However, when analyzing PEEPtotL variation due only to posture change (and not EELV change), a rise of almost 3 cmH 2 O of PEEPtotL was seen in PP(0°) compared to SP(0°) or SP(30°). Clinicians might therefore consider lowering PEEP level in PP.

Competing interests
None.  Table 6. Conclusion Cases of severely ill suspected influenza infections in epidemic period may benefit as soon as they are admitted in ICU from antiviral therapy. Our results add to the existing observational datas on the effectiveness of starting antiviral treatment with Neuraminidase Inhibitors as soon as Influenza is clinically suspected, even if suspicion appears at ICU admission and with delay from symptoms onset.

Competing interests
None.
Introduction Non invasive ventilation (NIV) is a widely used technique to treat hypercapnic acute respiratory failure (ARF) in Chronic Obstructive Pulmonary Disease (COPD) patients. Nebulization of beta 2 agonists is recommended to treat COPD exacerbations. Nebulizing beta 2 agonist on NIV circuits could be potentially helpful but had been studied only in stable COPD patients. Our aim was to compare effectiveness of salbutamol and placebo nebulized in NIV circuit in patients admitted to the intensive care unit (ICU) for exacerbation of COPD. Aerosols were generated by a vibrating mesh nebulizer, whose optimal place in the circuit (just after the Y-piece) was determined by a preliminary bench study.

Patients and methods
We conducted a double-blinded trial comparing salbutamol and placebo delivered in a random order as two aerosols separated by 60 min, during NIV, in 43 patients admitted to the ICU for ARF. A spirometry was obtained at several points before and after nebulizations, as were clinical and biological safety parameters. Results Forced expiratory volume in one second (FEV1) increased significantly from baseline to 40 min after the end of salbutamol nebulization, when compared to placebo (+5.6 vs. −8.9%, p = 0.04).
At the same time, potassium level decreased slightly in the placebo arm (p < 0.01), when glucose and lactate levels increased (NS). Symmetric changes were observed in the salbutamol arm after the second nebulization (+13.7 vs. −8.9%, p < 0.01). Nebulization was well tolerated in all patients. Conclusion Placing the nebulizer immediately after the Y-piece is appropriate to deliver aerosols in NIV circuit and compatible with crimped circuit. Salbutamol nebulization is well tolerated during NIV and increase significantly FEV1 when compared to placebo. If this physiological effect is associated with clinical benefit remains to demonstrate.
Competing interests I, the undersigned, certify that the BANNISTER clinical trial «Beta Agonist Nebulization in Non Invasively ventilated Chronic Obstructive Pulmonary Disease (COPD) patients: Safety and Therapeutic Efficacy Range» , was funded by Association pour la Promotion à Tours de la Réanimation Médicale and the bench study by Inserm. The nebulizers were provided by Aerogen; the company was not involved in study design, data exploitation or drafting of the manuscript.
Introduction One of major determinants of non-invasive ventilation (NIV) success is its tolerance. Numerous strategies have been assessed to improve this tolerance, such as pharmacological sedation or sophrology. Music therapy has been showed to be effective in healthcare settings, in particular in ICU invasively ventilated subjects [1]. We therefore investigated the effect of a musical session provided by a trained caregiver on NIV tolerance and efficacy in ICU patients with acute respiratory failure. Patients and methods MUS-IRA is a randomized 3-centers, 3-arm open-label trial (PHRIP 2013). Subjects included were adult acute respiratory failure patients for whom the physician in charge considers NIV as indicated with a level of consciousness allowing a benefit from the musical intervention (Glasgow Coma Scale >11). Non inclusion criteria were contra-indication to NIV; severe hearing impairment; withdrawal of life sustaining therapies with expected survival of less than 48 h; subjects included in another trial dealing with acute respiratory failure. NIV was conducted in the same fashion in the three arms of randomization. The "musical intervention" subjects received a 30-min L-type music session with the MUSIC CARE© software (2), with a sleeping mask concealing the eyes. The "sensory deprivation" group had the sleeping mask and the insulating around-ear headphone, during a 30-min period. The "NIV alone" group had their NIV conducted as it is usually in our ICUs. The main objective was to determine if a musical intervention improved NIV tolerance, measured by respiratory comfort, and ventilation parameters at 30 min of NIV in comparison to conventional care. The respiratory comfort was assessed by a nurse or nurse assistant blinded to the treatment arm with a numeric visual scale (from 0 to 10) at the initiation (T0), after 30 min (T30), and at different time points of each NIV session. The primary endpoint was the change in respiratory comfort at initiation and after 30 min of the first NIV session. The secondary endpoints were NIV failure and the percentage of patients requiring anxiolytics or sedative during NIV sessions. The comparison performed in a pre-specified fashion was between "musical intervention" group and "NIV-only" group. A total of 99 subjects had to be randomized in order to show a 2-unit difference in respiratory comfort between two groups. This number was extended to 114 because of missing data on the primary endpoint.
Results Among the 114 subjects randomized (May 2015 to May 2016-"musical intervention" group n = 37; "sensory deprivation" group n = 38, "NIV-only" group n = 39), median age was 67 years (60-74), and 63 were men (55.3%). Mean baseline respiratory comfort for "musical intervention", "sensory deprivation" and "NIV-only" was 4.34 ± 3.01, 4.24 ± 2.59, 3.89 ± 3.03 respectively for the first NIV session (p = 0.74). Mean change in respiratory comfort between T0 and T30 for "musical intervention", "sensory deprivation" and "NIV-only" was respectively 0.54 ± 3.57; 0.55 ± 2.33; 0.66 ± 2.9. The comparison between "musical intervention" and NIV-alone group yielded a p-value of 0.91. NIV failure during ICU stay was evidenced in 16.2%, 10.5%, 12.8% ("musical intervention", "sensory deprivation" and "NIV-only", respectively, p = 0.74). No subjects required the administration of anxiolytics or sedatives to cope with the first NIV session in all groups. Discussion Our study failed to evidence a significant effect of a musical intervention on the reduction of respiratory discomfort during the first NIV session of acute respiratory failure ICU patients. One may hypothesize that the improvement provided by NIV (which patients received in all three groups) was such that it outweighed and/or masked a potential effect of the musical intervention. Further analysis are planned to investigate the respiratory comfort and the evolution of physiological parameters during the subsequent sessions. Conclusion Applying a musical intervention early in the course of a treatment by NIV during acute respiratory failure did not modify respiratory comfort. Further research is warranted.
Introduction Tracheal intubation in the intensive care unit (ICU) is associated with severe life-threatening complications including severe hypoxemia [1]. Preoxygenation before intubation has been recommended in order to decrease such complications. Noninvasive ventilation (NIV)-assisted preoxygenation allows increased oxygen saturation during intubation procedure [2]. However, the NIV-mask has to be taken off after preoxygenation to allow the passage of the tube through the mouth. High-flow nasal cannula oxygen (HFNC) has a potential of apneic oxygenation during the apnea period following the preoxygenation with NIV. We hypothesized that application of HFNC combined with NIV was more effective at reducing oxygen desaturation during the intubation procedure compared to NIV alone for preoxygenation in hypoxemic ICU patients with acute respiratory failure.

Patients and methods
We did this randomised, controlled, singlecenter trial with assessor-blinded outcome assessment in patients admitted to ICU. Hypoxaemic patients requiring orotracheal intubation for respiratory failure were randomised to receive preoxygenation using HFNC [flow = 60 L/min, fraction of inspired-oxygen (FIO 2 ) = 100%] combined with NIV (pressure support = 10 cmH 2 O, positive end-expiratory pressure = 5 cmH 2 O, FiO 2 = 100%) in the interventional-group or using NIV alone in the reference-group (Fig. 6). The primary outcome was the minimal oxygen saturation (SpO 2 ) during the intubation procedure. Secondary outcomes were intubation-related complications and ICU-mortality. After exclusion from analysis of two patients for protocol violation, no (0%) patients in the interventional-group and five (21%) patients in the reference-group had a SpO 2 below 80% (p = 0.050). We recorded no significant difference in intubation-related complications and ICUmortality between groups. Conclusion A novel strategy of preoxygenation adding apnoeic highflow oxygen to NIV to perform orotracheal intubation in hypoxaemic patients is more effective at reducing oxygen desaturation in comparison to the reference-method using NIV alone.
Introduction Dyspnea is common and severe in intensive care unit (ICU) patients. In these patients, dyspnea is associated with a poorer outcome. The measurement of dyspnea involves a self-assessment by the patient with a visual analogic scale (VAS-Dyspnea), which by definition requires a certain level of communication. However, many intubated critically-ill patients are unable to reach this level of communication, which makes very difficult the assessment of dyspnea.
Recently, our team has developed, a behavioral score named IC-RDOS [1], which is reliable to predict severe dyspnea in ICU patients. This score has been only validated in non-intubated conscious ICU patients. The objective of our study was to validate this score in intubated communicating and non-communicating (NC) patients. IC-RDOS was compared to VAS-Dyspnea in communicating patients and to electrophysiologic markers of dyspnea such as surface electromyographic (EMG) activity of extra diaphragmatic inspiratory muscles [2], in NC patients.

Materials and methods
Between February and August 2016, invasively mechanically ventilated patients with clinical respiratory discomfort were included. The IC-RDOS and EMG activity of the Parasternal and Alae Nasi were measured before (baseline, BL) and after one or two interventions aiming at reducing dyspnea. These interventions consisted in an optimization of ventilator settings (Opt) followed, if necessary, by the injection of opioids. The VAS-Dyspnea was only recorded in communicating patients. Non-communicating state was defined by the impossibility to perform a VAS-Dyspnea. This work was supported by the "Fond de dotation Recherche en Santé Respiratoire" and the "Fondation du Souffle". Introduction Pleural effusions are common in critically ill patients. Bedside lung ultrasonography (LU) outperformes chest radiograph for the diagnosis of pleural effusion. Pleural effusion drainage improves oxygenation and respiratory mechanics of patients under mechanical ventilation. However, the role of pleural effusion during the weaning process is unclear.

Patients and methods
We performed a prospective multicenter study in six ICUs in France. We used ultrasonography to screen for pleural effusion during the weaning process. Weaning failure was defined as failure of spontaneous breathing trial (SBT) or extubation. Patients were included the day of a first weaning test after at least 24 h of mechanical ventilation. We assessed factors associated with pleural effusion and its evolution.
Results Two hundred and forty nine patients were included. Median duration of mechanical ventilation was 4 [2][3][4][5][6][7][8] days before a first SBT. Forty-seven patients (19%) failed a first SBT and 31 (15%) had extubation failure (reintubation or death within the 7 days following extubation). Pleural effusion was detected in 139 patients (56%) the day of SBT. Most of pleural effusions were homogeneously anechoic (98%) and associated with pulmonary condensation or atelectasis (70%). Interpleural distance was higher among patients with left ventricular diastolic dysfunction, cancer history, ARDS or dialysis before the first SBT. A higher interpleural distance was associated with SBT failure and weaning failure. Patient with weaning failure had more often large pleural effusion (34 vs 12%, p < 0.01). Among patient failing the first SBT and followed up during the weaning process, either diuretic nor fluid balance change altered the interpleural distance within 24-48 h. Conclusion Pleural effusion is frequent during the weaning process and is associated with weaning failure. Further studies are needed to test whether a strategy aimed at draining large pleural effusions has the potential to decrease duration of weaning process.

Competing interests
None. The collaboration between these two entities remains poorly described. Therefore, we aimed to characterize those recourses. Patients and methods This is prospective observational multicenter study conducted during 1 month (march 2016) in the ED and ICU of three hospitals of Paris suburb: Hôpital Antoine Béclère (398 beds university hospital), Hôpital André Mignot (702 beds) and Centre hospitalier intercommunal de Meulan les Mureaux (517 beds). We first conducted a preliminary survey describing the baseline evaluation of those recourses, with a questionnaire distributed to all physicians in the participating units. Therefater, we did a prospective work to characterize the reality of theses recourses. A questionnaire was fulfilled at each call, by the ED and the ICU physicians, blinded from his colleague. Finally, to evaluate the "non recourse" too, we recorded the characteristics of ICU patients admitted from another ward of the hospital, within seven days after triage in the ED. Results Preliminary survey: 40 emergency physicians and 24 intensivists answered (response rate: 87%). Intensivists declared that generally they estimate that the recourse is justified (87% of answers), in an appropriate timing (72%), and that they generally agree with the ED's physician about the emergency level (92%).

O41
-Recourse characterization: during the study period, there were 111 recourses among the 12 674 patients admitted in the ED. We had bilateral data in 50% of cases. The mean recourse frequency was 1.2/ day, representing 1/114 patients admitted in the ED, with a great variation between hospitals: from 1/71 to 1/158 patients. 50% of these recourses occurred during daytime (8 AM to 6 PM).
Intensivist went to the ED in 81% of cases. For the remaining cases, a direct admission was decided in 40% of cases and a phone advice was given in 60%. The mean time of arrival to the ED was 10 min and the mean time spent in the ED was 32 min. Life sustaining treatment limitation was a frequent recourse's reason (9.4%). There was a non significant trend toward an estimation of higher severity by emergency physicians (37.3 vs 19.8%, p = 0.19). The exchange's climate was rated excellent or good in 81% of cases. It was more often "neutral" when an ED's resident was calling (45 vs 5.7%, p = 0.048). In those cases (8% of recourses) intensivists complained about an insufficient supervision of the resident. Main demand of the ED physician was an ICU admission (80%), and disagreement occurred in 36% of cases. Eventually, 55% of patients were admitted in ICU. "Non recourse" involved 20 patients. It was estimated according to their file that 7 of them (40%) harbored severity signs at the initial ED evaluation, but the intensivist wasn't solicited. ICU admission was denied for one patient despite intensivist's evaluation and evidence of severity.
Discussion The ICU physicians seem to be solicited by the ED physicians roughly once a day. There is some "non-recourse" with delayed admission in ICU which could be deleterious for the patient. The climate of medical exchanges is usually good, although this appreciation may be downgraded by a resident's call.
Except for the Glasgow coma scale (which was suggested in our questionnaire), no severity or prognostic score was used as objective assessment of severity. Life sustaining treatment limitation is a frequent reason for ICU recourse. These complex situations require a collegial discussion and recourses were estimated justified by the intensivists.
Conclusion Recourses to ICU physicians by ED physicians seem to be adapted and the medical exchanges happen in good relationship conditions. The frequency of these recourses is very variable among hospitals. ED physicians are mainly asking for an admission in ICU. Our work seems to point out some improvement ways, such as the use of severity scores, and a better supervision of the residents when asking for an intensivist's evaluation.
Introduction If French activity in emergency departments is similar to that of other countries, the concept of pre-hospital medicalization remains specific. Invasive mechanical ventilation (IMV) is a common practice in French pre-hospitals; however, few data is available. Materials and methods An epidemiological study was conducted at two French pre-hospital departments throughout 2015 and 5 months of 2016. Eligible patients were those receiving IMV at pre-hospitals and aged >18 years. The aim was to define pre-hospital IMV practices and determine their consequences on mechanical and blood gases. Results Eighty patients were included with 71.3% being men and middle-aged 55 years. J28 survival was 60%. The aetiologies of intubation were 38.8% traumatic, 16.5% of cardiac origin, and 12.5% of brain vascular origin. The preferred mode of ventilation was assist-control volume-control (98.75%). The median tidal volume was 6.9 ml/kg of predicted body weight (PBW; interquartile range (IQR) 6.4-7.7 ml/kg PBW) with a range of 5.1-10.4 ml/kg PBW. Lung protective volume ventilation was employed in 70.7% of patients (6-8 ml/kg PBW). The median PEEP was 5 mmHg (IQR 5-5 mmHg). 34% of patients had a change of PEEP's level compared to factory settings. The average respiratory rate was 15 cycles/min (SD 14-16 cycles/min). The median FiO 2 was 100% (IQR 60-100%), with 58% of patients exposed to FiO 2 = 100%. In all, 56% of patients had hyperoxemia gases (PaO 2 : 126 mmHg, IQR 96-307 mmHg) and 56% hypercapnia gases (PaCO 2 : 46.3 mmHg, IQR 40.1-52.9 mmHg). Overall, 46.3% of patients needing brain protection (brain injury, stroke) had hypercapnia gases. Discussion Lung-protective ventilation is not defined strictly by tidal volume between 6-8 ml/kg PBW, but is a strategy that includes PEEP, respiratory frequency, and FiO 2 (1). This study suggests infrequent use of PEEP titration in favour of delivery of a high level of oxygen and infrequent utilisation of lungprotective ventilation in French pre-hospital medicalization.
ARDS is a syndrome that should be prevented by generalization of a lung-protective ventilation strategy for all of patients (2). Conclusion French pre-hospital invasive mechanical ventilation must be optimized with a full strategy of lung-protective ventilation with regards to tidal volume, pressure, oxygenation or capnia. A comparative study following a guided ventilation strategy is currently underway to determine if a ventilatory protocol would produce better results than the free will of the emergency physician. Therapeutic interventions were volume resuscitation in 13 cases, optimization of sedation in 12 cases, vasopressor managment in 12 patients and cardiopulmonary resuscitation in 3 cases. The occurrence of complications during transport was significantly increased in mechanically ventilated patients (p = 0.009), especially with inspiratory oxygen fraction >0.5 (p = 0.00), sedation before transport (p = 0.001), vasopressor requirement before transport (p = 0.03) and with the night shift team (p = 0.007). Sedation and mechanical ventilation were the independent risk factors of IHT complications. Conclusion This study confirms that the intrahospital transport of compromised patients leads to a significant number of complications. This finding emphasises the need of improving medical skills during IHT.

Competing interests
None. The aim of this study was to describe clinical characteristics and outcomes of patients who experience medical emergency team calls and to evaluate initial care provided in medical specialty wards. Patients and methods It's a two-months prospective study (03/2016-04/2016) including all in hospital emergency calls. These calls were given by medical wards teams (dermatology, Gastroenterology, internal medicine, dermatology, cardiology and radiology) to the emergency physician when a deterioration in their patient's condition was documented. Emergency and intensive care department emergency calls were excluded. Results There were a total of 51 calls for 40 patients, principally hospitalized in internal medicine ward (21) and cardiology ward (14). Emergency calls were given by a physician in 47 cases and a nurse in 4 cases, during the night in 32 cases (62%). The greater number of calls was received Monday (13) and the week-end (11). Cardiac arrest (7), respiratory distress (28) and hemodynamic instability (16) were the reason's call. The median emergency team arrival time was 10 min. The majority of patients (88%) required acute care for on average 35 min spent time. The vital signs were monitored and adequate therapy was initiated by the attending physician for 17 (33%) patients. Cardiovascular support was necessary for 24 (47%) patients, respiratory assistance was necessary for 3 patients. A transfer for the intensive care unit was indicated for 16 patients (31%) and delayed for 13 patients because of a lack of beds. In hospital mortality was 23% (n = 12). In hospital cardiac arrest and coma (Glasgow coma scale ≤8) were the principal prognostic factors. Delayed emergency time arrival was not associated with a greater mortality. Conclusion In hospital emergencies were associated with an increased mortality. A better recognition of the instable patients may improve their prognostic.

Competing interests
None. Introduction Legionnaires' disease (LD) is an important cause of community-acquired pneumonia. A high mortality rate has been reported in the most severe LD cases, for which the optimal antimicrobial regimen remains controversial. Patients and methods This is a retrospective, multicentre observational study including patients admitted to ten intensive care units (ICU) for severe LD over a 10-year period. Patients were stratified according to the antibiotic strategy administered: (1) fluoroquinolones versus no fluoroquinolones and (2) monotherapy versus Results 211 patients with LD requiring ICU admission were included. A fluoroquinolone-based and a combination therapy were administered to 159 and 123 patients, respectively. 146 patients (69%) developed an acute respiratory distress syndrome, 111 (53%) a shock, 56 (27%) an acute renal failure that required renal replacement therapy, and 54 (26%) died in the ICU. In-ICU mortality was lower in the fluoroquinolone group as compared to the no-fluoroquinolone group (21 vs. 39%, p = 0.01), as well as in the combination therapy group as compared to the monotherapy group (20 vs. 34%, p = 0.02). In multivariable analysis with covariate adjustment including a propensity score for fluoroquinolone treatment, a fluoroquinolone-based therapy, but not a combination therapy, was independently associated with a reduced risk of mortality (HR 0.4, 95% CI [0.2-0.7], p = 0.002) ( Table 7). Discussion Our results, which suggest a beneficial effect on mortality of a fluoroquinolone-based therapy in patients with LD who required ICU admission, are consistent with previous studies showing a nonsignificant trend in favor of fluoroquinolone use in patients hospitalized for LD [1]. The limited number of patients in our study precluded assessing the individual effects of distinct molecules within the fluoroquinolone class. Conclusion Patients with severe LD receiving a fluoroquinolonebased antimicrobial regimen in the early course of the management in ICU had a lower in-ICU mortality, which persisted after adjusting on significant covariates. A combination therapy did not provide significant mortality benefit in the current study.

O46
Introduction Data on necrotizing pneumonia in ICU are scarce. This potentially devastating condition is generally thought to be due to highly virulent bacteria such as Panton-Valentine leukocidin-secreting Staphylococcus aureus or to slowly-growing organisms such as mycobacteria and other Actinomycetales. The goal of this study was to describe the clinical, microbiological and radiological features of NP in intensive care unit. Patients and methods Monocenter restrospective study. All patients hospitalized between January 1st 2009 and July 1st 2016 with an ICD-10-CM code for pneumonia (J1X) in an 18-bed medical ICU were included. All chest X-rays and CT-scans obtained during the stay were screened in order to select patients with NP, which was defined as a new cavity or a zone with no contrast enhancement within a pulmonary consolidation. Diagnosis was confirmed by a senior radiologist. Medical records of included patients were reviewed to describe clinical and radiological presentation, demographic characteristics, microbiological results and outcome. Results Among 1009 screened pts, a definitive diagnosis of NP was made in 37 pts (3.7%). Median age was 57 [interquartile range 48-69] years, 68% were males, median SAPS2 was 43. Half of the pts were immunocompromised due to hematologic malignancy (7 pts), solid tumor (7 pts), immunosuppressive therapy (2 pts) or HIV infection (2 pts). Eight pts had chronic respiratory disease, 8 had diabetes, 11 were alcoholic. Pneumonia was community-acquired in 21 pts, including 3 pts with aspiration pneumonia. The remaining NP were nosocomial diseases, including 5 pts with ventilator-associated pneumonia. In 4 pts, pneumonia was considered to be acquired by hematogenous spread, including 1 patient with endocarditis. Most pts had severe hypoxemia, with a median minimum PaO 2 /FiO 2 ratio of 153 [102-203] mmHg. Twenty-three pts were intubated, and median duration of mechanical ventilation was 4 days. Eighteen pts developed septic shock, and 3 had renal replacement therapy. In-ICU mortality was 32% (12 pts). Chest CT-scan was performed in all pts but 1. One or more cavity was observed in 33 pts, whereas lung necrosis without excavation was present in 4 pts. Most pts (25) had more than one lesion, involving more than 1 pulmonary lobe in 22 pts and both lungs in 15 pts. Median size of the largest lesion was 42 [24-78] mm. Broncho-alveolar lavage was performed in 25 pts, and bronchial aspirate in 18 pts. One or more pathogen was isolated in 29 pts; in 4 pts, the final diagnosis was non-infectious (neoplasia in 3 pts and 1 crack lung), and all samples were sterile in 4 additional pts. Fifteen pts had polymicrobial infection. Overall, the most frequently isolated pathogen was Pseudomonas aeruginosa (9 pts), followed by Staphylococcus aureus (7 pts) and fungi (7 pts). Klebsiella pneumoniae and other Enterobacteriaceae were isolated in 5 and 6 pts, respectively. All cases involving P. aeruginosa were diagnosed in immunocompromised patients. In contrast, most patients with S. aureus infection (6/7) and all patients with K. pneumoniae infection were immunocompetent. Among pts with proven infection, none underwent surgical treatment, but percutaneous drainage was performed in 3 pts.
Conclusion NP was rarely observed in our ICU. In immunocompromised patients, P. aeruginosa and fungi were the most frequently isolated pathogens, whereas highly virulent bacteria such as S. aureus and K. pneumoniae were more frequently encountered in immunocompetent patients.

Competing interests
None. Introduction Symmetric peripheral gangrene (SPG) is a rare but severe complication of septic shock often leading to multiple amputations. Epidemiology of SPG and long-term outcome remain poorly known. Amputations are serious offense to body integrity but can benefit from rehabilitation and prosthetics. Our objectives were to describe epidemiology of SPG and to assess health-related quality of life (HRQOL) once rehabilitation was achieved. Patients and methods A prospective and retrospective, multicentric study was performed. Adult patients hospitalised between 2005 and 2015 were included. They must have undergone at least two major amputations (whatever the level) and been discharged in a specialised rehabilitation center. HRQOL was assessed with generic scale EQ-5D-3L by phone call. Epidemiologic data were extracted from hospitalization reports. Results Nine centres on 13 participated, 30 patients were recruited and medical letters were available for 25 of them. SPR was observed in a majority of female (60%), aged around 50 yo. Mean intensive care unit (ICU) length of stay was 39 (± 21) days. Infectious agents were in majority Gram positive cocci (64%), but Escherichia Coli took an important part (20%), similar to Meningococcus (16%). All patients were amputated of the two lower limbs and 80% were quadruple amputees. HRQOL estimated wit EQ index was inferior to the French reference. However, patients rated themselves their health state as similar to the reference and even superior to the reference French value before SPG, using visual analogue scale (VAS). Main decrease in VAS was 22 points (95% CI 13-31). Intense pain due to phantom pain was the main factor of impaired EQ index. Painkillers use was statistically dependant of antidepressants use. All patient except one said they would be willing to be treated again for SPG. Conclusion Symmetric peripheral gangrene is mainly due to Gram positive Cocci but also Escherichia Coli. It leads to severe amputations with impaired HRQOL which could be improved by better analgesic strategies. However patients consider themselves as being in good health and would be willing to be treated again. This should be taken into account before withdrawing life-sustaining therapies.

Competing interests
None. Introduction Early surgical intervention in bacterial infective endocarditis seems to increase survival compared with conservative management or delayed surgery [1]. However, in order to decrease the bacterial inoculum on valves, European society of cardiology guidelines suggest, in cases without emergency, to postpone surgery to allow 1 or 2 weeks of antibiotic treatment [2]. We looked for valve culture in a population of endocarditis in intensive care unit and we aimed to identify associated factor for positive valve culture and to study outcomes of those patients.

Patients and methods Retrospective study between 2002 and 2016
for all consecutive adult patients who underwent cardiac surgery during the acute phase of definite left sided infectious endocarditis (Duke Criteria) in two referral centers and requiring intensive care unit (ICU) hospitalization. The following variables were recorded: demographic characteristics, echography findings before surgery (vegetation size, abscess…), cardiac timing surgery between the first adequate antimicrobial therapy and surgery and was classified as <7 days, 7-13 days and more than 14 days. Continuous variables were expressed as median and quartiles range and were compared by nonparametric tests (Mann-Whitney U-test). Proportions were compared by a χ 2 test. A descendant stepwise logistic regression analysis was performed to determine the variables independently associated with positive valve culture. The study was approved by the Rennes institutional review board (N°12-70).
Results During the study period, 165 patients underwent cardiac surgery. Among them, 13 had endocarditis without microbiological documentation and 4 had no information about valve culture. Finally, 148 patients were included in the study with a median age of 65 years [53-73], a majority of men (74%) and obesity in 23 cases (16%). In most cases, it was a native valve endocarditis (n = 100, 68%) with a predominance of methicillin susceptible Staphylococcus aureus (n = 50, 34%) and a majority of acute endocarditis began less than 1 month (n = 102, 69%). In 95 cases (64%), the vegetation size was measured >10 mm and abscess was noted in 74 cases (50% The only factor independently associated with positive valve culture was the surgery performed in the week following the beginning of adequate antimicrobial therapy (OR 8.16, p < 0.001) and there was a strong tendency for larger vegetations >10 mm (OR 2.39 but p = 0.066). Performing surgery in the second week was not significantly associated with positive valve culture. In our study, positive valve culture was not associated with death or post-surgical complications except for acute respiratory distress syndrome (ARDS) which was more frequent in case of positive valve culture (37 vs 16%, p = 0.008). Overall in hospital mortality was 23%. The mortality of patients for which surgery was performed in the first week was 16/50 (32%) versus 18/96 (18%) but this tendency was not significant (p = 0.156). Discussion Performing early surgery was independently associated with positive valve culture which may be explained by the too short time of antimicrobial therapy for reduction of bacterial inoculum. However, valve surgery in patients with positive valve culture was not associated with death or post-surgery complications except ARDS and clinicians must be aware of this risk. Conclusion Our results suggest that valve cultures are more frequently positive when surgery is performed before 7 days of adequate antimicrobial therapy. We did not find that positive valve culture was associated with worst ICU outcomes except for ARDS.

Competing interests
None. Introduction The outcome of septic shock has undoubtedly improved over the last two decades, both in immunocomptetent and immunocompromised patients. The combination of anti-infective treatments and aggressive organ failure supports often allows stabilisation of the clinical condition, but patients become then exposed to intensive care unit (ICU)-acquired infectious and noninfectious complications that significantly impact on prognosis. Whereas immunosuppression is commonly viewed as a major risk factor of death in septic shock, whether it is associated with an increased incidence of ICU-acquired complications is unclear. The present study aimed at addressing the course of septic shock across clinically significant subgroups of patients with and without immunosuppressive conditions. Patients and methods This was a 8-year (2008-2015) monocenter retrospective study performed in a 24-bed medical ICU. All consecutive adult patients diagnosed for septic shock within the first 48 h of intensive care unit (ICU) admission were included. Septic shock was defined as a microbiologically proven or clinically suspected infection, associated with acute circulatory failure requiring vasopressors despite adequate fluid filling. Patients were considered either as nonimmunocompromised or immunocompromised, this latter subgroup being categorized in three subgroups according to the underlying immunosuppressive condition: solid neoplasm, hematological malignancy and HIV-or drug-induced cellular immunosuppression. Furthermore, administration of intravenous chemotherapy during the last 3 months and leuconeutropenia defined by leucocyte count <1000/mm 3 and/or neutrophil count <500/mm 3 were collected as well. Survival status was assessed at ICU and hospital discharge. We focused on the most likely infectious, hemorrhagic and ischemic complications occurring after the first 48 h in the ICU. The impact of the underlying immune status on 3-day and late mortality was assessed using a cause-specific proportional hazard model. To assess the impact of immune status on ICU complications, we performed two competing risk analysis model (Fine-Gray and cause-specific proportional hazard model). A landmark analysis with the choice of a 3-day fixed landmark point was performed to correct for the substantial immortal time bias. The SAPS2 was used to adjust for admission severity.

Results
Eight hundred one patients were included. Their median age was 59 [57-79] years. The main source of infection was the lung in 49.6%. About 38% of patients had underlying immunosuppressive conditions, distributed into solid neoplasms (15.2%), hematological malignancy (13.2%) and non malignant immunosuppression (9.6%). The overall in-ICU and in-hospital mortality rates were 37.3% and 41.3%, respectively. With respect to the immune status, immunocompromised patients displayed worse outcome than immunocompetent ones, those with solid neoplasms having the highest risk of death. 113 patients died within the first 3 days, leaving 633 patients at risk of ICU-acquired complications. Among the 3-day survivors, the crude incidence rates of ICU-acquired infectious, ischemic and hemorrhagic complications were 27, 11 and 9%, respectively. The incidence of secondary infections was similar across the immunocompetent and the three immunocompromised subgroups of patients. In contrast, non malignant immunosuppression was associated with an increased risk of ischemic complications while hematological malignancies were associated with an increased risk of bleeding.
Conclusion We herein show that patients with septic shock display various clinical courses in relation with their underlying immune status. The burden of ICU-acquired complications in septic shock calls for every efforts of prevention and early detection, in order to improve the overall outcome.

Competing interests
None. Empirical antimicrobial therapy was effective in respectively 31 and 3 cases. Most frequent pathogens were Escherichia coli (n = 10), Enterococcus faecium n = 8, Enterobacter cloacae (n = 7), Klebsiella pneumoniae (n = 6), Pseudomonas aeruginosa (n = 3), Candida spp (n = 5). Infection was polymicrobial for n = 16 (29.6%). Involvement of small intestine including terminal ileum was associated with fungal infections (p = 0.01). ICU mortality was associated with fungal infections (p = 0.003) and polymicrobial infections (p = 0.03). Five patients underwent surgery (2 perforations, 1 digestive hemorrhage, 1 abscess, 1 negative laparotomy), with a ICU mortality of 60%. Conclusion NE is a life threatening complication in patients with malignancies who receive intensive chemotherapy. NE is associated with high mortality rates, especially in patients with fungal infections. Antifungal therapy should be systematically discussed in NE patients with small intestine involvement.

O52
Introduction Heart transplantation is the reference treatment of end-stage heart failure. National priority heart transplantation for severe acute heart failure has increased, leading to a change in the population of heart-transplant recipients (older patients with more comorbidities). Infectious complication is one of the biggest concerns after solid organ transplantation, with high rates of morbidity and mortality. The impact of early infectious complications has never been assessed in a large cohort of ICU patients. We aimed to determine the characteristics, the determinants and the impact of infectious complications after heart transplantation.

Patients and methods
We retrospectively studied all consecutive heart-transplant recipients from Bichat university hospital, between January 1st, 2011 and June 30th, 2015. All infectious complications that occurred within 6 months after transplantation were considered for analysis. The early post-operative period was defined as the first 8 days following heart transplantation. The primary endpoint was the rate of infectious complications at 6 months. We used multivariate logistic regression to identify independent risk factors for infections in the ICU. Conclusion This study confirms the high rate of early infectious complications after heart transplantation. ECMO following transplantation is an independent factor associated with bacterial infections. Bacterial infection in ICU was associated with a longer stay in ICU. The risk of death at 3 months after transplantation increased dramatically with the third episode of "not-viral infection".

Competing interests
None.

O54
Clinical course and prognosis of severe cryptococcosis in the intensive care unit: a retrospective multicenter study

Competing interests
None.  [1]. In particular, the GBS has been observed in a previous epidemic in French Polynesia [2]. Our data confirmed the association between ZIKV infection and GBS, as the annual incidence of GBS in our ICU is usually 3-5. Interestingly, several cases of encephalitis were observed, which has not been reported previously.

O55 Fungemia caused by Yarrowia lipolytica in ICU (about 55 cases)
These are preliminary data. As the outbreak is still ongoing in the Caribbean, we can expect more cases and more precise data within the next months.

O59
Closed-loop adjustment of oxygen flowrate with FreeO 2 in patients with acute coronary syndrome, comparison of two SpO 2 target and manual adjustment: a randomized controlled study Minh-Tu Huynh-Ky 1 , Pierre Alexandre Bouchard 1 , Erwan L'Her 2 , Jean-François Sarrazin 1 , François Lellouche 1 Introduction Supplemental oxygen has been used in the management of patients with acute coronary disease and investigated for more than a century [1]. The rational for a systematic oxygen supply in patients with acute coronary syndrome is based on a limited number of data. The links between hypoxia and ischemic or arrhythmic electrocardiographic abnormalities have been known for a long time. The risks associated with hyperoxia have been well described in physiological studies (increased coronary resistances, decreased coronary blood flow). A recent large RCT demonstrated that liberal oxygen administration during the acute phase of myocardial infarction may increase the level of cardiac enzyme and the infarct size [2]. FreeO 2 is a recently developed device that automatically titrates oxygen flowrate with the aim to maintain constant the SpO 2 , around the target set by the physician. We evaluated two SpO 2 targets using automated oxygen titration with FreeO 2 during acute coronary syndrome and evaluated oxygenation delivered with usual manual oxygen therapy. Our hypotheses were that high rates of hypoxemia and hyperoxia would occur with manual titration, and would be reduced with automated oxygen titration.

Patients and methods
We conducted a pilot randomized controlled, single blind monocentric study to evaluate oxygen therapy administration at the acute phase of acute coronary syndrome. Patients with acute coronary syndrome were included (based on AHA criteria). Severe COPD patients were excluded. Patients were randomized one of the three arms: Automated oxygen titration with FreeO 2 at two different SpO 2 targets (92 and 96%) and manual administration of oxygen. The study lasted a maximum of 24 h for each patient, including one night. All patients were continuously monitored with FreeO 2 (set in recording mode in the manual group), continuous cardiac telemetry was performed for all patients, cardiac enzymes were collected as per usual care. The primary outcome was the frequency of desaturation (SpO 2 < 90% for 30 s). The secondary outcomes were the frequency of arrhythmias, the rate of tachycardia episodes and the level of cardiac enzymes in patients with acute coronary disease.
Results Sixty patients were included in the study, the mean age was 63 ± 12 years, 73% of the patients were men. The average recording time was 11.5 ± 2.8 h. Preliminary data are presented here.
-21 patients were included in the control group (manual adjustment of oxygen therapy).
Conclusion Automated oxygen titration with FreeO 2 with SpO 2 target set at 97% reduced the rate of desaturation and severe desaturation in patients managed for acute coronary syndrome. Automated oxygen titration with FreeO 2 with SpO 2 target set at 97% reduced the rate of ventricular extrasystoles. SpO 2 target set at 92% with FreeO 2 may not be sufficient in this population.
With manual titration the risk of hypoxemia and severe hypoxemia was increased, however, the rate of significant hyperoxia was not increased in this study and could not be evaluated. Additional data are required to evaluate intermediate levels of oxygenation target and potential impact in this population.
Introduction During acute circulatory failure, volume expansion (VE) aims at increasing cardiac output (CO). However, CO is seldom measured to manage VE [1]. Increase in systolic, mean and pulse arterial blood pressure (ΔveSBP, ΔveMBP and ΔvePP) or decrease in heart rate (ΔveHR) are often used as surrogates for VE-induced increase in CO (ΔveCO) despite their poor performance. VE-induced changes in end-tidal carbon dioxide (ΔveEtCO 2 ) could be a surrogate for ΔveCO. Indeed, EtCO 2 , the amount of exhaled carbon dioxide (CO 2 ), depends on CO 2 production by body tissues, its delivery by CO and its elimination by alveolar ventilation. If, during VE, alveolar ventilation is kept unchanged, i.e., during fully controlled ventilation, and if we hypothesize that CO 2 production only mildly changes, then ΔveEtCO 2 may reflect ΔveCO [2].
Other appealing surrogates were also scarcely evaluated: Doppler measurements of VE-induced increase in femoral artery flow (ΔveFemFlow) or, provided the absence of inspiratory efforts and arrhythmia, VE-induced decrease in respiratory pulse pressure variation (ΔvePPV). The objective of this study was to compare ΔveEtCO 2 , ΔveFemFlow, ΔveSBP, ΔveMBP, ΔvePP and, when applicable, ΔvePPV and ΔveHR as surrogates for ΔveCO for the identification of patients who responded/ did not respond to VE. Patients and methods Adult patients were prospectively included if (1) they already had an arterial line, (2) they were receiving controlled mechanical ventilation, (3) their blood pressure (BP) was stable over 5 min, (4) the attending physician prescribed a VE and (5)  seem necessary to precise the best antibiotic management in these patients.
Introduction ICU-acquired infections are common in critically ill patients and result in prolonged ICU-stay and increased healthcare costs. Moreover, the necessary antibiotic treatment may lead to a selection of multiresistant bacteria. Therefore, health authorities mandate the registration of specific nosocomial infections such as ventilator-associated pneumonia and catheter-related blood stream infections and wound infections. Unfortunately, there is no unified definition of nosocomial infection and its incidence depends on the screening intensity. We therefore hypothesized that a computer algorithm using the prescribed antibiotic treatments may be an objective alternative for the detection of nosocomial infections. Patients and methods As a gold standard, infections were scored by an infectious disease specialist. The Java algorithm uses the following inputs: patient identifier, admission/discharge day, antibiotic name/dose/route of administration, administration days of antibiotics. The algorithm distinguishes infections on admission and nosocomial infections. Sensitivity needs to be increased while maintaining specificity for actual use in an environment with varying antibiotic prescription practices. The algorithm should also be further validated in lager datasets for different ICUs.
Conclusion In this proof-of-concept study, a computer algorithm was shown to be an acceptable alternative to trained infectious disease specialists in scoring nosocomial infections.

Competing interests
None.

O68
Temporal trends in ICU-acquired bacteremia due to Staphylococcus aureus annual incidence in a French national ICU Bone marrow smears were performed in the ICU in 77 cases and were mostly useful to contribute to HLH diagnosis (83% hemophagocytosis). However, only 23 (30%) identified HLH-etiology. Among tissues that were sampled, the highest and the lowest diagnostic yields were provided by spleen resection and liver biopsy, respectively (7/8 [87.5%] splenectomies and 11/30 [37%] liver biopsies allowed to establish a definite diagnosis). Figure 10 displays respective contributions of other tissue examination. Interestingly, the best feasibility/contribution ratio was achieved by minimally invasive lymph node biopsy. Namely, among 97 patients with clinically relevant lymph nodes, 57 (59%) could be explored under echography or CT, and diagnostic yield was 74%.
Severe adverse events included two cases of reversible hemorrhagic shocks, one following transjugular liver biopsy and one after splenectomy. No post-procedure infection was recorded. Discussion Identifying HLH-etiology is a mandatory step for short and long term recovery. Moreover, when HLH patients are critically ill, identifying and targeting HLH etiology is the cornerstone of ICU management. However, in patients with multi-organ dysfunction, clinicians need guidance to understand which procedure is likely to identify HLH-etiology. In patients with HLH, pancytopenia or spleen enlargement are the rule. However, other organ involvement has to be sought in each specific case. Even if splenectomy achieves the highest diagnostic yields, minimally invasive lymph node biopsy or skin biopsy are of value based on their feasibility/diagnostic yield ratio. Only few complications were identified in this work and minimally invasive procedures seem to be safe. Conclusion This study provides important guidance to establish HLHetiology. Timing for identifying HLH-etiology has nevertheless not been associated with mortality. Studies to assess whether targeting HLH-etiology in the initial management are warranted.

Competing interests
None.  [1]. 288 patients who were neutropenic at ICU admission were included. The main endpoint was respiratory status deterioration at day 7, defined as any increase in respiratory SOFA score or death within the first 7 days of ICU stay. Using a propensity score (PS) based on the probability of receiving G-CSF during the first 48 h of ICU stay, we estimated the association between GCSF administration and respiratory function at day 7 on the matched sample, using a logistic regression model, adjusted on respiratory SOFA score at admission. Results 288 neutropenic patients were included in the study. 201 (70%) did not receive G-CSF during the first 48 h of ICU stay. 87 (30%) received G-CSF at day 1 or day 2. 142 patients were selected by PS-matching. 57 (40%) were male, the median age was 58 (46-65) years. The most frequent malignancy was acute leukemia in 60 (42%) patients. The median SOFA score at admission was 6 (4-9). The respiratory SOFA score at admission was 0 for 103 patients (73%), 1-2 for 18 patients (12%) and 3-4 for 24 patients (15%). 11 (8%) patients had an invasive pulmonary aspergillosis. After propensity score matching (71 patients/group), there was no significant association between G-CSF administration and respiratory status deterioration during the following 7 days (OR 1.08; 95% CI 0.55-2.13; p = 0.83), even though neutropenic patients who received G-CSF had better neutrophil recovery at day 7. 70 (49%) patients died or showed an increase of the respiratory SOFA score. Among them, 37 (53%) received G-CSF and 33 (47%) did not. Additional sensitivity analysis in patients admitted for acute respiratory failure showed similar results (OR 0.96; 95% CI 0.36-2.55; p = 0.93). Discussion This is the largest study focusing on the association between G-CSF administration and respiratory status in critically ill neutropenic patients. Although the benefit of G-CSF in terms of mortality has never been demonstrated in this population, our data do not confirm G-CSF related pulmonary toxicity. Conclusion In this study, G-CSF administration was not associated with deterioration of respiratory status in critically ill neutropenic patients with haematological malignancy.

Competing interests
None. By multivariable analysis with mortality as the outcome variable, admission after day-3 tended to be associated with mortality but did not reach significance when adjusted on patient's characteristics and processes of care. An ongoing analysis to adjust on severity at ICU admission will be communicated in January.
Conclusion Patients admitted to the ICU more than 3 days after hospital admission have an increased mortality. As one-third of these patients were admitted after an initial refusal, these results identify the typology of patients for whom delayed ICU admission impacts on outcomes. Interventional studies are warranted.

Competing interests
None.

O74
The effects of passive leg raising can be detected by the plethysmographic oxygen saturation signal Alexandra Beurton 1 , Jean-Louis Teboul 1 , Valentina Girroto 1 , Galarza Laura 1 , Christian Richard 1 , Xavier Monnet 1 Introduction Volume expansion is aimed at increasing cardiac output. Nevertheless, a cardiac output monitoring technique is not always available. The perfusion index (PI) (Masimo Corp., Irvine, CA, USA), is automatically calculated from the plethysmographic waveform of oxygen saturation as the ratio of the pulsatile fraction, caused by blood flow, and the non-pulsatile fraction. We hypothesized that PI could be proportional to stroke volume and that it could detect changes in cardiac output during passive leg raising (PLR) and volume expansion (VE).

Patients and methods
We included patients for which a PLR test was planned. We measured the changes in cardiac index (CI, PiCCO2 device, Pulsion Medical Systems, Munich, Germany) and PI before and during the PLR tests. If a VE (500 mL of saline infusion over 10 min) was performed, we also measured its effects on CI and PI. Results Fifty-five PLR tests were performed in 30 patients. One case was excluded because of a poor oxygen saturation signal. Norepinephrine was administered in all cases at a mean dose of 1.6 ± 1.0 mg/h. The PLR test was positive (increase in CI ≥ 10%) in 26 "preload responsive" cases and negative in 29 "preload unresponsive" cases. During PLR test, in preload responsive cases, CI significantly increased by 30 ± 15% and PI significantly increased by 79 ± 45%. During PLR test, in preload unresponsive cases, neither CI nor PI changed significantly. PI was able to detect a positive PLR test with good accuracy (area under the receiver operating characteristic curve: 0.99 (95% confidence interval 0.90-1.00, p < 0.001). If PI increased >23%, a positive response to PLR could be diagnosed with a sensitivity of 100% (84-100%) and a specificity of 90% (74-98%). Volume expansion was administered in 15 cases with a positive PLR test. Taking into account both PLR and volume expansions, the changes in CI and PI (27 ± 11% and 73 ± 47%, respectively) were correlated (r = 0.73, p < 0.001). The PI value at baseline was <2 in 15 patients, but the ability of PI to track changes in CI was not poorer in these patients than in the other ones.

Conclusion
The results of this preliminary study are that the perfusion index using pulse oximetry seems to accurately reflect changes in CI during PLR test and volume expansion. This could be a reliable way to assess preload responsiveness in a totally non-invasive way. Conclusion In a non-selected ICU population, half of the patients required fluid resuscitation or vasoactive drugs in the 2 h following TI. Sepsis was the strongest predictor of hemodynamic instability. Tissue hypoperfusion parameters, especially mottling score, identified a subpopulation of septic patients with a high risk of hemodynamic instability. In contrast, global hemodynamic parameters such as cardiac index before TI did not predict bad hemodynamic tolerance of the procedure.

Competing interests
None.
Introduction Elevations of cardiac troponin values, suggesting a myocardial necrosis, are common in critically ill patients with sepsis and are associated with a poor prognosis. In this context, myocardial ischaemia is one of the main underlying mechanisms leading to necrosis of myocardial cells defining the myocardial infarction (MI). However identifying which septic patients with suspected myocardial infarction are at risk of mortality and major cardiac events is a clinical challenge [1]. Clinical evaluation may lack sufficient precision, leading to inappropriate medication and discharge. It is uncertain whether risk scores derived from cardiologic populations apply in this context. We aimed to assess whether the «  Conclusion These preliminary data suggest that therapeutic drug monitoring could optimize the achievement of pharmacokinetic objectives associated with an effective antibiotic therapy. These data also suggest that, in most patients, the recommended doses of imipenem at 1 g three times a day and aminoglycoside at 20 to 25 mg/ kg, do not respect the pK objectives reported in the literature.

Introduction
The control of health-care associated infections and multidrug-resistant bacteria (MDRB) is a public health priority. The MDRB prevalence rate has dramatically increased, mainly extendedspectrum beta-lactamase producing Enterobacteriaceae (ESBLE

Competing interests
None. Introduction Selective decontamination with various regimens using topical antimicrobials has been reported to reduce acquired infections (AIs) and mortality in the ICU. The efficacy of decontamination on the prevention of P. aeruginosa AIs is controversial [1] and potential impact on the emergence of antimicrobial-resistant strains is a concern. We assessed the incidence of P. aeruginosa AIs (PAAIs) with the routine use of a multiple decontamination regimen with oropharyngeal and digestive tobramycin/colistin/amphotericin B and nasal mupirocin/ chlorhexidine body wash over 5 years with a special attention to antimicrobial resistance.

Competing interests
None.   Conclusion Staphylococcus spp. is essentially responsible for low respiratory infections, septicemias and infectious endocarditis, mostly nosocomial. These infections concern severely ill patients, with a low survival rate, without any causal link proved in this study. Empiric anti-MRS antibiotic therapy must be improved, particularly in a nosocomial condition. Vancomycin is still the first anti-MRS used 10 years after the provision of linezolid in France. Easier to use, empiric prescription of linezolid for nosocomial lung disease could help to improve patients' management.

P7
No meticillin resistant staphylococcus aureus, nor carbapenemase producing pseudomonas aeruginosa or acinetobacter baumanii were isolated. Conclusion MDR bacteria (100% extended-spectrum beta-lactamaseproducing enterobacteriacae) is present in more than one-third patients transferred from the emergency department. The risk factors forMDR bacteria acquisition and a strategy to contain this phenomenon are needed in our hospital.

Competing interests
None.  [1]. Similarly, in our Introduction The convulsive crisis is a symptom indicating a hyper synchronous discharge of a more or less extensive portion of the cerebral cortex. It is a frequent disease in the emergency (0.3 to 1.2%) [1]. The aim of our study was to report the epidemiological, clinical, therapeutic and evolutionary aspects of the convulsive crisis in the emergency.

Patients and methods
Prospective study extending on year (01/01/2015 -31/12/2015), including all patients aged over 18 consulting in the emergency for convulsive crisis.

Results
We included 41 patients. The average age was 45 ± 19 years with a sex ratio of 1.5. The convulsive crisis occurred at a known epileptic in 53.6% of cases. The average duration of crisis was 10 ± 4 min. The average time of consultation was 54 ± 21 min. Generalized tonic-clonic seizures were observed in 30 patients. Partial seizures in 11 patients. The status epilepticus has been reported in 12 patients (29.2%). Discontinuation of antiepileptic therapy was noted in 20 cases (48.7%), intoxication in 6 cases (14.6%), hypoglycemia in 5 cases (12.1%), head trauma in 5 cases (12.1%), accident cerebrovascular in 3 cases (7.3%) and preeclampsia in 2 cases (4.8%). The convulsive crisis was accompanied by a post-critical deficit in 6 patients (14.6%) and a head injury in 2 patients (4.8%). A brain imaging was indicated in 18 patients (43.9%). Benzodiazepines were prescribed in 78% of cases and barbiturates in 24.3% of cases. The use of invasive mechanical ventilation was reported in 12 patients (29.2%). Thirteen patients were hospitalized in intensive care, 12 in neurology and 6 in medicine department. There were no deaths. Conclusion The management of convulsive crisis in the emergency is multidisciplinary. It is based on a good mastery of the definitions and recommendations on this subject.
Introduction Stroke is a frequent and serious disease. Stroke units (SUs) have demonstrated efficiency in reducing death and disability after stroke. In France, public health autorities's reports recommended that every patient presenting with stroke symptoms should be admitted in SUs. But beds in these units are still insufficient, so triage seems unavoidable. We aimed to explore admission criteria in SUs in case there is only one bed left. In this situation, is the time window allowed thrombolysis the main criteria for neurologists to accept a patient? Patients and methods This was a postal and e-mail survey that took place from January 1st to July 31st 2016. An anonymous questionnaire was sent to the 164 neurologists who usually work in the 21 SUs in the region of Ile de France. The survey was in two parts: the first one encompassed questions about the physicians themselves (age, professional status, current job tenure), and, then, in the second part we presented different clinical cases including the type of stroke (ischemic, hemorrhagic, or transient ischemic attack), age ranges of the patients, and different time slots during nyctohemeral period. Finally, data were collected in July 2016. Results were expressed as absolute numbers and percentages, and for statistical analyses, we used the Fisher's exact test from the on-line software BiostaTGV.
Results One hundred four questionnaires were completed, representing 63% of all. Majority of physicians were hospital practitioner (53%), and 56% have been working in SUs for more than 5 years. We asked to neurologists if they already refused a patient when it stay one bed in their unit, to keep it for a patient who could be eligible for thrombolysis, and the answer was positive in 64%. There was no significant difference in accepting or denying regardless of the type of stroke or the different time slot. Concerning the age ranges, we found a significant relationship only with ischemic stroke (p = 0.014), as the oldest patients above 90 years old were more refused than the youngest. Discussion Our study has several limitations, and probably the main is the risk of declarative biases due to descriptions of practices that differ from those encountered in real life. A more rigorous methodologic analysis, including a prospective study following patients suffering from stroke, would perhaps more reflect the reality. Conclusion In case there is one bed left in their SUs, neurologists seem to accept most patients, even if thrombolysis is off limits. Only the age ranges may influence their decision if the patient suffers from ischemic stroke. More studies are necessary to confirm or not these results.

Competing interests
None. Introduction The management of hemorrhagic stroke (AVCH), considered the second cause of death in the world, has in recent years major changes. Early diagnosis and optimization of Early and specialized load have also shown their interest on the prognosis of patients Purpose Through this analytical study, we emphasize the quality of the management of the hemorrhagic stroke.

Materials and methods
We realized a retrospective study about 50 patients hospitalized at the resuscitation department of the surgical emergencies at the UHC Casablanca during a period of 4 years from January 2012 to December 2015. It included 31 men and 19 women, the meanagewas51 ± 15 yearsold. The epidemiological, clinical, paraclinical, therapeutic, and evolutive data were collected and gathered on a card then entered and analyzed on Epi Info software.
Results 50 patients were collected whose average age is 52 plus or minus 15 years male predominance is noted. Hypertension is the most common risk factor for 41% followed by diabetes with a rate of 26%. In our study, the recruitment rate was highest in autumn period 33.33%. The mode of early neurological symptoms was brutal in 74% of 37 patients. On admission clinical examination showed impaired consciousness with a Glasgow score ≤10 33 patients or 66%. Hemiplegia is the main focal neurological signs met the most common site of cerebral hemorrhage is lobar. Brain CT showed an intra-parenchymal hemorrhage in 40 patients or 79% of cases with a predominance of the parietal location (23%). A pure subarachnoid hemorrhage was noted in 4 patients 10% of cases. 46% of patients had ventricular flooding. Mechanical ventilation was required in 37 patients or 74%. Antihypertensive therapy was prescribed in 30 patients or 59%. Antiepileptic therapy was indicated in all patients. The surgical indication was raised in 22 patients or 44%. 14 patients died after surgery. The outcome was favorable in 14 patients or 28%, and negative in 36 patients or 72% of cases. Conclusion In the literature, various predictive factors of mortality are observed. The prognosis of the hemorrhagic stroke in the units of intensive cares is still dark, hence the importance to develop specialized centres in order to improve the management of the patients who suffer from this affection.

Competing interests
None. Introduction Optic nerve sheath diameter (ONSD) is an easy to obtain parameter, which has been associated to absolute value and variation of intracranial pressure. However, the normal value of this parameter (observed in patients without raised intracranial pressure) remains debatable (from 5 to 5.9 mm) highlighting the importance measurement precision. Is this pilot observational study, we aimed to compare the ONSD values obtained with two different US device. Introduction Delayed cerebral ischemia (DCI) is the main cause of disability after subarachnoid hemorrhage (SAH) but its pathophysiology remains unclear. We currently have no effective treatment to reduce the incidence of DCI. Study of pathophysiology has long suffered from a lack of consensus definition of DCI. It is only recently that experts have proposed that in observational studies and clinical trials aiming to investigate strategies to prevent DCI, the 2 main outcome measures should be: (1) cerebral infarction identified on CT scan or magnetic resonance imaging or proven at autopsy, after exclusion of procedure-related infarctions; and (2) functional outcome [1]. Currently, experimental studies must therefore meet this new definition. The aim of our study is to identify the extent preclinical studies using this new definition of DCI. Materials and methods Systematic review from PUBMED using key words. Consulting additional bibliographic databases including gray literature, scientific meeting abstracts and review of the bibliography of selected articles completes the research. Inclusion criteria: (1) description and/or modification of an SAH model in rats or mice; (2) study of cerebral vasospasm and/or DCI. This research is performed in accordance with the PRISMA recommendations [2]. Results Seventy-one publications from 47 teams are included. 11 teams are responsible for 77% of publications. 8 different methods are used for induction of SAH. The most used model is the direct blood injection into the cisterna magna. Vasospasm is studied by brain imaging (n = 18) mainly in the first 72 h and histology (n = 32) until the tenth day. A positive diagnosis of vasospasm is placed in histology in 28 of 32 studies. Cerebral ischemia is sought in 24 publications including one in cerebral imaging. Neurological deficit is wanted by sensorimotor tests in 13 publications. Three models use animals with comorbidities (hypertension, obesity and chronic inflammation). No model uses female population. Among the 10 most cited studies, none of them is studying cerebral ischemia by imaging. Discussion Differences with human pathology of SAH are raised about the pathophysiology leading to SAH and its anatomoclinical characteristics. In particular, there is an over-representation of models of posterior circulation. Similarly, study of vasospasm remains the most studied criteria to the detriment of the ischemia and neurological deficit, even though it is the first cause of disability after SAH (Table 11). Furthermore, we can note an over-estimation of the incidence of vasospasm in the included publications compared to the clinical epidemiological data. Study methods of vasospasm and cerebral ischemia may be criticized. Results 25 cases were collected, 12% of deaths during 2015, with average age of 7, 32 ± 6, sex ratio was M/W 1.5, the average weight of our population was 34, 72 ± 32 kg, reason for admission was head injury following an PRA for 8 cases, cranial traumatism following defenestration for 2 cases, stroke for 10 cases, and epilepsy for 6 cases. The Glasgow Average on admission was 6.88 ± 3, mean hospital duration was 6.75 days. All patients were ventilated on admission, use of vasoactive drugs were needed for 20 patients. Brain death diagnosis was made by two intensive anesthetist doctors for each patient looking for brain death clinical signs. Cerebral angiography was made for 10 patients. The procedure of organ donation has been discussed in 17 cases, parents have accepted donation in two cases, refusal of both parents in 9 cases, refusal of one parent in 6 cases. In 8 cases the patient died before starting the procedure. We successfully removed both kidneys and liver for our first donor and just one kidney and cornea for the second one. Discussion In front of persisting imbalance between the needs and the number of grafts, removal of pediatric organs should be enhanced by optimizing children potential donors and their families management. It is important to encourage the involvement of a health care team and humanly participate with the coordination teams to deal with parents and dare to seek their consent to donate. Conclusion This study shows out a significant pool of potential organ donors in pediatric intensive care. Hence the need to lead out activities within our society to make of them effective donors.

Patients and methods
Introduction Interest in bedside lung ultrasound in the intensive care unit has exploded in recent years [1]. The aim of this study was to evaluate the contribution of lung ultrasound in an intensive care unit and its place as a diagnostic tool and therapeutic implication.

Patients and methods
Retrospective study extending over 6 months. It included ICU patients who underwent lung ultrasound. Results Forty patients underwent a lung ultrasound due to various respiratory etiologies: 60% patients for acute respiratory failure, 30% for ultrasound-guided pleural punctures, 20% for exploration of a white lung on chest radiograph. Ultrasonography consists in exploration the chest over the six regions. Of the 24 (60%) acute respiratory failure, the review found: 15 (62.5%) fluid pleural effusions, 2 (8%) pneumothorax, 4 (16%) alveolar-interstitial syndrome and 3 (13.5%) pulmonary condensation. Clinical and gasometric improvement was observed in 70% of patients. Lung ultrasound has allowed a change in management in 43% of patients and has provided new information in 72% of cases and led to successful ultra sound guided pleural drainage. Discussion Lung ultrasound has allowed a major improvement of routine practice in intensive care unit. Once the appropriate equipment and training acquired, lung ultrasound has advantages: Safety, speed, acuity, reduced costs and increased patient comfort Conclusion Responding immediately to questions for which only sophisticated approaches were used, lung ultrasound untangles these daily problems in the emergency and intensive care.

Introduction
The use of large tidal volume ventilation can be deleterious and lead to Ventilator Induced Lung Injury. It has been suggested that Lung-protective ventilation using low tidal volume should be used in all critically ill patients. We conducted a study to evaluate the tidal volumes related to body weight prescribed during Assist Control Mechanical Ventilation (ACV) in our Unit and to determine the proportion of patients receiving Low Tidal Volume Ventilation (LTVV).

Patients and methods
Retrospective study conducted in a single 12 bed adult intensive care unit (ICU). Patients requiring mechanical ventilation admitted from January to June 2014 were included. Medical files were reviewed to determine the minimal and maximal tidal volumes related to predicted body weight (Vt min and Vt max) delivered during Assist Control Ventilation. LTVV was defined as Vt ≤ 8 ml/kg of predicted body weight (PBW).

Results
We included 129 patients (63.5% men) with a median age of 64 years (IQR 52.5-71), a mean SAPS II Score of 46 ± 16 and a median length of mechanical ventilation of 6 days (IQR 2-14.5). Patients were predominantly non-surgical (77.5%) and 18.6% had a diagnosis of ARDS. Mean height was 168.4 ± 9.3 cm, actual and predicted body weight were 75.5 ± 21.7 kg and 63.1 ± 9.8 kg. Median Vt max was 7.9 ml/kg (IQR 7.3-8.9) and 72 patients (56%) had permanent LTVV (Vt max ≤ 8 ml/kg). Median Vt min was 7.4 ml/kg (IQR 6.9-8.2). Women had Vt max and Vt min significantly higher than men [median respectively 9.4 (IQR 8.3-9.8) vs 7.4 ml/kg (IQR 7-8) and 8.4 (IQR 7.5-9.3) vs 7.2 ml/kg (IQR 6.7-7.6); p < 0.0001] (Fig. 11). Only 17% of women had permanent LTVV versus 78% of men. Discussion Even though the benefit of LTVV applied for all ICU patients is still to be demonstrated, our results show that women are exposed to much more "aggressive" mechanical ventilation with potentially harmful consequences. LTVV seems to be achievable for all intensive care patients since 78% of men admitted in our unit are ventilated with maximal Vt ≤ 8 ml/kg of predicted body weight. Exclusion criteria were mixed lung condition, presence of chest tube, prone position, and severe hemodynamic impairment. Patients were classified according to standard definition in: normal lungs, COPD or ARDS. Patients were ventilated in INTELLIVENT-ASV ® mode and positioned in semi-recumbent. Respiratory mechanics variables were measured once per patient during the first 24 h after ICU admission together with arterial blood gases. Airway pressure and flow was measured by a proximal pneumotachograph. Static compliance (CSTAT) and inspiratory resistances (RINS) were measured using the least square fitting method [2]. CSTAT was also calculated as the ratio between tidal volume and driving pressure. Expiratory time constant (RCEXP) was measured as the ratio between volume and flow at 75% of maximum expiratory flow and calculated as the product of CSTAT MEAS and RINS. Anova was performed to compare results between each lung condition. Data are presented as median [25th-75th quartiles].  Table 12. Conclusion This study reports clinically-based values of respiratory mechanics for passive mechanically ventilated ICU patients. These results should help bench studies design to simulate realistic conditions.

Usefulness of a respiratory comfort scale in titration of the NAVA (Neurally Adjusted Ventilatory Assist) level for patients under spontaneous mechanical ventilation: A physiological study
Bertrand Canoville 1 , Cédric Daubin 1 , Jennifer Brunet 2 , Damien Du Cheyron 1 , Frédéric Lofaso 3 , Nicolas Terzi 4 Introduction Although Neurally Adjusted Ventilatory Assist (NAVA) is currently used in many intensive care units, the best method for setting the optimal level of assistance in this proportional ventilatory mode is not yet established. In order to determine a method based on aiming the best respiratory comfort of the patient, the authors studied the impact of several levels of assist in NAVA, on the respiratory behaviour and comfort, of patients undergoing weaning of the mechanical ventilation. Patients and methods This study was single-center prospective, open, randomised, of current care. Patients staying in a critical care unit, undergoing invasive mechanical ventilation for an acute respiratory failure, and recovering spontaneous ventilation, were included. For every patient, an optimal level of assistance in pressuresupport ventilation was determined and called AI 100, with which an equivalent level of assistance in NAVA called NAVA 100 was set. From this level, six other NAVA levels were determined, by increase or decrease of 25% stepwise, respectively called NAVA 25, NAVA 50, NAVA 75, NAVA 125, NAVA 150 and NAVA 175, and applied in a randomized order. For every situation, after obtaining a stability of the end-tidal carbon dioxide pressure (ETCO2), pressure, volume, flow, and electromyographic activity of the diaphragm (Edi) were recorded during 5 min. An arterial gazometry was performed, and respiratory comfort was assessed by using the dyspnea numeric scale, the Multidimensional Dyspnea Profile (MDP) by Banzett, and the Respiratory Distress Observational Scale (RDOS). All parameters were compared between ventilatory modes, by using variance analysis ANOVA. Results Ten patients were included. Between NAVA levels, no difference was observed regarding tidal volume (Vt, p = 0.143) and arterial carbon dioxide pressure (PaCO 2 , p = 0.141). The Ti/Ttotneu ratio and AUC Editot (the area under curve of Edi superior to trigger) raised significantly for NAVA 25, the lowest NAVA level, respectively p = 0.005 and p < 0.001. The dyspnea numeric scale and MDP by Banzett showed no significant difference along the NAVA levels, respectively p = 0.412 and p = 0.107. However, RDOS score was significantly higher for the low levels of assistance NAVA 25 and NAVA 50; p < 0.001. Conclusion With variation of the NAVA level, our patients showed a respiratory behaviour similar to the findings described previously in the litterature. The level of assistance did not seem to influence the patient's feeling of dyspnea, despite an obvious raising of the ventilatory demand. However, Ti/Ttotneu ratio, AUC Editot and RDOS clinical score appear as potential useful tools to detect underassistance issues in patients undergoing NAVA.
Introduction The aim of the study is to assess the usefulness of the Full Outline of Unresponsiveness (FOUR) score in predicting extubation failure in critically-ill intubated patients admitted with disturbed level of conscious in comparison with the Glasgow Coma Scale (GCS). Materials and methods All intubated critically-ill patients with disturbed level of consciousness were assessed using both the FOUR score and the GCS. The FOUR score and the GCS were compared regarding their predictive value for successful extubation at 14 days after intubation as a primary outcome measure. The 28-day mortality and the neurological outcome at 3 months were used as secondary outcome measures. Results Eighty-six patients were included. Median age was 63 [50-77] years. Sex-ratio (M/F) was 1.46. Median GCS was 7 [3-10] while median FOUR score was 8.5 [2.3-11]. A GCS ≤ 7 predicted the extubation failure at 14 days after intubation with a sensitivity of 88.5% and specificity of 68.3% (Youden index = 0.57 95% CI [0.35-.7]) whereas a FOUR score <10 predicted the same outcome with a sensitivity of 80.8% were transferred to a rehabilitation centre, 10 were transferred in acute care hospital in step-down unit, 29 were discharged directly to home. Weaning success was observed in 3 patients with NIV at admission and 24 were educated to self use of NIV. Of the 16 patients without mechanical ventilation at admission only 3 patients with tracheostomy were discharged with long-term tracheostomy.
Of the 54 patients with invasive mechanical ventilation and tracheostomy at admission: • 30 were weaned and decannulated • 7 were discharged with long-term tracheostomy without ventilation • 14 with long-term tracheostomy were only partially weaned from invasive ventilation • 1 with tracheostomy and 24/24 ventilator support.
Of the patients who were dependent on high oxygen flow (>6 l/mn) at admission 4 were weaned from high oxygen flow at discharge. Discussion The current data demonstrate that, in our centre, most of the patients with weaning failure can be successfully weaned. Despite a prolonged ICU stay before referral, weaning outcomes were similar to those reported in other countries and organizational models. Introduction Chronic Obstructive Pulmonary Disease (COPD) exacerbations are now the third leading cause of mortality all over the world and in Tunisia. There still are many controversies in their management and mainly the indications of antibiotic treatment and corticotherapy. In Tunisia and despite many valuable studies, only few studies were focused on the "Tunisian" daily practice in case of acute and severe exacerbations of COPD. Thus; we decided to perform this study. We aimed to describe the clinical characteristics of our patients, the blood gas exchange findings, the medical treatment, the ventilator support and to analyze their outcome.

Materials and methods
It was a Single-center retrospective study performed in the teaching department of emergency and intensive care medicine in Zaghouan in Tunisia conducted between 1st January 2015 and 31th December 2015. All patients admitted in our ICU for acute exacerbation of COPD were enrolled in the study. Anamnestic characteristics, primary diagnosis, mode of mechanical ventilation were noted. The results of laboratory tests, blood gas data and the outcome of patients were also collected.
Results 40 patients were included in the study. The mean age was 68 ± 11 years with a sex ratio of 4 (32 males and 8 females). The mean SAPS II score was 35 ± 10 with an APACHE II mean score of 22 ± 5. The most common cause of COPD in our patients was tobacco smoke with a mean consumption of 32 ± 10 pack-year history of smoking. The most frequent COPD condition in our patients was chronic bronchitis (n = 29; 72.5%). Six patients received home oxygen therapy. Three patients had lung emphysema and one patient had both restrictive and obstructive syndrome. At admission, the mean Glasgow coma scale score was 12 ± 3. The mean Ph was 7.21 ± 0.13 with a mean bicarbonate level of 32 ± 10 mEq/l. Lung infection was the most common cause of exacerbation (n = 26; 65%). The other recorded causes were: pulmonary edema (n = 3), pulmonary embolism (n = 2) and pleural effusion (n = 5). Twenty-two patients (55%) received systemic corticotherapy and 33 patients received antibiotic therapy. Sixteen patients (40%) did not receive bronchodilator medicines. Fifteen patients (37%) required noninvasive ventilation (NIV). Eighteen patients (45%) was intubated and required invasive mechanical ventilation. Ventilatory support was not necessary in the other seven patients (17.5%). The mean duration of hospital stay in our ICU was 111 ± 98 h. Sixteen patients died in our ICU with a mortality rate of 40%. The most frequent causes of death were septic shock (n = 5), acute respiratory distress syndrome and multiorgan failure (n = 2). Four patients were transferred and twenty patients were discharged from hospital. Only SAPS II was independently associated with poor outcome (p < 0.001). Conclusion Our patients may be classified severe COPD according to GOLD recommendations. There is a big gap between our daily practice and the international recommendations. We still have a widespread use of corticotherapy and antibiotic therapy. We need effective interventions to implement existing evidence-based guidelines into daily practice in exacerbations of COPD.
Introduction Acute exacerbation of chronic obstructive pulmonary disease (AECOPD) is linked to morbidity and mortality increase; high bicarbonate levels are linked to chronic hypercapnia and hypoventilation in this population. The consequences of bicarbonate levels change during ICU stay are not well studied. The aim of the present study is to investigate the association between bicarbonate concentration change during ICU stay for AECOPD and long-term survival. Long-term patient's status (surviving or deceased) was checked by consulting the register of civil status. Results During the study period 440 patients were consecutively admitted for hypercapnic COPD exacerbation (84.5% males, median age: 68 years, median pH at admission: 7.28 and median HCO 3− : 30.6, NIV was the first ventilation modality in 65% and survival at ICU discharge: 84.3%). Blood gas analyses (at ICU admission and discharge), and long-term vital status were available in 239 patients (survival at a median follow-up of 7 years was 31.2%). Bicarbonate levels change were inversely associated with the long-term survival (Table 14; Fig. 12). Conclusion Our study suggests that bicarbonate levels change during ICU stay for hypercapnic COPD exacerbation could be considered as prognostic factor.

Competing interests
None. Introduction Dyspnea is frequent and intense in intubated patients. It is one of the worse experiences of the intensive care unit (ICU) stay and is associated with adverse outcomes. Despite the increasing use of non invasive ventilation (NIV) as the cornerstone therapy of acute respiratory failure (ARF), little attention has been given to that threatening sensation. The objectives were: (1) To quantify the prevalence and intensity of dyspnea in patients receiving NIV for ARF, at admission and in response to NIV; (2) To examine the factors associated with dyspnea;

P28 Dyspnea in patients with acute respiratory failure requiring non invasive ventilation in the ICU: prevalence, factors and prognosis-a prospective cohort study
(3) To investigate the impact of dyspnea on NIV success or failure and on the quality of life and post-ICU burden. Patients and methods Second analysis of a prospective observational cohort study in patients who received NIV for ARF in 54 ICUs in France and Belgium, in 2010/2011. Dyspnea measurement was assessed with a modified Borg scale at admission and in response to the first NIV session. Patients with a dyspnea intensity <4 (defined as light to moderate) were compared to those with dyspnea intensity ≥4 defined (moderate to severe). Results Among the 426 patients included the median dyspnea on admission was 4 [3][4][5] and decreased to 3 [2][3][4] in response to the first NIV session (p = 0.001). Moderate to severe dyspnea in response to NIV was independently associated with severity as assessed by the SOFA (OR 1.09, p = 0.023), anxiety (OR 1.84, p = 0.009), leaks (OR 2.15, p = 0.002) and poor NIV tolerance (OR 2.01, p = 0.012). In addition, moderate to severe dyspnea was an independent risk factor of NIV failure (OR 2.24, p = 0.001). Finally, dyspnea was associated with higher length of stay and mortality (7 vs. 18%, p = 0.001) but was not associated with higher post ICU burden and quality of life. Conclusion In patients receiving NIV, dyspnea is frequent, intense and exposes patients to a higher risk of NIV failure. NIV is associated with poorer outcome.
Introduction COPD patients have often polyglobulia because of associated hypoxemia especially in patients at the stage of chronic respiratory failure. Little is known about the prevalence of anemia and its impact on prognosis in patients with severe AECOPD admitted to ICU. The aim of this study was to determine the prevalence of anemia in AECOPD patients and its impact on prognosis. Patients and methods In a prospectively collected database including consecutive patients admitted between 2007 and 2015 for AECOPD in our ICU, we retrospectively assessed hemoglobin levels at ICU admission. Anemia was defined according to WHO criteria: Hb < 13 g/dl in males; Hb < 12 g/dl in females. Continuous variables expressed as median (25-75 percentiles interquartile ranges, IQR) and compared with the Mann-Whitney test. Results The cohort included 210 patients (median age 67, median pH 7.30, 87.6% males, NIV as first ventilator mode in 86.2%). Anemia was observed in 77 of the 210 patients (36.6%). Median haemoglobin levels were at 10.7 and 14.5 g/dl, in patients with and without anemia, respectively, Table 15 summarizes characteristics and clinical outcomes in the two study groups.  In multivariate analysis anemia was not identified as independent factor associated with ICU mortality. Conclusion Anemia was observed in one-third of patients admitted to our ICU for severe AECOPD, and was not associated with ICU mortality.
Introduction Severe acute exacerbation of COPD (AECOPD) reduces quality of life and represents a major burden for health care systems. SUA is well known to increase during hypoxia and systemic inflammation. This final product of purine degradation is one of biomarkers showing promise in the prediction of outcomes in AECOPD [1]. The aim of this study was to assess the possible role of SUA and SUA/ creatinine ratio as biomarkers for the prediction of mortality and future exacerbations of patients hospitalized in intensive care unit (ICU) for severe AECOPD. In addition, SUA/creatinine ratio levels were higher in patients having extended evolution time prior to ICU admission (p = 0.05) and those having high SOFA score (p = 0.05). SUA levels were not associated with increased risk for AECOPD in the 6 months period after ICU discharge. High SUA was an independent factor associated to circulatory failure in multivariate Cox regression analysis (HR = 12.32, p = 0.037).

Patients and methods
In multivariate Cox regression analysis, high SUA/creatinine ratio was an independent predictor of 30-day mortality (HR = 15.03, p = 0.032), and it was also independently associated with acute heart failure (HR = 7.08, p = 0.046). Discussion SUA is considered as a useful biomarker in the identification of high-risk patients admitted for mild AECOPD [1][2]. The present study suggests a role of SUA in the prediction of mortality in severe AECOPD associated with heart failure or shock requiring ICU admission Conclusion High SUA was only independently associated to circulatory failure among patients with severe AECOPD. High SUA/ creatinine ratio on ICU admission was associated with acute heart failure and was an independent predictor of 30-day mortality in ICU patients with severe AECOPD.

Competing interests
None.

Fig. 13 See text for description
Introduction Central venous catheters constitute invasive devices that are important for patients in intensive care units. They expose to multiple complications such as: infectious, traumatic and thrombo embolic complications [1]. The aim of our study is to look for the incidence of thrombo embolic events and to target its predictive factors. Results 122 catheters were analysed, thrombus incidence was 24.5%. The comparison between the two groups is mentioned in Table 16. The thrombus was mobile in 83% with a mean length of 4.1 mm. Conclusion Thrombo embolic events on CVC are frequent, varying from 8% to 25% of cases and depend on the site. The Doppler helps us to diagnose it quickly and to prevent heavy complications. In our study, predictive factors were: the neoplasy (p = 0.045), parenteral feeding and the use of triple lumen catheter.

Introduction
The primary goal of volume expansion in acute circulatory failure is to increase cardiac output. Nevertheless, techniques that measure cardiac output are not available in every patient. Changes in heart rate could be used instead of cardiac output, since increases in cardiac output induce a decrease in the reflex sympathetic stimulation that may be reflected by significant decreases in heart rate. Our goal was to test whether the changes in heart rate could be used to differentiate positive and negative response to volume expansion.

Materials and methods
We measured heart rate and cardiac output before and after volume expansion (500 mL saline over 10-30 min) performed in patients with acute circulatory failure. We tested the ability of changes in heart rate to detect fluid responsiveness, defined by an increase in cardiac output by more than 15% with volume expansion.
Conclusion Except for changes of very large amplitude, decreases in heart rate cannot reliably detect a positive response of cardiac output to volume expansion. Changes in heart rate only allow the detection of large changes in cardiac output.

Competing interests
None. Introduction Mean systemic filling pressure (MSFP) defines the pressure measured in the veno-arterial circulatory system when the cardiac output is nil. Representing the upstream pressure of the venous return gradient, it is considered as a major hemodynamic parameter. Because of the difficulties of its measurement, it is not routinely used in clinical practice but some authors proposed to use heart-lung interaction under mechanical ventilation for its estimation. Our objective was to demonstrate how MSFP is modified by positive pressure ventilation. Patients and methods We conducted a bi-centric non interventional, prospective and observational study to measure the pressure encountered in the circulatory system at the time of death in criticallyill patients on an arterial catheter and a central venous catheter. We included 112 patients with previously inserted catheters and mechanical ventilation with a positive end-expiratory pressure (PEEP) at the time of death. After calibration and supine positioning, arterial and venous pressures were recorded in five conditions: at endexpiration and end-inspiration with and without PEEP and finally after disconnection from the ventilator.

Results
Pressure measured on arterial and central venous catheters did not differ, both representing MSFP. Both arterial and central venous pressures increase with tidal ventilation and with PEEP (Fig. 14). The position of the arterial or the venous central catheters did not have any impact on these results. Conclusion MSFP measured both on arterial and central venous catheters is altered by mechanical ventilation, with a tidal volume and PEEP effect, questioning the use of cardio-pulmonary interactions under mechanical ventilation for its extrapolation.

Competing interests
None. Introduction Hemodynamic monitoring is an indubitable device in the diagnostic and therapeutic management in critically ill patients. The pulmonary artery catheter (PAC), since 1970, is the gold standard of recommended devices due to the diversity and relevance of provided parameters. Pulse contour cardiac output (PiCCO) represents an interesting alternative to the PAC. However invasiveness and delicacy in the collection and interpretation of data make that the risk/ benefit ratio of these techniques is to be worrying. Our purposes were to assess the diagnostic, therapeutic and prognostic utility of invasive monitoring by PAC and PiCCO compared to non-invasive or less common tools. Patients and methods Evaluative retrospective study comparing two arms of ICU patients with invasive versus non-invasive hemodynamic monitoring. Were included, the ICU patients who had hemodynamic and/or respiratory failure and how required the use of hemodynamic exploration. Exclusion criteria were age less than 18 years and the lack of data on the medical record. Outcomes: diagnostic utility was judged on the consistency and quality of data interpretation, therapeutic usefulness on the therapeutic changes and favourable response following these therapeutic adjustments during the 1st 24 h of monitoring, and prognostic assessment was judged on the ventilation period, length of stay and mortality. Results 131 cases of hemodynamic monitoring were included (PAC/PiCCO: n = 71 and common tools: n = 60). The superiority of the invasive monitoring was objectified with the consistency of interpretation of the collected parameters (91.5 vs 78.4%, p = 0.044) and therapeutic changes derived by interpretation (83 vs 66.7%, p = 0.041). No benefit found on the clinical improvement following the modifications guided by monitoring (50 vs 57%, p = 0.72) and (31 vs 47%, p = 0.074) for the hemodynamic and respiratory responses respectively. The duration of ventilation and stay-length did not differ with those of the standard group. A higher mortality was found with age >50 years (  Introduction Septic cardiomyopathy is commonly encountered in patients with septic shock. Most studies suggest no correlation between left ventricular ejection fraction (LVEF) and mortality in patients with septic shock. In the other hand, the initial left ventricle global longitudinal strain (LVGLS) seems to be a better prognostic factor. However its time course evolution remains to be precisely defined. The present study aims to describe the evolution during the first days of the septic shock and to establish the prognostic value of this novel parameter in critically ill patients. Materials and methods A prospective observational single center study was performed in the ICU. After approval of the local ethics committee, all patient admitted to the ICU for septic shock without known pre-existing heart disease were eligible. In these preliminary results, 13 of the 100 planned patient were included and analysed. Echocardiography was performed on the first day, and repeated daily during ICU stay until norepinephrin was stopped. LVEF and LVGLS were acquired in apical two-chamber, four-chamber and long-axis views. Patients were divided into two groups: survivors and non survivors. Results Of the first 13 patients included mortality in the ICU was 23%. Left ventricular ejection fraction at admission (LVEF) was not significantly different among the survivors and non survivors (respectively 53% ± 12 vs 28% ± 9; p = 0.089). Initial LVGLS was lower (which indicates better function) in survivors as compared to non survivors (−13% ± 12 vs −6% ± 0.5; p = 0.044). Time course evolution of LVGLS is displayed in Fig. 15.

Conclusion
The preliminary results of the present study suggest that LVGLS has a better prognostic value than LVEF in the initial stage of septic shock.
Introduction Left ventricular diastolic dysfunction (LVDD) represents about 50% of the causes of heart failure [1]. However little is known about LVDD in the intensive care unit (ICU) setting. The purpose of our study was to determine the prevalence, clinical characteristics and outcome of LVDD in the ICU.

Materials and methods
We performed a prospective monocentric study in a medical ICU. Between august 2014 and December 2014, for all consecutively admitted patients, transthoracic echocardiography was realized during the first 24 h after ICU admission. LVDD was established according 2 definitions: 1 complete definition based on American Society of Echocardiography criteria [2], and 1 simplified definition. Results LVDD prevalence was 61.4% (102/166) according to the complete definition and 56.0% (93/166) according to the simplified definition. LVVD patients were older (66.5 vs 55 years, p < 0.0001). Male sex predominated (67.6 vs 48.4%, p < 0.0001). LVVD was more associated with cardiogenic pulmonary oedema, diuretics use, and new onset of atrial fibrillation, whether the full definition (25.5 vs 0%, p < 0.0001; 52.9 vs 34.4%, p < 0.0001; and 27.5 vs 14.1%, p = 0.044 respectively) or the simplified definition (28 vs 0%, p < 0.0001; 53.8 vs 35.6%, p = 0.002; and 29 vs 13.7%, p = 0.018 respectively). Unlike the full definition, LVDD patients detected with the simplified definition were associated with more important intra-hospital mortality (28 vs 12.5%, p = 0.016). Discussion LVDD prevalence found in our study is similar as in previous works in the cardiology setting. Higher incidences of atrial fibrillation and cardiogenic pulmonary oedema in the LVDD patients can be explained by a more important left atrial dilation, and a higher incidence of other cardiomyopathies in the LVDD cohort. Conclusion LVDD is a frequent abnormality in the ICU setting, associated with an important morbi-mortality. LVDD identification should alert the physician for an higher risk of new onset of atrial fibrillation and cardiogenic pulmonary oedema.

P39 Precision of measurements with transthoracic echocardiography in critically ill patients
Mathieu Jozwiak 1 , Julia Gimenez 1 , Jean-Louis Teboul 1 , Pablo Mercado 1 , François Depret 1 , Christian Richard 1 , Xavier Monnet 1 Introduction Hypovolemia is common in intensive care patients. It may reduce cardiac output and O 2 delivery relative to tissue need. Low mixed Central venous O 2 saturation (ScvO 2 ) has been used to optimize cardiac output and O 2 delivery relative to tissue needs. The use of ScvO 2 to predict fluid responsiveness is unclear. Objective: To determine whether ScvO 2 is a good indicator of fluid responsiveness. Patients and methods We carried out a prospective study in the medical intensive care unit of the teaching hospital in Mahdia over a period of 48 months. All patients with circulatory failure were enrolled. At baseline (t = 0 min), hemodynamic measurements were performed with a PICCO. After baseline measurements and blood sampling, 500 cc of fluids were given over 30 mn and hemodynamic measurement were performed at the end of fluid perfusion. Concomitant vasoactive and sedative drugs and ventilatory settings remained unchanged. Fluid responsiveness was defined as an increase in cardiac index >15%. Results A total of 68 patients requiring volume expansion were included. The causes of acute circulatory failure were septic shock (n = 45), cardiogenic shock (n = 11), and dehydration (n = 12). Among the 68 included patients, 33 (49%) were responders. ScvO 2 was significantly lower in no responders group (62 ± 11 vs 68 ± 12, p = 0.046). However ScvO 2 variation was significantly higher in responders group (10 ± 8 vs 6 ± 7, p = 0.041). The area under the ROC curve for ScvO 2 was 0. 37 (95% CI 0.23-0.5). The best cutoff value of ScvO 2 was <55% (sensibility = 36%, specificity = 8 8%, positive predictive value = 75% and negative predictive value = 60%). The area under the ROC curve for ScvO 2 variation was 0.65 (95% CI 0.52-0.78). The area under the ROC curve for delta ScvO 2 was 0.5 (95% CI 0.43-0.7). Conclusion In this study, we can't recommend strongly the use of ScvO 2 as a predictor of fluid responsiveness.

Introduction
The application of tactical tourniquet is one of the most important lifesaving intervention (LSI) on the battlefield and in theaters of terrorist attacks. Despite its apparent simplicity, the failure rate of tactical tourniquet application remains high in real life experience. Some personality traits of the combat lifesavers could explain this high failure rate. The aim of this study was to analyze the association between different personality traits and the performance for applying the tactical tourniquet. Materials and methods This observational, cross-sectional study concerned French soldiers of two theaters of operations in Africa (SANGARIS in Central African Republic and BARKHANE in Malia), between October 2015 and April 2016. The performance for application of the tactical tourniquet SOFFT ® was evaluated during simulation sessions and included the time for application and its effectiveness, demonstrated by the elimination of the popliteal pulse Doppler signal. A performance score included both an application time shorter than 60 s and the interruption of the Doppler signal. Personality traits were assessed using validated questionnaires: the Self-Esteem Scale of Rosenberg (SES), the HEXACO personality inventory for empathy (HEXACO), the Freiburg Mindfulness Inventory (FMI) and the Büssing Altruism Scale (BAS). Results A total of 72 participants (mean age 27 ± 4) were included in the study. The effectiveness rate of tactical tourniquet use was 51%. Only 26% of the participants performed this LSI in accordance with the performance score criteria. Some of the personality traits were associated with significant differences in the performance of the SOFFT ® application. Thus, the most empathetic soldiers (high HEXACO levels) had weak performance scores, in comparison with those with a lower HEXACO level (p = 0.016). Participants with a high BAS level were also less performant for application of the tactical tourniquet. However, combattants with a high SES applied more quickly and more efficiently the tactical tourniquet (Fig. 16). Conclusion Personality traits are associated with significant differences for the performance of tactical tourniquet application, a crucial LSI in setting of both combat and terrorist attacks. Personality traits can be evaluated easily, using validated questionnaires. Finally, the "adaptive learning" could represent an adequate approach to improve the LSI training, according to the personality traits of the trainees.
Introduction Delivering bad news regarding prognosis or care-related damages are difficult situations generating anxiety for patients but also for the medical teams who feel generally insufficiently prepared and trained. We have studied the interest of training paediatric and anaesthesiologist residents and fellows in difficult communication through simulation. Materials and methods This multimodal training with multidisciplinary teachers (psychologists and doctors) included theoretical interactive sessions (behaviour and environment to inform parents in an acute situation, modalities of communication, and knowledge about the defense mechanisms…) and simulation sessions with professional actors playing child's parents. The scenarios promoted situations after significant events involving a child in PICU or in the operative room. Each scenario was created from real case. Standardized debriefing by teachers and a psychologist was done after each scenario with actors and every participant. A pre and post-test using Lickert scale was fulfilled to evaluate the interest of this program and the confidence in this kind of meeting. The results are expressed in median [interquartile] and comparison used the U Man-Whitney test. A p value <0.05 was considered significant. Results Thirty-one medical residents (17 anaesthesiologists, 14 paediatricians) and 5 fellows (1 anaesthesiologist and 4  See Fig. 17 Discussion The participants reported a significant lack of initial training and a lack in confidence in communication skills with relatives' despites their strong interest for this topic. The majority has still encountered such a situation. The post test evaluation showed a significant improvement in confidence even though the level of anxiety remained high after the session. The multimodal aspect of the program combining theoretical courses and realistic simulation with professional actors in a "secured" environment was emphasized by the learners. Conclusion This study confirms the feasibility, the realism and the learners' satisfaction. It's of major interest to pursue development of such interprofessional training with others caregivers (nurses…) and other medical and surgical specialists.

Competing interests
None.  Introduction Though efforts have been made in the previous decades to teach medical students the basics of clinical research, their own involvement in such studies infrequently occurs before they become residents or assistants. The medical courses include lessons on experimental and clinical research methods, which may be difficult to integrate out of a concrete context. On the other hand, much has to be learned for medical students in ICU within a short training period. Involving medical students in research could increase their workload, and may affect their evaluation of the training period. We therefore aimed to have medical students making their first foray in clinical research through a practice evaluation. We asked them to evaluate the change in workload and their global satisfaction afterward. Patients and methods Four medical students (4th year of medicine) agreed to participate to a clinical trial through an evaluation of medical practices. None of them had had a previous experience in this field. Gastric stress ulcer prophylaxis (SUP) was chosen by the students among other subjects. Two references were provided, in French, to give them clues on how to conduct the research. The first one was a free web access medical thesis exploring the medical practice on ulcer prophylaxis in three ICU in Toulouse. The second was a chapter of a reference textbook. Three weeks thereafter, it was decided in a meeting to have a 2 steps approach. First, they realized an evaluation of the local practices through a daily monitoring of patients admitted during a 2 weeks period. All patients admitted for an estimated period of more than 48 h were eligible. Exclusion criteria were baseline ulcer treatments and active gastrointestinal bleeding. Second, a brief survey of seniors and residents' knowledge on the subject of matter was performed. In both steps of investigation, the recorded criteria were those described in the French health authority's recommendations. At the end the study period students participated to statistical analysis and interpretation of results. The study was presented to the department's staff by the students. A few weeks later, they were asked to evaluate this experience. Results During the study period, 27 patients were screened and 14 patients (Mean age 57.4 ± 23.5 years, 4 female, 10 men, mean SAPS II 48 ± 25) were included and observed during a total 105 ICU days (7.5 ± 6.4 days/patient). 6 (42.9%) patients received stress ulcer prophylaxis at least once, and during 71.6 ± 32.6% of their ICU stay. According to French recommendations, SUP was not justified at least once in 5 out of the 6 patients. In 29% of SUP days, no criteria supported this prescription. 8 patients did not receive SUP. These patients were observed during 66 days. According to guidelines, SUP should have been offered in 2 patients during 15 ICU days. In this subgroup of patients no criteria for SUP was found in only 13 (19.6%) days. Among the studied criteria, a strong link between SUP and enteral nutrition was found (p < 0.0001, Fisher exact test). 8 senior intensivists and 6 residents answered the survey. The estimated prevalence of this condition was evenly distributed between less than 1% and more 10% among responders. 30% of them ignored whether guidelines were available, and 80% of all responders ignored the strength of evidence of the guidelines. All the known risk factors for stress ulcer were cited, but the two major criteria (mechanical ventilation and coagulation abnormalities) were recognized in less than 30% of responders. 86% of responders declared that only proton pump inhibitor were appropriate for SUP. All students declared that adding this study to the global burden of the training period was manageable. They pointed out that building a study was a difficult task and declared that it was an added-value to the training course, and a good experience in their medical course. They would agree to do it again. Discussion This study is limited by its single center design, and the little number of participating students and included patients. Conclusion Helping medical students make their first steps in clinical research seems to be feasible in ICU. The students greatly appreciated this experience and judged it improved their training period. It could also represent a mean for patient management improvement.

Competing interests
None.

P46 Does observational role allow skills acquisition in cardiopulmonary resuscitation simulation training?
Emilie Duburcq-Gury 1 , Léa Introduction Technical progress and development of indications may lead non-specialized ICU implanting ECMO, with an investment in equipment and appropriation of new proccess. Acquisition and retention of new skills to manage these specific situations isessential but difficult because these events are infrequent at an individual level. Full scale high-fidelity simulation conducted in multi professional team is a way to improve skills and confidence of the teams for these rare situations but with immediate vital risk. Materials and methods A simulation program on emergency management in relation with ECMO has been jointly set up by members of the intensive care unit for disciplinary expertise in connection with the CEnSIM team for pedagogical expertise. Three priorities have been identified: • In situ high fidelity simulation • Multi-professional team simulation (1 intensivist physician, 2 nurses with at least 1 ECMO referrer, 1 caregiver) by 4 h session. • Program proposed to the whole service to ensure support for uniformity in healthcare These sessions were conducted in an intensive care room adapted to meet the requirements of high-fidelity simulation. Each session consisted of two simulations each followed by a debriefing. Two objectives were targeted: technical achievement and teamwork. Beyond the reactions, the main assessment was performed by measuring the evolution of the participant self-efficacy defined by Bandura as "an individual's belief in its ability to organize and execute the course of action required to produce desired results. " Measurements were performed before and after training using analogic visual scale of self-efficacy tailored specifically to each profession.
Results Two test sessions and ten training sessions were conducted over a period of 6 days, allowing the participation of all the physicians and ECMO referral nurses (total participants n = 44: 10 physicians, 4 residents, 20 nurses, 10 caregivers). This training was appreciated by all participants, both in terms of emotional perception (overall satisfaction: M = 4.9/5; SD = 0.3) and in terms of practical reinvestment (perceived usefulness: M = 4.9/5; The results show a beneficial effect of this program on self-efficacy for the three professional categories: Delta (pre-form-post-form) caregivers score = 2.63; p < 0.005; Delta (pre-form-post-form) nurses score = 2.6 nurses; p < 0.001; Delta (pre-form-post-form) physicians score = 3.46; p < 0.001. Conclusion Inter professional full scale simulation allows the acquisition of experiential knowledge and skills essential to the management of these complex situations. Beyond teamwork and technical skills, significantly increased self-efficacy is an important factor in achieving optimal behavior in emergencies. In situ simulation performed in interprofessional teams is probably an interesting method to optimize management of ECMO patient. Implementation and perpetuation of such projects need a strong institutional commitment as part of a policy.
Introduction Mechanical ventilation is a life supporting treatment commonly indicated in patient's ICU [1]. However, discrepancy in ventilatory settings can lead to patient-ventilator asynchrony. This adds a burden on the respiratory system and may increase the morbidity and mortality in the critically ill.
The discrepancies in ventilatory settings may be largely underestimated because of a frequent lack of monitoring [2]. The aim of the study was to evaluate frequency, typology and severity of discrepancies in ventilatory settings in a Tunisian medical ICU, and to identify factors associated with patient-ventilator asynchrony. Patients and methods An audit observational study was conducted in a 7-beds medical ICU during 1 month period (August 2016). All consecutive ICU patients requiring invasive or non invasive mechanical ventilation were included. The data collected were: patient's characteristics, initial diagnosis, SAPSII, PaO 2 /FiO 2 ratio, type of mechanical ventilation, ventilatory mode, prescribed ventilatory parameters and interfaces. Were analyzed, peak inspiratory pressure, plateau pressure, auto-PEEP, volume and inspiratory and expiratory flow waveforms. Patient-ventilator asynchrony was defined as: ineffective inspiratory efforts, auto-triggering, delayed cycling, double triggering and inspiratory waveform distorsion. Results During the study period, were performed a total of 160 ventilatory settings observations. Mean age was 59.9 ± 16.8 years, mean SAPSII score was 36.3 ± 11.1. 157 (98.1) patients were on invasive mechanical ventilation and 3(1.9) were on non-invasive mechanical ventilation. PaO 2 /FiO 2 , 286 ± 115; mean peak inspiratory pressure, auto-PEEP and plateau pressure were respectively: 35.5 ± 11.8; 8.4 ± 4.5 and 22.0 ± 4.7 cmH 2 O. Discrepancies in ventilatory settings were found in 55(34.4) patients. 23(42) patients had frequent patient-ventilator interactions. Ineffective efforts and double triggering were the two most common asynchronies (43.4 and 13.1% respectively). Patient-ventilator interactions were assessed as severe in 9(16.3). Patient-ventilator asynchronies were associated neither to the severity (SPAS II, PaO 2 /FiO 2 ) nor to the respiratory system mechanic (peak inspiratory pressure, autoPEEP, plateau pressure and driving pressure). A simple intervention on the ventilatory settings corrected the asynchronies in all the patients. Conclusion The discrepancies in ventilatory settings reveals frequent. Patient-ventilator asynchronies are the most observed discrepancies. This could be the consequence of a frequent lack of monitoring.  11 janvier 2006). Large-scale feasibility of such a training, retraining and assessment remains a big challenge. We conducted a study to evaluate the training efficacy of a serious game designed for cardiopulmonary resuscitation and applied to secondary school students. Materials and methods Students of 6th and 7th grades from Paris region were randomized (1:1) to one of the 2 groups using a computerized system. Randomization was stratified by classroom. In the Control group, children received a 30-min teaching on nutritional wellbeing. In the Interventional group, children played a serious game reproducing real-life cardiac arrest situation (3D real-time simulation software, Stayingalive ® , iLUMENS-Dassault Systemes) without any adult intervention. Two months later, each student was observed and evaluated by 2 assessors blinded to the study group during a cardiopulmonary resuscitation session using mini manikins MiniAnnePlus ® (Laerdal Medical). The assessment scale included 15 items divided into 4 categories (cardiac arrest recognition, call for help, chest compression and use of defibrillator). Primary endpoint was the total score obtained by each student. Secondary endpoints included performance for each item. Results A total of 97 children were included in the analysis (Interventional group n = 50, Control group n = 47). Median total score was significantly higher in the Interventional group (7 [6-9] vs 6 [5-8], p = 0.02, Fig. 18

Fig. 18 Mean difference between groups and according to class grade
Conclusion A 30-min experience with a cardiopulmonary resuscitation serious game without any adult intervention improves basic life support performed by secondary school students. It may be further improved by adult supervision, repeated use, and gamification using leaderboards. These results are promising to implement on a larger scale such serious games to improve knowledge and skills of children.

Competing interests
Mini manikins were provided by Laerdal Medical. Among 519 pregnant women who tested positive, only two cases of microcephaly has been identified. One patient a 76-year-old man died. Our objective was to describe severe forms of Zika-associated GBS complicated by cardiovascular disorders in patients admitted to the Intensive Care Unit (ICU) during the onset of the endemic period.

Patients and methods
Prospective study of all Zika-associated GBS in patients admitted in our ICU from 1/12/2015 to 5/25/2016 and presenting at least one organ failure using the SOFA score. During this period, all our ICU patients were systematically tested for dengue fever, Chickungunya and Zika viruses. Cases included in this study were defined by the presence of GBS clinical signs (e.g., distal paresthesias) with biological confirmation using RT-PCR blood and urine and also CSF, and/or positive serology for IgM or IgG. Conclusion Zika infection may lead to ICU admission due to the development of GBS. Physicans particularly cardiologists should be aware of additional cardiovascular complications secondary to GBS-related dysautonomic complications. An emergent need for collaborative and multidisciplinary approach is required to mitigate the imminent Zika outbreak in the Caribbean, by providing diagnostic laboratory facilities, ICU beds availability and assistance for appropriate management of the environment. A prospective study is currently conducted to evaluate the cardiovascular complications associated with fatal cases of Zika infection virus.

Introduction
The outcome of septic shock in cancer patients has dramatically improved over the last two decades, but the impact the acute inflammatory insult on the further prognosis of cancer is unclear. Indeed, sepsis-induced immune dysfunctions may directly impact on the prognosis of the underlying malignancy in survivors. We are currently developing a research project in order to investigate the reciprocal relationships between bacterial sepsis and malignant tumor growth. We first reported that sepsis-induced immune suppression promoted malignant tumor growth when tumor cells were inoculated in post-septic mice. Conversely, we observed that sepsis may inhibit the growth of previously established local and metastatic tumors. The present study aimed at investigating the cellular and molecular mechanisms of sepsis-induced tumor inhibition.

Materials and methods
We used 8-12 w.o. C57BL/6 J mice. Mice were first subjected to malignant tumor inoculation by subcutaneous injection of the MCA205 fibrosarcoma cell line. Seven days after tumor inoculation, mice were subjected to polymicrobial sepsis induced by cecal ligation and puncture (CLP), without any subsequent antibiotic treatment. Controls were cancer mice subjected to sham surgery. The features of anti-tumoral immune response were compared between CLP-and sham-operated cancer mice at day 1 (early assessment) and day 7 (late assessment) following surgery. The intra-tumoral infiltration of immune cells within tumor tissues and draining lymph nodes was assessed by flow cytometry, and the intratumoral cytokine production was assessed by ELISA. We also investigated the role of bacteria and related pathogen-associated molecular patterns on growth of MCA205 cells. The expression of Tlr2 and Tlr4 genes was assessed by southern blot after DNA amplification.
Results We first confirmed that polymicrobial was able to inhibit malignant tumor growth when applied in mice previously inoculated with fibrosarcoma MCA205 cell line. Not surprisingly, the bacterial load in blood, kidneys and tumors were higher at day 1 in CLP-operated mice than in sham-operated counterparts. Of note, the bacterial contamination of tumors was not sustained at 7 days. We checked that MCA205 cells expressed functional Tlr2 and Tlr4 receptors. MCA205 cells cultured in the presence of Tlr4 and Tlr2 agonists, respectively lipopolysaccharide (LPS) and heat-killed Staphylococcus aureus, did not display accelerated cell growth in vitro, suggesting that the antitumoral effect of sepsis is mediated by the septic host's response. The distribution of myeloid and lymphoid immune cells within tumor tissue was quite similar in sham-and CLP-operated mice, with the exception of increased proportions of inflammatory monocytes and Tgd lymphocytes in septic mice. The intra-tumoral cytokine pattern of CLPoperated mice was skewed towards an increased production of both the pro-inflammatory Tnf-a and the anti-inflammatory IL-10 cytokines at day 1.

Conclusion
Polymicrobial sepsis applied to cancer mice inhibits the local growth of a malignant fibrosarcoma tumor. This anti-tumoral effect is not directly driven by pathogens, but rather involves an enhanced intratumoral inflammatory response. In the light of mild quantitative cell alterations, our results suggest that sepsis induces a functional modulation of immune cells resulting in potent antitumoral activity.
Introduction Hematological malignancies may be directly responsible for life-threatening organ failures through tumor lysis syndrome, tissue infiltration, coagulation disorders and obstruction of anatomical structures. Respiratory and renal dysfunctions are commonly encountered in this setting. In addition, some patients may develop a systemic inflammatory response syndrome, responsible for acute circulatory dysfunction, so-called sepsis-like syndrome. Besides advanced life support, the treatment of those cancer-related organ failures relies on timely administration of chemotherapy. In this study, we addressed the features and outcomes of patients with sepsis-like circulatory dysfunction related to hematological malignancies, with particular emphasis on the impact of chemotherapy on organ failures. Patients and methods This was a 9-year (2007-2016) single-center retrospective observational study performed in a 24-bed medical ICU. Inclusion criteria were age ≥18 years AND presence of a hematological malignancy, either already known at the time of ICU admission or diagnosed during the ICU stay AND development of acute circulatory dysfunction requiring vasopressors without any evidence of underlying infection. Patients with hematological malignancies who received chemotherapy in the ICU were retrieved through the information systems from hospital pharmacy units involved in the delivery of cytostatic drugs, and all medical files were individually checked for inclusion criteria. Data were collected from individual files, and included the overall severity through the APACHE2 and SOFA scores computed at the time of ICU admission. The SOFA score was thereafter computed daily. Endpoints were the in-ICU and in-hospital vital status. Results Over the study period, 24 patients (12 men, 12 females) fulfilled the inclusion criteria. Their mean age was 62 ± 16 years. Most of them were in good functional condition since 22 had a performance status of 0. Two patients were previously immunocompromised (severe combined immune deficiency and kidney transplant). The underlying haematological malignancies were distributed as follows: non-Hodgkin lymphoma (n = 18) including 6 patients with diffuse large B-cell lymphoma, acute myeloid leukemia (n = 4), Hodgkin's lymphoma (n = 1) and chronic lymphocytic leukemia (n = 1). Seven patients had malignancies newly diagnosed in the ICU. Otherwise, the median time from diagnosis to ICU admission was 8 days (min 2 days; max 4357 days). The primary reasons for ICU admission were respiratory (n = 11), renal (n = 7), neurological (n = 6) and hemodynamic (n = 5) failures. The mean admission APACHE II and SOFA scores were 18.8 ± 9.0 and 8.6 ± 3.4, respectively. The mean blood lactate level was 4.9 ± 1.4 mmol/L. Mechanisms of organ failures were related to tumor lysis syndrome (n = 10), hemophagocytic lymphohistiocytosis (n = 4), lung (n = 4) and liver (n = 2) malignant infiltration, and disseminated intravascular coagulation (n = 2). Acute circulatory failure requiring vasopressors was present prior to chemotherapy in 18 patients, and was secondarily triggered by chemotherapy in 6 patients. In the first group, administration of chemotherapy was associated with a dramatic improvement in the circulatory conditions (Fig. 19) as assessed by the trend in the hemodynamic SOFA variable. The in-ICU and in-hospital mortality rates were 75 and 79%, respectively. The main cause of death was untractable multiple organ failure (n = 14).

Discussion
The mechanism of organ failures in this setting is questionable, but is presumably linked to a massive release of pro-inflammatory cytokines and endogenous danger signals by malignant cells.
Modulation of this overwhelming inflammatory response may represent a field of investigation as well as a potential therapeutic target in the future. Conclusion Urgent administration of chemotherapy is associated with fast improvement in circulatory conditions in patients with sepsis-like shock related to hematological malignancies. However, the overall outcome remains very poor.
Introduction Glucotoxicity is known to occur during hyperglycemia and glycemic variability, situations that are associated with poor outcomes in sepsis patients. Although cardiovascular dysfunction is a major cause of mortality during sepsis, the impact of glucotoxicity on the cardiovascular system during sepsis remains unknown [1]. Beneficial effects of protein tyrosine phosphatase 1B (PTP1B) deletion, a negative regulator of insulin signaling, on glucose homeostasis and cardiovascular dysfunction during endotoxemia have been reported. We hypothesized that exogenous glucose administration during inflammation increases cardiovascular dysfunctions by activating glucotoxicity pathways and that this is prevented by endothelial PTP1B gene deletion.

Materials and methods
For this purpose, we generated an endotoxinic model with glucose administration. EndoPTP1B−/− or wild type (WT) mice received LPS (1 mg/kg) or saline solution followed by five injections of glucose (2 g/kg) or saline solution each hour 12 h after LPS. Endothelial deletion of PTP1B was generated by crossing LoxP-PTP1B with Tie2-Cre mice. The exploration of the cardiac function is performed 20 h after LPS injection (H 2 O) by the non-invasive The fractional shortening tended without significance to decrease in mice receiving LPS with glucose compared LPS alone. Analysis of cardiac tissue showed a significant deleterious impact of LPS and glucose association on oxidative stress (gp91phox) and inflammatory responses with increased cytokines production and expression of adhesion molecules (VCAM-1 and ICAM-1). The cardiac gene expression of CHOP and GRP78 (endoplasmic reticulum stress) in WT showed no change in the condition LPS with glucose. Moreover, expression of PARP1, sensor of DNA damages and activator of glucotoxicity pathways, is increased with glycemic variability. Despite an increased expression of PTP1B when LPS and glucose are combined in wild type mice, endothelial PTP1B deletion gene only reduced significantly the cardiac expression of iNOS, PARP1 and PTP1B. Finally, cardiac systolic function tended without significance to be impaired by glucose and LPS combined. Discussion The absence of variation in the gene and protein expression of CHOP and GRP78 shows a lack of endoplasmic reticulum stress in our model despite cellular stresses generated by glucose. Conclusion In endotoxinic model, the variability of blood glucose aggravates the effects induced by an inflammatory trigger in the cardiac tissue without significant impaired cardiac function in this model. Endothelial deletion in PTP1B, involved in the regulation of glucose homeostasis, provides improved glycemic control, with a reduction of pathological activation of iNOS, a marker of abnormal vascular function, and PARP1 marker DNA damage. Unité espri-biobase, CHRU Nancy, Vandoeuvre-lès-Nancy, France;

P54 Effects of low doses of esmolol on cardiac and vascular function in experimental septic shock
3 Service de biochimie, CHRU de Nancy, Vandoeuvre-lès-Nancy, France; Introduction Administration of a selective β1-blocker, such as Esmolol, in septic shock has demonstrated cardiovascular protective effects related to a down-regulation of inflammation. However, the administered dose systematically induced a reduction in heart rate of approximately 20%, thus limiting its prescription at bedside. The present study aimed to determine whether a non-chronotropic dose of Esmolol still maintains its protective cardiovascular and antiinflammatory effects in experimental septic shock. Materials and methods Four hours after cecal ligation and puncture (CLP), Wistar male rats were randomly allocated to the following groups (n = 8): CLP, CLP + E1 (Esmolol: 1 mg kg −1 h −1 ), CLP + E5 (Esmolol: 5 mg kg −1 h −1 ), CLP + E18 (Esmolol: 18 mg kg −1 h −1 ). An additional eight rats underwent Sham operation. All rats received a continuous infusion of saline, analgesic and antibiotics 4 h after the surgery. Assessment at 18 h included in vivo cardiac function by echocardiography and ex vivo vasoreactivity by myography. Circulating cytokine levels (IL-6 and IL-10) were measured by ELISA. Cardiac and vascular protein expressions of p-NF-κB/IκBα, iNOS, p-AKT/AKT and p-eNOS/eNOS were assessed by Western blotting.
Results CLP induced tachycardia, hypotension, cardiac output reduction, hyperlactatemia and vascular hypo-responsiveness to vasopressors. Compared to CLP animals, heart rate was unchanged in CLP + E1 and CLP + E5 but was reduced in CLP + E18. Stroke volume, cardiac output, mean arterial pressure and lactatemia were improved in CLP + E1 and CLP + E5 while vascular responsiveness to Phenylephrine was only improved in CLP + E5 and CLP + E18. Plasma IL-6 levels were decreased in all Esmolol groups. p-NF-κB was decreased in both cardiac and vascular tissues in CLP + E5 and CLP + E18. Conclusion In experimental septic shock, low doses of Esmolol, still improved cardiac function and vasoreactivity. These benefits appear to be associated with a modulation of inflammatory pathways.
Introduction For patients treated with cisatracurium, it has been shown that proinflammatory cytokines (IL-1, IL-6, and IL-8) significantly went down during the 48 h of molecule administration as compared to placebo [1]. This anti-inflammatory effect is mediated in part by an inhibition of nicotinic acetylcholine receptor [2]. We did not find in the literature any study exploring the post-discontinuation phase of neuromuscular blocking agent apart the fact that blood oxygenation improved after interruption of cisatracurium infusion. The goal of present study was to demonstrate inflammatory rebound defined from the systemic inflammatory response syndrome (SIRS) criteria after discontinuation of cisatracurium. Patients and methods It was a prospective, single-center, observational study. We included adult patients admitted to the medical intensive care unit of the Croix-Rousse hospital in Lyon, who received mechanical ventilation and cisatracurium infusion between February and August 2016. A rapid intravenous infusion of 15 mg of cisatracurium was administred followed by a continuous infusion of 37.5 mg per hour for at least 24 h. After inclusion in the study, SIRS criteria were monitored daily from cisatracurium onset until the 72th hour after molecule interruption. The SIRS was defined by the presence of at least 2 of the followings: temperature >38 °C or <36 °C, heart rate >90/min, respiratory rate >20/min or PaCO 2 <32 mmHg and white blood cells count >12 G/L or <4 G/L or >10% immature neutrophils. Inflammatory rebound was defined as the increase of SIRS by at least 1 criterion from the baseline defined as the cisatracurium discontinuation. The primary outcome was the prevalence of inflammatory rebound 24 h after the cisatracurium discontinuation.
Results Thirty-nine patients were enrolled. The prevalence of inflammatory rebound 24 h after the cisatracurium discontinuation was 56.4%. No risk factor in the multivariate analysis was associated with the presence of inflammatory rebound and this rebound did not affect mortality. We also noticed an increased risk of inflammatory rebound with time after cessation of cisatracurium (Fig. 20) and this risk was not depending on the length of cisatracurium administration before its discontinuation.
Conclusion This study demonstrated that inflammatory rebound occurred in more than half of patients after interruption of cisatracurium infusion and its risk of occurrence increased with time.
Introduction The Veno-Arterial ExtraCorporeal Life Support (VA-ECLS) is an efficient therapy in refractory cardiogenic shock. Nevertheless, VA-ECLS is associated with serious infectious complications. The objective of this study was to investigate the epidemiology and the risk factors of infection of femoral cannulation site in patients with VA-ECLS after cardiac surgery.  . 21). Pseudomonas aeruginosa, Enterococcus faecalis, Escherichia coli and Enterobacter cloacae were identified in 16, 12, 12 and 10% respectively. The other bacteria were klebsiella pneumoniae in 15% and staphylococcus in 10%. In univariate analysis, risk factors of cannula site infection were a multiresistant bacterial colonization, a bacteremia event and/or a bleeding event in VA-ECLS cannulation site. In multivariate analysis, only a bleeding event in VA-ECLS cannulation site was identified as a risk factor cannulation site infection. The intrahospital mortality was similar in both groups but the hospitalization length stay increased in the «infected» group, 45 vs. 26 days (p < 0.0001) respectively. Furthermore, reoperation increased significantly in the «infected» group, 71 vs. 2% of patients (p < 0.0001) respectively. After reoperation, a reconstructive surgery of the cannulation site was necessary in 45% of cases. Conclusion In patients treated with VA-ECLS after cardiac surgery, inguinal cannulation site infection occurs in 27% of patients. Bleeding event is the main risk factor.

Materials and methods
Introduction Few studies focused on patients with severe infective endocarditis (IE) in Intensive Care Unit (ICU), and despite major advances in both diagnosis and therapeutic procedures, it still carries poor prognosis. In 2009, European guidelines on antibiotic prophylaxis have changed and became more restrictive. The goal of this study is to describe changes in incidence, characteristics of patients, management, in-hospital mortality and pathogens in an overall population of patients admitted to ICU for a definite IE. Through comparison of two We observed a significant increase in crude annual incidence (153.0 ± 14.9 vs 217.6 ± 46.4; P = 0.018) and in incidence density relative to number of stays (0.79 ± 0.08 vs 1.1 ± 0.02% P = 0.004). Despite a trend towards increasing number of cases over the first period (1997-2009), slopes of incidence density curves clearly indicate an acceleration of the number of cases since 2009 (P = 0.017). Patients treated during the second period are significantly older and more severe than those treated before 2009. Surprisingly, use of invasive ventilation, renal replacement therapy, and vasopressor were significantly lower during second period. Contrariwise, resort to surgery has doubled between two periods. ICU mortality is significantly lower in the second period but in-hospital mortality remains unchanged. Concerning pathogens, we found a significant increase in incidence of Streptococcus spp. and Staphylococcus spp., and no changes concerning intracellular bacteria, Enterococcus spp., Candida spp. or Gramnegative bacilli.
Conclusion Despite some limitations inherent to its retrospective design and to potential diagnostic coding bias, our study highlights a quick shifting landscape in the epidemiology of infectious endocarditis in intensive care, characterized by a strong increase in the incidence and changes in bacterial epidemiology. Restrictive bend in antibiotic prophylaxis guidelines could be substantially responsible for these trends.
Introduction Heparin is widely prescribed in patients admitted in surgical intensive care units (SICU) to prevent venous thromboembolic events. Heparin-induced thrombocytopenia (HIT) is a rare but potentially life-threatening complication of heparin therapy. HIT requires the emergent discontinuation of heparin and the prescription of an alternative anticoagulant therapy that could be difficult to manage in SICU patients and enhance the risk of hemorrhagic complication. The early diagnosis of HIT in SICU patients remains a challenge. As thrombocytopenia could reveal several SICU complications, the 4T score of Warkentin is a useless tool to efficiently discriminate patients having or not HIT and the biological confirmation of HIT is delayed. The HIT Expert Probability (HEP) score has been reported to have a higher predictive value than the 4T score in non-ICU patients. The purpose of the study was to compare the HEP score to the 4T score in the early diagnosis of HIT in SICU patients.

Materials and methods
We conducted a one-center prospective observational cohort pilot study (www.ClinicalTrials.gov Identifier NCT02790567), included all consecutive patients admitted in our SICU between October 2013 and May 2015 and suspected to have HIT. Non-inclusion criteria were pregnancy, age <18 years old and treatment with fondaparinux. The day the diagnosis of HIT was suspected, the HEP and the 4T scores were calculated and the following blood analyses were performed: the ID-PaGIA Heparin/PF4 test, the ELISA test, the heparin-induced-platelet-aggregation (HIPA), and the serotonin release assay (SRA). After completion of the study, all medical files were reviewed by a multi-disciplinary independent committee to discriminate patient having (HIT group) or not (SAFE group) a HIT. The final diagnosis was based on the medical history of the patient, on the time-variation of the platelet count while heparin was discontinued or not, and on the results of the venous Doppler of the 4 limbs. The committee was blinded from the value of the HEP and of the 4T scores, and from the results of the ID-PaGIA Heparin/PF4 test. The ROC curves of the HEP and of the 4T score were constructed. The sensibility (Se), the specificity (Sp), the positive predictive value (PPV), and the negative predictive value (NPV) of the HEP and of the 4T scores were calculated, and the areas under the ROC curves (AUC) were compared by using a Chi2 test. Conclusion Based on a small prospective observational pilot study, a HEP score ≥5 could have a higher predictive value than a 4T score ≥6 in the early diagnosis of HIT in SICU patients. This result needs to be confirmed in a larger multicenter study.

Competing interests
None.

Fig. 22 Receiver operating characteristic curves for the HEP and for the 4T scores
Introduction Drug poisoning is a frequent cause of hospital admission especially in intensive care unit (ICU). Despite advances in treatment, hospital mortality of severe acute poisoning admitted in ICU seems to increase. Purifying methods, continuous haemodiafiltration veinoveinous (CVVHDF) and molecular adsorbent recirculating system (MARS) were developed with promising clinical results [1]. However no analytical study has quantified their accurately purifying efficiency. It has not been assessed efficiency of the different compartments of MARS nor its advantages over other methods of dialysis and filtration. The objective of this study was to quantify the purifying efficiency of the different compartments of the CVVHDF and MARS and to compare their respective efficiency in an ex vivo model in the most favourable conditions for these methods to assess their maximum capacity purification.

Materials and methods
We performed an ex vivo study based on a manipulation bench simulating intoxication pentobarbital at a plasma concentration of 40 mg/l injected into a central compartment (5 l) devoid of transporter proteins (200 mg of pentobarbital). The EC extraction coefficient [EC = (in concentration − out concentration)/in concentration] were calculated for each compartment of CVVHDF and MARS as well as the amounts withdrawn by the sum each compartment allows to assess the overall capacity of each technique.
Results At the end of 6 h simulation with CVVHDF, the remaining material in the central compartment was 3% of the total quantity injected. The cumulative amount removed in total effluent was equal to 95%. The non-recovered amount was equal to 2%. EC CVVHDF was almost constant with an average value of 14% with large variations however (range 2.5-27%). It was not observed to release at the end of manipulation.
At the end of a 6 h simulation with MARS, the remaining amount in the central compartment was undetectable. The cumulative amount removed in total effluent from the hemodiafiltration column was equal to 11.5% the amount removed by carbon column was 88.5%. The sum of the amounts removed by the effluent and the carbon column reflects the total amount injected. The EC hemodiafiltration compartments and charcoal hemoperfusion were 25 and 70% respectively. Purification of pentobarbital was complete after 3 h of a MARS session. Discussion MARS is the most effective purifying method in a pentobarbital-simulated intoxication. This rapid and complete treatment is mainly due to purifying coal capacity. It was not observed the cartridge saturation for the amount of administered pentobarbital. Conclusion Future studies should determine the parameters that may affect the treatment capacity by both methods.
Introduction Plant poisoning is generally not life-threatening. Its occurrence in children is commonly accidental. In adults it results from a suicide attempt or from its use for addictive or therapeutic purposes. Because of their specificity in clinical presentation and outcomes, we have conducted this study. The objective of this study was to report all plant poisoning cases collected in Tunisian toxicological intensive care unit, with their epidemiological, clinical characteristics and outcomes.
Patients and methods A retrospective study was performed between January 2007 and December 2015.
Epidemiological data and clinical outcomes were reviewed. Data were analyzed using SPSS.
Results During the study period, 38 patients were included. Sex ratio was 1. Mean age was 43 years (5-65). Poisoning was accidental in 88% of cases. Most frequently incriminated plants were Datura stramonium (40%) with an anticholinergic toxidrom present in all cases, Ricinus communis (23%) with gastrointestinal manifestations (present in 88% of cases), Nerium oleander (9%) with digitalis toxicity-like symptom in 66% of cases, Hyoscyamus Niger (12%) with anticholinergic symptoms in 75% of cases and hallucinogenic effects in all cases, Atractylis gummifera (9%) with gastrointestinal and acute liver failure symptoms, and Peganum harmala (9%) with only gastro-intestinal effects. All patients received supportive care. Mortality rate was 8, 5%, interesting children, and was secondary to multi-system organ failure due to ingestion of Atractylis gummifera.
Conclusion Each year 2.2% of our poison center calls report exposures to toxic plants. Most of these exposures are of minimal toxicity largely because of the fact that they involve pediatric ingestions, which are of low quantity. Public education is important to minimize these poisonings and must be oriented primarily towards children to reduce mortality.
Introduction Herbal remedies have been used for centuries to treat a variety of diseases. Mediterranean Buckthorn (Rhamnus alaternus) has been used for therapeutic purposes and no toxicity effects have been documented. Rhamnusalaternus (Rhamnaceae) is a small tree located mainly in the North of Tunisia, where it is known as ''Oud El-khir". It has traditionally been used as a diuretic, laxative, hypotensive drug and for the treatment of diabetes, hepatic and dermatologic complications. Previous phytochemical studies have shown potent antioxidant, free radical scavenging, antimutagenic and antigenotoxic activities of flavonoids and phenol isolated from Rhamnus alaternus roots and leaves. Patients and methods It was a retrospective study reporting a family collective poisoning which occurred in the region of Zaghouan in Tunisia in July 2015. All the members of the family ingested accidentally a traditional preparation of a plant "Oud El-Khir" in a juice prepared for a traditional marriage.
Results On 1st July 2015, a family composed of seventeen members (ten men and seven women) was admitted to the teaching emergency and intensive care department of the regional hospital of Zaghouan (Tunisia). All members of the family presented dizziness, weakness, anorexia and dyspnea. They reported the ingestion 10 h before of a juice in a traditional marriage. This juice was prepared using a plant called "Oud El-Khir" because of its capacity to be a lucky charm for the new married. The mean blood pressure was 130/60 mmHg. All members of the family experienced nausea, vomiting, anuria and hematuria. On physical examination, five members of the family had myalgia without other clinical signs. For all patients: cytological reports and sputum smear were negative (three times) for pulmonary tuberculosis. Hepatitis B and C serology were also negative. Chest x-ray was normal; Blood and urine culture were negative. In renal ultrasonography performed in five members of the family, there was a significant difference in kidney sizes and the corticomedullary differentiation was altered. Laboratory tests showed hyperglycemia and renal failure with metabolic acidosis in ten patients. Three dialysis sessions were performed. Samples of the herbal decoction were obtained from the juice. It was a dark brown suspension with fine brown deposit and a clear supernatant. It smelled a strong penetrating odor. Samples of both Rhamnus alaternus roots and its decoction were sent to be analyzed in the laboratory of toxicology in the Center for Emergency Medical Assistance of Tunis in Tunisia. Screening by GC-MS of both Rhamnus alaternus roots and infusion extracts, revealed the presence of anthraquinone glycosides such as 4,5-dihydroxy-9,10-dioxoanthracène-2-carboxylic acid (Rhein), 1,8-dihydroxy-3-(hydroxymethyl)-9,10-anthracenedione (Aloe-emodin) and 1,8-dihydroxy-3-methoxy-6-methylanthracene-9,10-dione (Physcion). The retention times were 8.95, 9.67 and 10.25 min respectively. Anthraquinone glycosides were detected in a dichloromethane extract and ethyl acetate extract at pH value = 9 and only in a dichloromethane extract at pH value = 7 by GC-MS analysis. Conclusion Rhamnus alaternus can be toxic when used in an abusive way besides its strong antibacterial, antioxidants and antidiabetic activities. To our knowledge, this is the first report of cases of renal failure and rhabdomyolysis which is possibly associated with an accidental consumption of Rhamnus alaternus roots. We present these cases to illustrate the role of both clinical and biological investigations in handling cases of herbal poisonings. We aimed also to increase awareness among emergency physicians about patients presenting to the emergency department with unexplained symptoms (renal failure, rhabdomyolysis…) that requires prompt diagnosis so that such lifethreatening complications can be avoided.
Introduction Lithium overdose may result in encephalopathy and electroencephalographic abnormalities. Three poisoning patterns have been identified based on the ingested dose, previous treatment duration and renal function. Whether severity of lithium-induced encephalopathy depends on the poisoning pattern is not established.

Materials and methods
We designed a rat study to investigate lithium-induced encephalopathy and correlate its severity to plasma, erythrocyte, cerebrospinal fluid and brain lithium concentrations previously determined in rat models mimicking human poisoning patterns. Lithium-induced encephalopathy was assessed and scored using continuous electroencephalography.

Results
We demonstrated that lithium overdose was consistently responsible for encephalopathy which severity depended on the poisoning pattern. Acutely poisoned rats developed rapid-onset encephalopathy which reached a maximal grade of 2/5 at 6 h and disappeared at 24 h post-injection. Acute-on-chronically poisoned rats developed persistent and slightly fluctuating encephalopathy which reached a maximal grade of 3/5. Chronically poisoned rats developed rapid-onset but gradually increasing life-threatening encephalopathy which reached a maximal grade of 4/5. None of the acutely, 20% of the acute-on-chronically and 57% of the chronically lithium-poisoned rats developed seizures. The relationships between encephalopathy severity and lithium concentrations fitted a sigmoidal Emax model based on cerebrospinal fluid concentrations in the acute poisoning and brain concentrations in the acute-on-chronic poisoning. In the chronic poisoning, encephalopathy worsening paralleled the increase in plasma lithium concentrations. Conclusion Severity of lithium-induced encephalopathy is dependent on the poisoning pattern, previously shown to determine the lithium amount accumulated in the brain. Our data supports that electroencephalography is a sensitive tool to score lithium-related neurotoxicity.
Introduction Amitriptyline is one of the common drug poisoning that induces severe cardiovascular and neurological complications. The incidence of electrocardiogram changes in Amitriptyline poisoning was not explored enough. Therefore, we conducted this study to determinate the incidence of electrocardiogram abnormalities in Amitriptyline poisoning. Patients and methods It was a retrospective study conducted from January 2012 to July 2016 in a toxicological unit including all admitted patients for Amitriptyline poisoning. A 12-lead electrocardiogram was carried out and analyzed at admission and every 6 h. Results One hundred and fifty patients aged 29 ± 12 years were included; their sex ratio was of 0.19. Among them, only 6 had a history of cardiovascular disease. A loss of consciousness was noted in 58% of patients (n = 87) and electrocardiogram changes in 42% of patients (n = 63). The most common electrocardiogram changes were sinus tachycardia (108 ± 13 beats/min) (60%), widening of the QRS complex (16%) and right bundle ranch block (12%). A changes in die S-T was observed in 4 patients and dominant secondary R wave (R' > 3 mm) in aVR in 3. Forty-four percent of patients with electrocardiogram changes were comatose. No patient had developed hemodynamic instability especially in the case of widening of the QRS. These electrocardiogram abnormalities took 6 to 36 h to regress without specific treatment.
Conclusion Although electrocardiogram changes are common in Amitriptyline poisoning particularly in a patients with loss of consciousness, it rarely induced hemodynamic instability.
naloxone to reverse tramadol-related effects is debated. We aimed at investigating the pathways involved in tramadol-related neurotoxicity and seizures in the rat, by using various antagonists of the different tramadol-mediated effects including naloxone, cyproheptadine, fexofenadine and diazepam and determining the turnover of brain monoamines. Materials and methods Body temperature (using telemetry), respiratory effects (using plethysmography) and neurological effects (using clinical scales and EEG) were studied. Brain (frontal lobes) monoamines (serotonin, dopamine and norepinephrine) and their respective metabolites were measured using HPLC coupled to fluorimetry. For each animal and each time, we calculated the difference between the parameter value at that time and baseline and the area under the curve of its time course. Comparisons were performed using two-way ANOVA followed by post-tests using Bonferroni correction.
Results Tramadol induced sedation (p < 0.01), seizures (early onset and peaking at 30 min) and increase in inspiratory time (p < 0.001) as well as a non-significant trend to hypothermia,. Diazepam completely suppressed seizures Naloxone prevented tramadol-related sedation and respiratory effects while did not inhibit seizures. In contrast to cyproheptadine which exhibited no effects, fexofenadine partially reduced seizures, suggesting the involvement of a histaminergic pathway. Monoamines turnovers were significantly reduced in the presence of diazepam (p < 0.01), suggesting that diazepammediated prevention of tramadol-induced seizures could be related to the inhibition of monoamines metabolism in addition to its usual GABAergic effects. Conclusion Tramadol-induced sedation and respiratory effects are mediated by mu-opioid receptors. Seizures involve complex mechanisms including histaminergic but not serotoninergic pathways. Diazepam-related anticonvulsive activity to prevent tramadol-induced seizures may be related to the inhibition of monoamines metabolism in addition to its GABAergic effects.
Introduction Suicide by injecting insulin is not uncommon in diabetics and non-diabetics but few series were described in literature. Our study aimed to describe insulin intoxication in a series of cases.

Materials and methods
Our study is retrospective. It included all patients admitted for insulin intoxication in the period between 2012 and 2015. Insulin poising was withheld on medical history of insulin injection without therapeutic indication, on clinical signs of hypoglycemia (sweats, neurologic and digestive signs…), and on capillary glucose under 0.5 g/L in non-diabetics and under 0.6 g/L in diabetics, at least once. Results Forty-three patients were collected, with sex-ratio at 0. Introduction Guillain-Barré syndrome (GBS) is an inflammatory disorder of the peripheral nervous system is idiopathic acute polyneuropathy, which has now become the most common cause of acute flaccid paralysis in children.
Patients and methods This is a retrospective study which runs from 2006 to 2015 performed in the pediatric reanimation Results Guillain-Barré syndrome was diagnosed in 50 patients aged 9-15 years, the predominant involvement in sex with male enrollment ratio of 1.7 men/women After a prodromal event, usually infectious (96%) and a free interval of 15 days on average, start motor disorders. These are of two types: Consider a hypo or aref lectique flaccid paralysis of the lower limbs (25%) of ascending evolution in 91.4% of cases. Let flaccid tetraplegia or hypo areflectique (75%) The respiratory reached was observed in 64% of cases also other serious signs such as swallowing disorders (70%) and autonomic disturbances (14%) were also observed which justified a decision support in intensive care for all our patients. The use of ventilation was required in 68% of cases and specific treatments based on immunoglobulins were administered in 88% of cases The death rate is still high (22%) and the majority of complications related to hospital. The study of these patients has identified some prognostic factors of the intensive care unit in disease as male gender, duration of administration of Ig, the occurrence of autonomic disorders such unstable blood pressure which remains the most discriminating the occurrence of nosocomial infection. Conclusion Guillain-Barré syndrome therefore a pediatric emergency that requires rapid diagnosis and immediate changes in severity criteria for the establishment of appropriate treatment.
Introduction Purpura fulminans (PF) is a major cause of mortality and morbidity in children. Despite therapeutic advances in the management of PF, mortality remains high up rates that exceed 50%. In Tunisia, the profile of such serious disease has not yet been described. The aim of this study was to determine the predictors of mortality in patients admitted in pediatric intensive care unit (PICU) with purpura fulminans. Patients and methods Retrospective review of case sheets was done. Sixty-nine children [median age 3 years (1.6 months to 11 years and boy to girl ratio 1.15) admitted between January 2000 and May 2015 with PF in PICU of the pediatric hospital Bechir Hamza in Tunis, were included. The diagnosis of PF was made in patients with severe sepsis or septic shock with an extensive purpura. The PF is considered secondary to meningococcal infection if Neisseria meningitidis or soluble antigens are found at the blood culture or cerebrospinal fluid. In patients whose samples are negative, the PF is considered secondary to meningococcal infection if no other bacterial or viral origin is found to explain the purpura. For each patient enrolled, we have clarified the demographic data, the severity of clinical presentation by pediatric risk score of mortality (PRISM) and Glasgow meningococcal septicemia prognostic score (GMSPS) and therapeutic data. Results A mortality rate of 52% was observed for a predicted death by the PRISM score at 24.6%. Seventy-seven percent of the deaths occurred during the first 24 h. Twenty-five (69.4%) children died of irreversible septic shock and 6 (16.6%) children died of refractory hypoxemiae. Independent predictors of mortality were the initial severity assessed by the GMSPS (p = 0.001) and use of high doses of vasoactive drugs evaluated by the vasoactive-inotropic score (p = 0.026). Conclusion The mortality in our PICU is high, higher than predicted by PRISM. It occurred mainly within 24 h of admission. Early recognition and prompt initial antibiotic therapy continue to be the cornerstones of the successful management of this dramatic disease, reducing mortality.

Competing interests
None.
Introduction S. marcescens, a gram negative bacillus, classified as an Enterobacteriaceae, was originally considered to be a non-pathogenic saprophytic water organism. Since several years, it is a well-recognised nosocomial pathogen involved in outbreaks in neonatal intensive care units [1]. Patients and methods This study describes two consecutive outbreaks of S. marcescens clones in a mixed neonatal and paediatric intensive care unit and discusses its investigation and control. Results Between October 2015 and February 2016, seven preterm new-borns (gestational age: 25-34 weeks) were identified with S. marcescens in our 15 beds unit, spread over 12 rooms. Four children presented severe infections (four pneumonias, one sepsis and one conjunctivitis). S. marcescens was isolated in the stools for three others. From May to June 2016, three new cases (gestational age: 26-34 weeks) were detected (two pneumonias and one lung colonisation). All infected new-borns have favourable clinical outcome.
A single S. marcescens strain was identified during each outbreak. The two strains were genetically unrelated and have a natural phenotype of resistance. In order to contain the outbreak, active surveillance was established to detect infected and/or colonised patients. Strict cohort nursing of those patients was implemented and admissions in the unit were limited. Health care workers of the unit were requested to reinforce infection control measures. Audit programs performed by the infection prevention team showed the necessity to reinforce meticulous hand hygiene using hydro alcoholic solution and adequate gloves wearing. Sixty environmental samples (water, room surfaces, medical devices…) failed to identify S. marcescens reservoirs. Actually, the current strategy of empiric antibiotic treatment is being reevaluated by a multidisciplinary group composed of a medical microbiologist, an infectious disease specialist and a paediatric intensive care clinician. Furthermore, a case-control study is being performed to assess the risk factors for acquiring S. marcescens in a mixed neonatal and paediatric intensive care unit. The cause analysis also highlighted that: • due to the mixed character of our ward, patients frequently changed rooms during their hospitalisation in order to increase the admission capacity. However, this policy may increase infectious risks. • the first outbreak was detected late when three infants with S. marcescens were identified. During the second event, infection control measures were applied after detection of two cases. This early intervention probably limited the spread. No new case was detected since June 2016.
Conclusion S. marcescens can cause rapidly-spreading outbreaks of severe infections in neonatal units, but with appropriate control measures these outbreaks can be contained at an early stage.
Introduction In order to limit the emergence of antimicrobial resistance, it is recommended to re-evaluate every prescription of antimicrobial treatment within 72 h and to proceed to a de-escalation whenever it is possible [1]. However, evaluations of such recommendations are scarce, especially in paediatrics. Our goal was to evaluate the proportion of modification of the antimicrobial treatment within the first 72 h in our PICU and to determinate whether it was relevant or not. Conclusion Antimicrobial treatment adjustments are globally well executed in our paediatric intensive care unit, but could be improved by obtaining expertise from an infectious diseases specialist more frequently. Hence antibiotic de-escalation should be improved. The results of this study constitute an adequate basis for a future antimicrobial stewardship that could be conducted in our unit while focusing on a systematic consultation with a paediatric infectious disease specialist.

Materials and methods
Introduction Development of an infection in a critically ill patient is a gravity element associated with increased mortality. Elements related to the patient and his environment used to target the empiric antibiotic therapy to bacterial flora community or nosocomial kind. where interest of a prescription adapted to prevent the emergence of resistant bacteria. This problem could be better controlled by optimizing antibiotic prescriptions also taking into account the economic consequences.
Introduction Continuous assessment of nociception intensity during anesthesia is crucial. Analgesia nociception index (ANI: 0-100) has been proposed to clinicians to objectively measure pain in both conscious and anesthetized patient. ANI calculation results from sophisticated computing of sympathetic to parasympathetic tones balance using heart and respiration rate variability. Because we aimed to evaluate nociception during oro-tracheal intubation performed under apnea conditions, we conducted a study to test the reliability of ANI at monitoring nociception when respiratory rate was zero. Materials and methods ASA physical status I healthy informed and consenting adult volunteers participated in this experimental prospective trial. Demographic parameters were recorded and noninvasive hemodynamics (HD), blood pressure (BP), heart rate (HR), pulsed oxygen saturation (SpO 2 ) and ANI monitoring tools were installed. After a 5 min resting and stabilization the volunteers lying on their back were asked to perform a 30 s duration apnea starting on command at Residual Functional Capacity (RCF). After a wash-out period following this first apnea period (P1) and return to baseline parameters, the volunteers were asked again to perform a second 30 s duration apnea (P2) starting on command at RCF. Initiation of P2 coincided with 30 s ulnar nerve stimulation at the wrist (single twitch, 1 Hz, 25 mA). The different component of ANI (energy, mean, and instantaneous values: iANI), HR, BP and SpO 2 values were recorded at predefined time points: T0 (baseline values, just before apnea), T1 (end of apnea), T2 (T1 + 30 s), for P1, and the same predefined time points (T0′, T1′, T2′) for P2. After the experiment, volunteers were asked to rate the pain intensity at the wrist during P2, using a Visual Analogue Scale (VAS: 0-100). Evolutions of measured parameters were compared between P1 and P2. Values are mean ± SD. Results 21 healthy volunteers aged 34 ± 10 were included. During P1 and P2, both iANI and energy were monitored. Mean VAS during P2 was 27/100. Figure 23 illustrates HD and iANI variations at T1/T1′ and T2/T2′. During P1, there is a remarkable stability of iANI at T1 (+1%) and no significant decrease at T2 (−3%). On the opposite, a progressive decrease of iANI was evidenced during P2 (−5% at T1′), reaching a nadir 30 s after the end of nociceptive stimulation (−17% at T2′). HD and Sp02 remained stable and comparable in P1 and P2. Conclusion Based upon our results, ANI seems discriminant and efficient at measuring nociception in the conditions of a short apnea of 30 s. ANI is thus probably a reliable tool to measure nociception during tracheal intubation. Our trial shows that HD is less sensitive than ANI to detect weak intensity nociceptive stimulus.
feasible, well tolerated and beneficial. We aimed to assess the feasibility and tolerance of early mobilization as part of routine care in a Tunisian medical ICU. Patients and methods A prospective period study conducted in a 7-bed-medical ICU with nurse/patient ratio at 1/3 over 12 months.
Were studied, incidence density of four predefined activity events (sit on bed, sit in chair, stand-up and ambulate), onset delay, limitation factors and tolerance. Six activity-related adverse events were defined as fall to knees, catheters or tubes removal, systolic pressure drops more than 30 mmHg, pulse increase by 20 beats/mn or bradycardia, SpO 2 drop <80% and extubation.
Results During the study period, we conducted a total of 1265 activity events in 127 patients. The mean length of stay was 11.8 ± 12.3 days and the global incidence density was 842 activities/1000 patient days.  [1]. Sricharoenchai in an observational study conducted in Baltimore ICU over 30-month period including 5267 activities showed a low incidence of early mobilization adverse effects (0.6%) [2]. Conclusion This study shows the feasibility and safety of an uncommon practice in a relatively poor resources and low patientnurse ratio medical ICU.
Introduction Traditionally, in French Hospitals, the team composition is different according of the moment: the daily team for day from Monday to Friday, the watch team for night, week end, and public holiday. It appears to be important for us to evaluate the workload in the different configurations. We need to know the way our ICU team is needed outside the ICU in order to improve the quality of the team, in order to anticipate the workload during different moments and in order to avoid reducing the team size.

Results
We collected 469 consultants. The average age was 43.45 years (16-104 years) for a sex ratio of 0.87. The main reasons were an oropharyngeal infection or bronchopulmonary (17.3%), asthenia (13.4%), chest pain (11.9%). More than 50% of the consultations were deemed inappropriate. These patients were younger than 35 years in 48.4% of cases, and jobless in 58.6% of cases. There was no significant difference in the number of inappropriate consultations between the two genders. The days and times of arrival did not differ between appropriate consultation and not appropriate. After review, 6% of patients were admitted to unit short hospitalization, 49.9% referred to a specialist consultation, other consultants put outgoing, equipped (39.7%) or not (4.5%) of a therapeutic prescription.CV Discussion The use of emergency would be linked to the absence of prior appointment, and continuity of care. The consultants have the option of availability, some for conditions not covered by the emergency. The high proportion of younger people without profession is explained by the free care. The elderly rarely autonomous and complex diseases carriers, are fewer in the "inadequate consultation" group. Conclusion Develop a care system and quality of care in basic health facilities, educate and convince patients to get there, would relieve hospital emergency services that will then take care of true emergencies. Productivity gains are necessary to cope the tariff evolution and may require reduction of costs, increase of activity and pricing optimization.

Introduction
Introduction Hemorrhagic complications during placenta previa are unpredictable and may indicate an extremely urgent C section. Our study goal is to determine the risk factors and prognosis of this urgent situation. Patients and methods This is a case-control study, prospective and analytic over the second semester of 2015. We included all patients who underwent emergency C section for placenta previa bleeding. The witness cases are represented by C section performed in emergency for another reason, at random.

Competing interests
None. highly retained, and depending on severity of patients and on intensivist's experience.
Conclusion TCD is performed in heterogeneous diseases in nontraumatic critically ill children. TCD has an important impact, dependent on the operator's experience and on patient's severity.
Regarding the difficulty of accurate interpretation of TCD's values, improvement of TCD's use is necessary.

Introduction
The sublingual microcirculation is impaired in states of severe sepsis and septic shocks. The microcirculatory dysfunctions are prognostic markers of survival and organ failures in critically ill adult patients. In children, the sublingual microcirculation is also impaired in septic shock [1] and severe respiratory failures [2]. We hypothesized that the microcirculation impairments related to the severity of respiratory and circulatory failures could be reversible under extracorporeal membrane oxygenation (ECMO). The aims of the study are (1) 3). The partial MFI restoration confirmed the microcirculation recruitment in these patients. Now, further studies are needed to confirm this microcirculation's recruitment in a larger pediatric ECMO cohort and to investigate the impact of this recruitment on the prognosis.

Conclusion
The microcirculation was impaired in the initial phase of ECMO in children with severe respiratory or circulatory failures and it was partially restored during the early phase of ECMO.
Introduction Impedance cardiography (IC) is a promising non-invasive, continuous cardiac output (CO) monitoring method. PhysioFlow ® (PF ® ) is a new IC technique never studied in pediatric intensive care unit (PICU). The aim of the study was to compare CO and cardiac index (CI) measurements in PICU patients by IC using PF ® with those obtained by the method of reference, trans-thoracic Doppler echocardiography (TTE), with regard to accuracy, precision of agreement and reproducibility.

Patients and methods
In this single-center prospective method comparison study, all PICU patients aged between 28 days and 10-year-old and requiring TTE were included, except those with complex congenital heart disease or poor IC signal and/or TTE quality. Simultaneous sets of three measurements were realized by TTE and PF ® . CO and CI measured by TTE (COTTE and CITTE) were compared with CO and CI obtained by PF ® (COIC and CIIC). Concordance correlation coefficient (CCC) and Bland-Altman analysis were used to analyze the concordance rate and compare accuracy and percentage error (PE) between the two methods. As required, data were logarithmically transformed prior to Bland-Altman analysis. Reproducibility was evaluated with intraclass correlation coefficient (ICC) and using the calculation of the precision of the method [1].  (2) an interaction between NO and prostaglandin synthesis via COX-2 in the juvenile P14 ischemic rat. Increasing evidences suggest a "cross talk" between NO and Pgs biosynthetic pathways but still remains to be better elucidated, namely during brain development.
Conclusion These results show that the juvenile rat brains mostly responds to ischemia by a COX-2-dependent prostaglandins production, and suggest that the transcriptional responses observed under 7-NI facilitate and reorient COX-2-dependent prostaglandins production.
Introduction Congenital diaphragmatic hernia is a rare anomaly that occurs in <1-5 of every 10,000 births. Pulmonary hypoplasia and pulmonary hypertension are commonly associated with congenital diaphragmatic hernia and explain its high morbidity and mortality. Despite significant advances in neonatal intensive care and protective ventilation, mortality remains as high as 50%. The development of extracorporeal membrane oxygenation (ECMO) support has improved the outcome but its use is controversial in this pathology.
The main objective of this study is to determine associated factors with extracorporeal membrane oxygenation (ECMO) in congenital diaphragmatic hernia and to explore ECMO's indications, causes of death, and evolution after discharge for survivors. Introduction Administration of fluid to increase cardiac output is a cornerstone of the hemodynamic resuscitation. The respiratory variation in aortic blood flow peak velocity (ΔVpeak) is the only variable to reliably predict fluid responsiveness in children. It requires assist control ventilation with high tidal volume (>10 ml/kg) [1]. The purpose of this study was to evaluate the clinical usefulness of assessing variation of stroke volume during a calibrated abdominal compression for the prediction of fluid responsiveness in children during acute circulatory failure. Patients and methods This study was approved by local institutional review boards (CPP Lyon Sud Est II). Written informed consent was obtained from parents. Oral consent of patients old enough to understand the study was also obtained. Patient, less than 8 years old of two paediatric intensive care unit (PICU), during a 1-year period inclusion, from September 2015 to September 2016 were selected. Stroke volume index was assessed with an echocardiography at baseline, after an abdominal compression (with a calibrated pressure of 25 mm Hg), at a return to baseline, and after a volume expansion (10 ml/kg of fluid challenge over 10 min). Pulse pressure (PP) Systolic (SAP), diastolic (DAP), and mean arterial pressure (MAP); heart rate (HR), respiratory aortic blood flow velocity (ΔVpeak), left ventricular ejection fraction (LVEF), and respiratory vena cava diameter variation (VCIvar) were also recorded before volume expansion. . With the multivariate analysis, older age and male sex were associated with an increased risk for anaphylaxis death of any cause (p < 10-4).

Patients and methods
Conclusion Fatal anaphylaxis decreased in France between 1979 and 2011. Higher rates of fatal anaphylaxis are observed in male sex, group aged 70 years and older and iatrogenic cause. In pediatric population, fatal anaphylaxis is low, but it's probably under-estimated by diagnostic difficulties.

Introduction
The burn out is an emotional state of exhaustion and loss of performance in response to chronic work stress. This study aims to assess the prevalence and determine the Burn out of risk factors in medical and paramedical staff in the anesthesiology department.

Patients and methods
This was a multicenter, transversal analytic study performed within the medical and paramedical staff of the operation rooms in the central anesthesiology department in IBN ROCHD hospital in CASABLANCA. Caregivers were asked to freely and anonymously complete a questionnaire involving demographic variables, professional and an assessment of the causes and consequences of stress at work. Results On a population composed of 60 caregivers, 41.7% had a score of high emotional exhaustion, 11.98% a high score of depersonalization and 47.38% a low score of professional achievement. Le burnout was found at 62.52% of our caregivers: 39.62% of them had low levels of burnout; 21.34% presented a moderate level and 1.58% had the highest level of burn out. In multivariate analysis, Doctors and Nurses Residents were most at risk of burn out. Fear of malpractice and the unsatisfactory salary multiplied the risk of burnout, respectively 2.14 and 1.83. Conclusion Burn out is a threatning reality of the anesthesiology environment. The consequences could be severe on the personal and professional performance of the health institution, which involves the implementation of preventive strategies highlighting the value of work organization and the contribution of the and paramedical medical staff.

Competing interests
None.
Introduction Evaluation of family satisfaction has become an essential part of the responsibility of intensivists and a major criterion for assessing the quality of care. Family satisfaction could depend on  Fig 26).
Conclusion The transfer to a new ICU building without implementing any new policy is associated with a higher a higher satisfaction of families. A communication strategy could help improving further family satisfaction. Results Records of these women were studied. The characteristics of women are the same for both populations, but there is a net reduction of maternal mortality 0-4 and the satisfaction rate increased from 31 to 71%. The features of management are summarized (see Table 18). Discussion Our institution demand for unscheduled care and frequency of emergencies are increasing in the service of emergency obstetric gynecological similar phenomenon reported in some publications. Our study marks a reduced waiting time in half, as in a study published in France [2]. Despite the protocol established for the second population, we note a major impact for the average time of care, and the average time to obtain blood components is as long, related organizational deficits at night and on weekends, but this has allowed us to reduce the mortality rate and duration so stay on the economic cost.

Conclusion
The human factor has a direct impact on the length of hospitalization and thus the economic cost, on maternal mortality, an indirect impact on satisfaction patient. Il is time to realize that the human factor is the success factor.

Introduction
In the absence of specific academic training in intensive care for paramedical personnel and in order to improve skills, training is necessary especially before the opening of a new unit. The aim of our study was to evaluate the training of newly recruited in intensive care paramedical staff. Patients and methods For initiation of a new intensive care unit, medical staff prepared a specific training program of resuscitation. The schedule of the training has been distributed to all relevant staff in advance. Presentations (basics and technical data sheets) on various resuscitation themes were prepared by the staff and reviewed by doctors' leaders. The evaluation form contained 16 tests (T). Each test consisted of five questions before the presentation and five after. An increase in the evaluation mark obtained after the presentation is considered an improvement in knowledge.
Results Twenty of the staff: twelve nurses, four technicians in intensive care and four healthcare assistant participated in the training. Fiftyfive percent of the staff were working in the private sector before their recruitment. The median time between graduation and recruitment was 1 year and a half. Technicians group had the highest level of knowledge (Table 19). The analysis of the post-formative evaluation noted an increase in knowledge. The healthcare assistant group had the lowest rate of improvement (Table 20). Conclusion The direct involvement of paramedical staff in their own training increases their motivation without harming the quality and the main objective: knowledge improvement.
Results Twenty of the paramedical staff: twelve nurses, four technicians in intensive care and four healthcare assistants participated in the study. Only one questionnaire has not been analyzed, it belonged to a nurse presenting a serious mood disorder. Forty-five percent of the study population was unemployed before their recruitment. Eight individuals (40%) showed burnout the first day of work, 11 (55%) at the end of training and 12 (60%) after 3 months of training (Table 21).
Regarding the sub-dimensions of burnout there was an improvement of professional fulfillment (increasing from 35 to 45% after training and practicum). We also noticed a worsening of emotional distress (increasing from 0 to 25%). There was also an improvement in mood disorders (decreasing of 35-15%). Conclusion Theoretical and practical courses enhanced the level of professional achievement but were not sufficient alone to prevent Burnout.
Introduction Procalcitonin (PCT) is a pro-hormone mainly produced by the thyroid C cells and routinely used as diagnostic biomarker of bacterial infection. PCT synthesis has never been described in the liver. To our best knowledge, no study investigated the possible predictive value of PCT in acetaminophen poisoning. Our objectives were to report the distribution of serum PCT values in acetaminophenpoisoned patients according to the onset and severity of their toxic liver injury in order to assess any possible predictive value for this biomarker. The delay between acetaminophen ingestion and N-acetylcysteine infusion was 4.5 h [2.9; 9.0]. Serum PCT was markedly increased above the 1 μg/L threshold in the patients who already presented or further developed significant liver cytolysis defined by serum alanine aminotransferase (ALAT) >100 UI/l (2 N) despite the treatment with N-acetylcysteine and independently of the onset of any bacterial infection, with a specificity of 97.9% and sensitivity of 69.6%. Conclusion Serum PCT measurement in acetaminophen-poisoned patients admitted to the ICU is helpful to early identify patients who present significant acetaminophen-related liver toxicity already established or in progress on admission despite the administration of N-acetylcysteine according to the international recommendations.

Patients and methods
Introduction Venlafaxine, an antidepressant drug with properties of serotonin and norepinephrine reuptake inhibition, may be responsible in overdose for life-threatening cardiovascular complications. Venlafaxin-related toxicity additionally includes a possible interaction at elevated concentrations with the membrane sodium channels resulting in membrane stabilizing effects. To date, vulnerability factors to develop such cardiovascular complications are unknown. Based on a limited number of reported cases, poor cytochrome P450 (CYP) 2D6 metabolizers have been suggested to develop increased cardiovascular toxicity. This liver enzyme metabolizes venlafaxine to O-desmethyl-venlafaxine (or norvenlafaxine) and altered norvenlafaxine-to-venlafaxine metabolic ratio was suggested to support cardiotoxicity onset. We aimed to describe venlafaxine-related toxicity in patients admitted to the intensive care unit (ICU) and test the proposed hypothesis of vulnerability.

Patients and methods
We conducted a prospective single centre observational study including all venlafaxine-poisoned patients admitted to the ICU from 2010 to 2016. Plasma venlafaxine and norvenlafaxine concentrations were determined using gas chromatography coupled to nitrogen-phosphorus detector (GC-NPD) after initial detection using high-performance liquid chromatography coupled to diode array detector and mass spectrometer (LC-DAD/ MS  and fatality (p = 0.04). No statistical relationships were found between cardiovascular toxicity and plasma venlafaxine concentration and norvenlafaxine-to-venlafaxine ratio on admission and at their respective peaks. When focusing on the patients with cardiovascular manifestations strictly attributed to venlafaxine toxicity, no statistical link was found with CYP2D6 phenotype. Conclusion Venlafaxine poisoning may result in severe complications including cardiovascular toxicity and even fatality. Cardiac toxicity is responsible for increased morbi-mortality but is not related to CYP2D6 phenotype. However, inclusion of additional patients is still warranted in our possibly underpowered study before any definitive conclusion.
Introduction Cyproheptadine (Periactin ® ) is a histamine H1 receptor and serotonin antagonist, it has also anticholinergic effects. It is used to treat allergic type symptoms and as an appetite stimulator. Cyproheptadine poisoning has been rarely reported. That's why we conduct this study, to identify the epidemiological characteristics and clinical outcomes of cyproheptadine poisoning and to define the association between the ingested dose and clinical manifestations. Discussion Children in pre-school age are often at risk of domestic accidents in relation to their age and environment. They ingest products wrongly deconditioned or transferred into current use containers. These by catch happen to be in the kitchen at a large household or in the bathroom. They also ingest drugs for adults and thus higher dosage: these drugs are too easily left to reach those children who are not suspected faculties to identify seize and swallow what they take to candy. A study in India positioned intoxication hydrocarbons 1st place (50.9%), followed by organophosphates (11.9%), medication and caustic identify (4.8%) cases each [1]. In Oran (Algeria) taking caustic remains at the level of poisoning in children. Overall mortality by poisoning is lower in children compared to adults. (The products in question are heterogeneous drugs, household products, plants, pesticides) We recorded 01 death secondary to acute respiratory distress syndrome following ingestion by siphoning gasoline. Conclusion At the end of our study, it appears that the acute poisoning in children is a very common event. However, developments remain favorable in most cases. This should not obscure the potential severity of the poisoning, nor lose sight of the necessary prophylactic measures. This prophylaxis should include education of the population, prevention of certain tragedies related to toxins or aberrant behavior, and triggering alerts to reduce morbidity and mortality related to poisoning. Our ACCIPED laboratory medical university Oran for a future mission secondary and tertiary prevention through continuous training for general practitioners, to respond to any request for advice on diagnosis and treatment of these accidents, assess risk through telemedicine and through a website continuously 24 h/24 h. The breathing patterns of every subject seems to influence the existence of inspiratory leaks. Some volunteers had more inspiratory leaks than others, and experienced a longer length of preoxygenation than others. This is probably explained by a different instantaneous inspiratory flow [2]. These results hypothesize that individual preoxygenation is needed for each subject. Conclusion Noninvasive ventilation allows effective preoxygenation despite the existence of leak.

Patients and methods
Introduction Manikin-based studies for evaluation of ventilation performance show high heterogeneity in the analysis and experimental methods used as we pointed out in previous studies. In this work, we aim to evaluate these potential limitations and propose a new analysis methodology to reliably assess ventilation performance. Patients and methods One hundred forty healthcare providers were selected to ventilate a manikin with two adult self-inflating bags in random order. Ventilation parameters were analysed using different published analysis methods compared to ours. Results Using different methods impact the evaluation of ventilation efficiency which ranges from 0% to 45.71%. Our new method proved to be relevant and showed that all professionals tend to hyperventilate and revealed a significant relationship between professional category, grip strength of the hand keeping the mask and ventilation performance (p = 0.0049 and p = 0.0297 respectively). Discussion Using adequate analysis methods is crucial to avoid many biases. Extrapolations to humans still have to be taken with caution as many factors impact the evaluation of ventilation performance. Healthcare professionals tend to hyperventilate with current devices.

Conclusion
We believe that problems related to manual ventilation efficiency could be prevented by implementing monitoring tools in order to give a direct feedback to healthcare professionals regarding ventilation efficiency and ventilatory parameter values.
Introduction Basic and advanced airway management as Bag-Valve-Mask (BVM) and Endotracheal Intubation (ETI) are procedures used in prehospital resuscitation. In this work, we compare manual ventilation performance with both techniques and analyse their variability and also try to assess what factors may affect their performance. Patients and methods One hundred forty healthcare providers were selected to ventilate an intubated and not-intubated manikin with two adult bags for 5 min each in a random order. Ventilation performance were analysed using a new analysis method we described in a previous study which takes into account ventilation variability.

Results
Five hundred sixty ventilation tests (280 for ventilation using a facemask and 280 for ventilation with an Endotracheal Tube (ETT)) were performed by the 140 healthcare professionals. Results show a significant difference between ventilation performance with a mask compared to ETT (p < 0.05) with more ventilation efficiency when healthcare professionals ventilate with an ETT than a mask (37 vs. 21 ventilation tests). Furthermore, a significant relationship is observed between participants' professional category, the size of the hand squeezing the bag and manual ventilation performance (p < 0.05). Discussion Healthcare professionals have performed 560 ventilation tests on an ETT and a facemask. Among them, only 58 (10.41%) were efficient and 502 (89.64%) are inadequate (of which 11.4% insufficient and 78.2% excessive). Whatever the kind of ventilation technique used, they are still struggling to perform manual ventilation efficiently according to international guidelines. Conclusion The high ventilation performance failure observed in this study shows that "to intubate or not to intubate the patient" is not the question, but we should focus on how to improve efficiency of manual ventilation performed by healthcare professionals.
Introduction Devices' ergonomics are major determinants of task failure, especially in stressful environments. The aim of the study was to provide a global evaluation of ICU ventilator ergonomics using novel exploration tools. Materials and methods Devices and physicians 6 new-generation ventilators were evaluated (Dräger Evita V300, Covidien PB980, Philips V680, Hamilton S1, GE Carescape R860, Maquet Servo-U) and compared to an old one (Carefusion Avea). 20 senior physicians were included, each testing up to 4 ventilators in a randomized order.
Objective task completion 11 specific tasks were to be completed. The test was a failure if the correct response was given ≥120 s, or in case of incorrect response or abandoned task User-friendliness evaluation System Usability Scale (SUS) was developed to measure device's usability. It has a range of 0-100, the highest score being the best value. Mental Workload evaluation (MWL) is an indicator for human-machine interface (HMI) comparison. 3 dimensions depending to user and 3 dimensions depending to HMI are subsequently weighed to obtain TLX result. The higher the TLX, the higher the MWL. Physiological measurement Pupil diameter modifications were assessed by eye-tracking (SMI ETG 1); heart, respiratory rate and thoracic volume variations were measured with a dedicated device (Hexoskin).
Results No users could set inspiratory flow on Servo U, and only 18% succeeded with S1. Servo U had the worst global results (failure rate = 42%) and Avea the best (failure rate = 13%) (p = 0.12). Among all ventilators, Avea had the best SUS and TLX values, and Philips V680 the worst (p < 0.05). Eye tracking, respiratory rate and tidal volume activation differed between ventilators (p < 0.05). V300 caused the higher eye-tracking activation rate when compared to Avea (p = 0.03) and R860 (p = 0.019) (Fig. 27).

Conclusion
Ergonomics evaluation is mandatory when evaluating new devices in the ICU. Most ICU ventilators presented poor HMI capabilities, thus allowing the occurrence of various hazards and failure.
Introduction Hypotension and oxygen desaturation are the most common critical events occurring in 17% of emergency transports. They are mainly related to ventilator issues and the type of ventilator might mitigate the rate of adverse events. We decided to compare the most suitable ventilators for out-of hospital transport. Materials and methods This experimental bench test was designed to evaluate the technical characteristics, ergonomics, accuracy of volume and pressure delivery, triggering performance, pressurization and depressurization capacity, patient asynchrony and leak compensation of transport ventilators. We simulated many patient profiles and conditions by adjusting lung compliances (30, 100 ml cmH 2 O −1 ) and resistances (5, 20, 50 cmH 2 O L −1 s), inspiratory efforts (2.5, 5, 10 cmH 2 O), and leakage levels (<1, 3, 6, 12 L min −1 ). The ventilators were used in pressure control, volume control, and pressure support ventilatory modes in non-invasive and invasive settings. Results Even if the technical characteristics of portable ventilators are quite similar, their ergonomics and performance are unequal. Major differences were found on tidal volume delivery, with mean relative errors ranging from +1.7 to −14.9%. Triggering delays and pressurization capacity also showed significant heterogeneity with a mean pressure time product (PTP) varying from 883 to 3018 cmH 2 O ms. The resistance of the expiratory circuit also differed across ventilators (from 0.96 to 1.82 cmH 2 O L −1 s), impacting the mean exhalation time and inducing air trapping and dynamic hyperinflation. Finally, while all ventilators were able to synchronize with patient's demand at baseline (leakage <1 L min −1 ), only one showed adequate patient synchrony at high leakage level.

Discussion
The main characteristics of last-generation ventilators are comparable, but major differences remain, especially in the conception of their basic ventilation modes. For instance, Hamilton T1 VC mode is actually a hidden pressure control mode that does not ensure Conclusion While it is still difficult to determine which ventilator best meets clinician expectations, this study will help to select the most appropriate device in regard to his own needs and make a clear comparison between ventilator performances.
Introduction Even though limitingtidal volume (TV) in ARDS patients is recommended, this goal may not be achieved once spontaneous breathing comes up and assisted modes are used. Furthermore ICU ventilators offer numerous assisted ventilation modes that work differently across the brands. We underwent present bench study to systematically investigate the effect of assisted mechanical modes on a single ICU ventilator on size and variability of TV at different breathing frequencies (f ), patient effort and ARDS severity.

Materials and methods
We performed a bench study in our university laboratory on an ICU ventilator (V500 Infinity, Dräger, Germany) using ASL 5000 lung model. Compliance was set at value mimicking mild, moderate and severe ARDS as recently reported. Thirteen assisted ventilation modes were tested falling into three categories, namely volume controlled ventilation with mandatory minute ventilation (VCV-MMV), pressure-controlled ventilation (PCV) including airway pressure release ventilation (APRV) and biphasic positive airway pressure (BPAP), and pressure support ventilation (PSV). fand effort were tested each at two levels for each ARDS severity in each mode. TV was expressed as median (first-third quartiles) and compared across modes using non parametric tests. The probability for TV > 6 ml/kg ideal body weight (IBW) was assessed by binomial regression and expressed as odds ratio (OR) with 95% confidence intervals (CI). The variability of TV was measured from the coefficient of variation.
Results The distribution of TV over all f, effort and ARDS categories significantly differed across modes (P < 0.001, Kruskal-Wallis test). TV was significantly greater with PSV (420 mL (332-527) than with any other mode except for the three modes accommodating a variable PS level. The risk for TV to be greater than 6 ml/kg IBW was significantly increased with spontaneous breaths assisted by PSV modes (for PSV OR 19.36; [12.37-30.65]) and significantly reduced in APRV (OR 0.44; [0.26-0.72]) and PSV with guaranteed volume mode. The risk increased with increasing effort and decreasing f. Coefficients of variation of TV were greater for low f and for VCV-MMV and PCV modes. APRV had the greatest within-mode variability. Conclusion The ventilation mode had an important impact on TV in this study. The risk of TV > 6 ml/kg IBW was significantly reduced in APRV and PSV with guaranteed volume mode. APRV had the highest variability. PSV with guaranteed volume could be tested in ARDS patients.
Introduction Single-Use Fiberscopes (SUF) have recently become available for ICU clinical practice to overcome 3 main problems associated with Re-Usable Fiberscopes (RUF) resulting in systematic: immediate availability of a sterile device, short lag time to endoscopic treatment and removal of the cross contamination risk. However endoscopy of the upper airway and lung (EAWL) is a medical act that generates significant costs. We thus conducted a medico-economical evaluation of RUF versus SUF. Patients and methods The study was conducted over 1 year (2015) in 3 French university hospitals. The medical evaluation was conducted in one of them while the cost analysis included the 3 hospitals. Medical evaluation was based upon a qualitative questionary filed by the clinician after each EAWL. Calculating the cost of RUF included the amount of money paid for the device acquisition and its equipment, and the cost related to both sanitary (systematic and post gesture) and technical maintenances. The cost of SUF was calculated based upon acquisition and the cost of disposable medical waste.
Results Medical evaluation performed upon 368 EAWL (60% SUF and 40 RUF) showed quality parameters either similar or better for SUF as compared to RUF except for one paramater: suction performance. Cost details of 922 EAWL were analyzed. Depending upon center organisations, cost per EAWL ranged between € 203 to € 240 for SUF, as compared to € 189 and € 269 for RUF. Saving related to SUF exclusive use would have ranged between € 2097 and € 7636 depending upon center's settings. Discussion SUF is efficient and performant for EAWL. Although costs associated with the use of SUF and RUF for ICU EAWL are close, selecting SUF induces substantial savings. The permanent and immediate availability of sterile equipment allowing to significantly shorten the lag time to treatment may have important clinical outcome for the patients. Conclusion At a time when health authorities are focusing their efforts on the prevention of nosocomial infections, elimination of endoscopes-related cross-contamination risk by using SUF, especially in ICU is a major issue. Improvement of Suction capability will promote clininal superiority of SUF over RUF.
operating room, where invasive blood pressure (BP), Bis-spectral (BIS) and Analgesia Nociception Index (ANI) monitoring were installed. The induction procedure of anesthesia with target-controlled infusion was standardized using sufentanil and propofol. Timed OTI maneuver was performed after monitored (corrugator supercillii muscle) deep neuromuscular blockade (atracurium) was installed. OTI maneuver which duration could not exceed 30 s was truncated in 3 periods (P) of predefined duration. P1 lasted 10 to 15 s and corresponded with direct laryngoscopy. P2 lasted 5 to 10 s and corresponded with tracheal tube (TT) manipulation. Finally, P3 coincided with TT cuff inflation to controlled 35 cm H 2 O lasted less than 5 s. Recorded HD, BIS and ANI data were analyzed at 6 predefined time points. T0: stable baseline values after monitors installation, T1: post induction of anesthesia and just before OTI, T2: end of P1, T3: end of P2, T4: end of P3, T5 and T6, respectively 1 and 5 min after the end of OTI. Recorded parameters evolution was compared using standard statistics. Results Thirty-nine patients were included in this trial, but 4 were secondarily excluded because of OTI lasting more than 30 s. Figure 28 shows mean (SD) variations of monitored parameters: mean arterial pressure (MAP), heart rate (HR), BIS, and ANI, from T0 to T6. P1 is characterized by a remarkable stability of all measured parameters. P2 and even more P3, are associated with intense and significant HD and ANI variations. ANI variations are completely superimposed, but opposite, to those of MAP and HR. BIS variations during TI were negligible. Conclusion Our data does not support the dogma but rather suggests that laryngoscopy may not be the most intense nociceptive stimulus applied to patients during OTI. Both laryngeal and/or tracheal stimulation resulting from tracheal tube passage and cuff inflation seem to promote very intense nociception. Interestingly, such phenomenon should be erased by either laryngeal mucous topicalization or superior laryngeal nerve block.
Introduction Low (<70%) mixed venous saturation (SvO 2 ) was one of the independent variable related to mortality in septic shock patients [1]. Nevertheless, values of central venous saturation (ScVO 2 ) may be considered as a surrogate to SvO 2 for early goal directed therapy in severe sepsis and septic shock. In sepsis, difference between ScVO 2 and SVO 2 (ΔSVO 2 ) was approximately of 5% [2]. This delta SVO 2 ignores the origin of sepsis: supra or infra-diaphragmatic where this gradient could be higher due to increased oxygen extraction by organs under diaphragm (gut, liver and kidney). We compared the correlation between ScVO 2 and SVO 2 , the time course of delta SVO 2 and effects of therapeutic challenge on delta SVO 2 between patients in severe sepsis or septic shock due to supra or infra diaphragmatic infection. Enter the text. Patients and methods We included patients in severe sepsis or septic shock monitored by Swan Ganz catheter and continuous ScVO 2 , hospitalized in a 24 bed medico-surgical ICU of CHU-Charleroi. Hemodynamic data and delta SVO 2 were recorded from catheter insertion to withdrawal and before and after each therapeutic intervention (fluid challenge, dobutamine initiation or dosage modifications). Demographic data and mortality were also collected. Graphpad prism was used for statistical analysis and graph, Bland-Altmann analysis was used and results were compared by Mann-Whitney U test when not specified.  )). Therefore, ΔSvO 2 was the same before and after interventions in the 2 groups of patients (p = 0.99) (Fig. 29).
Conclusion As in other studies, ScvO 2 overestimate SvO 2 by nearly 5.0% in patients with septic shock independently of the sepsis origin. ScVO 2 could be used as a surrogate of SVO 2 in the management of septic patients.

Introduction
To investigate whether respiratory variations of the inferior vena cava diameter (cIVC) assessed during a simple standardized breathing maneuver predict fluid responsiveness in spontaneously breathing patients with septic acute circulatory failure and irregular heartbeats.
Patients and methods This prospective, bicentric study, was performed in the intensive care units of a general and a teaching hospital. Spontaneously breathing patients with cardiac arrhythmia considered for volume expansion for clinical signs of acute circulatory failure related to sepsis were included. Echocardiography and Doppler ultrasound were used to record the stroke volume (SV) and IVC collapsibility index (cIVC) defined as [maximum expiratory diameter (edIVC) − minimum inspiratory diameter (idIVC)]/edIVC × 100 at baseline and after a 500 ml colloid infusion. Vena cava pertinent diameters were measured 15-20 mm caudal to the hepatic vein junction and recorded by bi-dimensional imaging on a subcostal long axis view. We measured the edIVC and idIVC during standardized (st) and unstandardized (ns) breathing, and calculated cIVCst and cIVCns before fluid loading. Standardized respiratory cycles consisted of a deep standardized inspiration followed by passive exhalation. A positive response to fluid loading was defined as an increase of the SV > 10%. Results Fifty-five patients were included, 29 (53%) with atrial fibrillation and 25 with atrial extrasystols >6/min. Twenty-nine (53%) patients were responders to fluid loading. A cIVCns > 19% predicts fluid responsiveness with a sensibility of 83%, a specificity of 68%, and an area under the ROC curve of 0.83 (95% CI 0.72-0.94). A cIVCst > 39% predicts fluid responsiveness with a sensibility of 93%, a specificity of 88%, and an area under the ROC curve of 0.93 [IC95: 0.86-1]. The area of uncertainty was restricted between 41 and 46%. Interestingly, idI-VCst < 12 mm predicts fluid responsiveness with a sensibility of 83%, a specificity of 88%, and an area under the ROC curve of 0.93 (95% CI 0.86-1). The area of uncertainty ranged between 10 and 13 mm. Discussion A simple standardized inspiratory maneuver improves the accuracy of cVCI to predict fluid responsiveness with high specificity and sensibility and a limited grey area in patients with cardiac arrhythmia. Moreover, idIVCst demonstrates very good performance to predict fluid responsiveness. This easy to acquire parameter may be more suitable for bedside patient's management. Conclusion cIVCst is a powerful index to predict fluid responsiveness in spontaneously breathing patients with sepsis and irregular heartbeats. Interestingly, idIVCst shows similar performance to predict fluid responsiveness with a much easier assessment, improving feasibility in clinical routine.
Introduction Acute pulmonary embolism (PE) is a common condition responsible for morbidities such as chronic pulmonary hypertension and deaths. The evaluation of PE severity is mainly based on arterial blood pressure and right ventricular abnormalities. According to recent ESC guidelines, patients with high risk of death had systolic hypotension and intermediate-high risk patients had right ventricular injury (defined by biological and echocardiographic alterations). However, the predictive value of tissue perfusion remains unknown. We aimed to investigate the relationship between 28-day mortality and tissue hypoperfusion parameters on patients admitted to intensive care units (ICU) for symptomatic acute PE.

Patients and methods
We conducted a retrospective observational study in a 18-bed ICU in a tertiary teaching hospital. From January 1993 to December 2015, all consecutive patients older than 18 years admitted for acute EP were included. Patients were identified by querying the electronic health records with the following keywords (in French): "pulmonary embolism", "anticoagulant", "deep venous thrombosis", "cardiac arrest", "heparin", "vitamin K antagonists" and "cava filter". General characteristics of the patients were recorded: demographics, biological data, diagnoses, severity of illness evaluated by the SOFA score and SAPS II. Global hemodynamic and tissue perfusion (arterial lactate level, skin mottling and urinary output) parameters were collected. We investigate the relationship between these parameters and 28-day mortality.
Results Over a 22-year period, we identified 317 patients admitted to our ICU with a diagnosis of acute PE. Of these 120 patients were excluded because PE was not the main diagnosis of ICU admission and 85 patients because of missing data, leaving 112 patients for study. At admission, the SOFA score was 2 [1][2][3][4][5][6] and the SAPS II was 31 . 31 patients (27%) were infused by vasopressors and 32 received mechanical ventilation support (28%). According to international guidelines, 49 patients (35%) were classified as high risk and 73 patients (65%) were classified as intermediate-high risk [1]. Thirty-nine patients (35%) received thrombolytic therapy. Day-28 mortality rate was 26%. At admission, when compared to 28-day survivors, non-survivor patients had higher lactate level (4.2 ± 5.9 vs. 1.5 ± 2.9 mmol/L, Mann-Whitney test P < 0.001), more frequent mottling around the knee area (50 vs. 24%, Chi squared test P = 0.02) and a lower urinary output (during the first 24 h) (0.5 ± 0.6 vs. 0.9 ± 0.8 mL/kg/h, Mann-Whitney test P = 0.02). Interestingly, 28-day mortality increased with the number of tissue perfusion alterations defined as hyperlactatemia (>2 mmol/L), oliguria (<0.5 mL/kg/h) and mottling presence. The mortality at day 28 increased from 10% for no failure to 22% for 1 alteration, 32% for 2 alterations and finally to 58% for 3 tissue perfusion alterations (P = 0.02, Chi squared for a trend test). Discussion Several studies in emergency and pneumology departments have investigated global hemodynamic parameters to predict outcome of PE patients. Low systolic blood pressure called "shock" and right ventricular dysfunction were identified as predictors of bad outcome. We were interested in the exploration of tissue perfusion because in septic shock several studies highlighted the discordance between global hemodynamic parameters and tissue perfusion. Moreover, during severe infections, hypoperfusion markers such as lactatemia, mottling or oliguria have been identified as predictive of ICU mortality. Recently, Vanni et al. [2] suggested lactate could be useful to identify normotensive PE patients at high risk of early death. Here, we found that, at admission, non-survivors patients at day-28 had more severe tissue hypoperfusion than non-survivors. In addition, mortality increased with the number of tissue hypoperfusion parameters. These results have to be confirmed prospectively in a large cohort of PE patients but suggest that tissue hypoperfusion could be helpful for triage and potentially for guiding treatment.
Conclusion In a retrospective study, we observed an association between tissue hypoperfusion parameters and 28-day outcome in ICU patients admitted for acute symptomatic PE.

Introduction
The administration of intravenous fluids is a key concern in the ICU. Fluids are used in many cases for renal protection, dilution of medications and as "maintenance" fluids. Moreover, they are considered as the cornerstone in resuscitation of patients with shock states [1]. Accordingly, large amounts of fluids are often administered with a resulting positive fluid balance and fluid overload. Currently, several studies have identified a relationship between fluid overload and mortality [2]. Most of them have estimated fluid overload by calculating the cumulated fluid balance. This calculation is subject to numerous inaccuracies. Another method is daily measurement of body weight. Few studies demonstrated the link between ICU mortality and early weight gain [2]. The aim of this study was to evaluate the association between early weight gain and ICU outcome. Patients and methods All patients admitted from the 1st January 2012 to 31th December 2013 in a medical ICU of a University Hospital for which a weight was recorded on admission (Day 0) and at Day 3 were included. Demographic data, comorbidities, SOFA score, cause of ICU admission, length of stay and mortality were recorded. Weight changes were calculated by subtracting body weight as measured on third day from one as measured on admission. Conclusion This study is one of the few to establish that early weight gain is an independent risk factor for mortality in ICU patients. The body weight, though subject to difficulties and possible errors in measurement is an indirect marker of fluid overload. It remains to establish if the control of fluid overload based on an early weight-gain control strategy is likely to improve patients outcome.

Introduction
The PROSEVA study was the first randomized controlled study having shown a positive effect of prone position (PP) on ARDS mortality [1]. Unlike previous randomized controlled trial, PEEP was adjusted with a PEEP-FIO 2 table, whose consequence was a small but significant decrease in the PEEP level. The aim of this study was to investigate the effect of PEEP variation during prone position on cardiac index in ARDS patients. Patients and methods Single center retrospective observational study performed on ARDS patients hospitalized in a medical ICU between July 2012 and March 2016. Patients included were adults fulfilling the Berlin definition for ARDS, undergoing at least one prone position session, under hemodynamic monitoring by the Picco ® device, with availability of hemodynamic measurements performed before (T1), at the end (T2), and after the prone position session (T3). Prone position sessions were excluded if they were performed >7 days after ARDS onset. The following variables were recorded: demographic, SAPSII, ARDS severity and risk factor, SOFA score and cumulative fluid balance at PP onset, delay between ARDS session and PP session, hemodynamic, arterial blood gas, ventilatory settings, plateau pressure, catecholamine dose and additional treatments. Statistical analyses were performed using prone position session as statistical unit and mixed models taking into account both multiple prone position sessions by patient and multiple measurements during a prone position session. Variables associated with cardiac index with a p value below 0.1 in univariate analysis were selected for inclusion in a multivariable mixed logistic regression model, using backward stepwise descending selection. p < 0.05 was chosen for statistical significance. Data are expressed as mean ± standard deviation.
Results 85 patients fulfilled the inclusion criteria over the study period, totalizing 149 prone position sessions (2 ± 1 sessions per patient). PEEP level decreased significantly from 9 ± 3 to 8 ± 2 cm H 2 O between T1 and T2. PEEP decreased by at least 5 cm of H 2 O in 18 (12%) of the PP sessions. Multivariate analysis identified 7 variables independently associated with cardiac index (model 1, cf.  [1][2][3] day. Thirty-three percent had comorbidities. Seventy per cent were transferred from emergency unit. The type of stay was medical for 95%. The primary reason for admission was essentially respiratory (44%) and neurologic (22%). Infection was the main cause of respiratory failure (79%). The destination was pediatric ward in 53% and home in 36%. Three per cent were transferred in pediatric intensive care unit and one patient deceased. Conclusion The patients hospitalized in pediatric intermediate care unit were young. The length of stay in these units was short and the respiratory failure was more frequent.

Competing interests
None.

Patients and methods
We did a prospective, observational, multicenter cohort study in seven French regional hospitals (09/2012-01/2014). All consecutive children <18 years were included. Three scores (PAWS, PEWS, and BedPEWS) were calculated each 8 h and more if deterioration. Binary outcome criteria were "medical call", and "admission to Pediatric intensive care Unit (PICU)". When one or two monitoring parameters necessary to calculate the score were missing, the score was still calculated and the missing value was considered normal and these scores were called "imputed scores". We used areas under the ROC curve (AUC) to estimate discrimination.
Results 2868 children were included for a total of 19,071 observations for calculating the three scores. Median scores for the three scores were respectively 2, 2 and 1. Medical call was observed in 11% (n = 2056), and admission to PICU in 0.45% (n = 85). AUCs calculated for the three scores for predicting medical call were ranged from 0.92 to 0.93. AUCs for predicting PICU admission were ranged from 0.89 to 0.92 (Table 23). Conclusion The three Pediatric Early warning scores, developed from pediatric ward and emergency department, can be used to detect deterioration requiring a medical intervention or PICU admission in hospitalized children in pediatric intermediate care unit.
still achieved to reach a big number of patients (84 patients in one epidemic season). We observed a very high rate of breastfeeding disturbance during hospitalisation for bronchiolitis in our hospital (over 50%). We expected to find severity of respiratory disease as the first risk factor but only 16.8% of mothers pointed out severity of disease as a reason for unwanted weaning and respiratory distress on medical charts was also not significantly different. Lack of support from caregivers was the first factor pointed out. It is a modifiable factor for further practice, especially in younger infants for witch breastfeeding has just begun and is still precarious. Conclusion Bronchiolitis is a high risk event for breastfeeding with about half of mothers in our hospital either stopping or diminishing their breastfeeding during hospitalisation. Correct advice and support could be a determining factor of breastfeeding's continuation and further studies should focus on interventions to prevent unwanted weaning.
during PP and only 10% reduce or stop it. Finally the mucocutaneous complications are the most frequent (54% of interviewed persons) just before the displacement of protheses (37%). The prevention of cutaneous pressure ulcers by the use of specific mattress is widespread (61% of interviewed persons) but no supplemental care in prone compared to supine positioning. Discussion The survey aimed to describe the techniques and modalities of PP in children, as well as several aspects of care associated with PP. The results show that the quality of care seems unaffected by the PP. The interviewed persons do not reported technical contraindication of PP. However, a special attention should be put on hemodynamic instability, tracheostomy or umbilical venous catheter. The survey reported the same complications in children and adult [2]. Finally, we also report the absence of paramedical care protocol for the PP. Future clinical studies will assess the impact of nursing protocols to limit complications, improve quality and comfort of care in children during PP periods. Conclusion The prone positioning in French-speaking pediatric intensive care units is not very well protocolized. Nevertheless, the care not seems to be significantly impacted by the prone positioning. Future research should focus on the evaluation of prone positioning nursing protocols.

Patients and methods
Introduction Patients on chronic hemodialysis are prone to develop overload pulmonary edema (OPE) but also infections, and have been reported to be regularly admitted to the intensive care unit (ICU) [1,2]. Acute respiratory failure (ARF) in this specific population can be challenging for the intensivist, as OPE requires ultrafiltration which could be deleterious in other causes. The objectives of this study were (1) to describe the spectrum of ARF causes in patients on chronic hemodialysis and (2) to identify predictors of pulmonary edema diagnosis.

Patients and methods
We performed a monocentric retrospective study in the 18-bed ICU of a non-teaching hospital with a nephrology unit. All patients on chronic hemodialysis admitted to the ICU for ARF in the period 2011-2015 were included. Patients with unknown diagnosis on ICU discharge were excluded. We defined ARF as either use of oxygen at ≥6 l/min dioxygen to achieve a pulse oximetry measurement of ≥90% (88% in COPD patients) or partial pressure of arterial oxygen (PaO 2 ) <60 mmHg on room air or need for mechanical ventilation plus of one of the following signs: a respiratory rate of >25 breaths/min, use of accessory respiratory muscles, or cyanosis associated to worsening dyspnea. The following data were collected: previous dry weight, time from last dialysis session, clinical, biological and imaging data on admission, number of dialysis sessions performed in ICU and total ultrafiltration volume required to withdraw oxygen therapy, weight on discharge, treatment and outcome, Patients were classified as having OPE (OPE group) or not (non-OPE group) based on the final diagnosis on ICU discharge. The two groups were compared with univariate analysis using non parametric tests. We did not find significant differences in temperature on arrival, white blood cells count nor NT-proBNP. Chest X-Ray interpretation was correct in 82% of cases of OPE patients but could not conclude in 46% of non-OPE episodes. Non-OPE patients had a more severe prognosis 38.5% of death (n = 5) versus 2.2% (n = 1).
Conclusion In this small retrospective monocentric study, ARF in patients on chronic hemodialysis is due to non-OPE cause in 22% of cases. A lower blood pressure, a higher CRP, and a shorter time from last dialysis session could prompt physician to search for a non-OPE cause and adjust cautiously ultrafiltration therapy.
Introduction Acute respiratory infection (ARI) is the most common infectious cause for admission to Intensive Care Unit (ICU) and the prevalence of infection increases with age. Changes in population demographics and comorbid illness may drive important changes in the composition of patients admitted to the ICU. Notably, how the population aging impacts on the incidence of hospitalized patients for ARI and how it increases the demand for critical care services is unknown. To address this knowledge gap, we sought to describe trends in demographics changes among elderly patients admitted to ICU for ARI on a 9-year period.

Materials and methods
We conducted a retrospective cross-sectional study based on hospital discharge databases (HDDs) from January 1, 2006 to December 31, 2014. We selected patients over 18 years old (y-o) who were hospitalized for ARI in a French region (Centre Val de Loire region, 2.5 million inhabitants, served by one university hospital, one regional hospital and 37 general and private hospitals). Cases of ARI were extracted from the HDD with an algorithm based on ICD-10 specific diagnosis codes, taking into account the type, number and position of these codes in the hospital discharge report. We previously validated the ICD-10 case definition reviewing a sample of medical charts as the gold standard. Giving the acceptable accuracy and precision of our algorithm, the following data were extracted from the

Fig. 30 See text for description
Introduction Pulmonary embolism is a major complication observed in critical ill patients. ICU environment patients are at high risk of deep vein thrombosis and pulmonary embolism. The objective of our study is to determine the incidence of this pathology and to analyze the prognosis of patients having a pulmonary embolism in ICU.

Patients and methods
Prospective study included all patients with confirmed pulmonary embolism by a spiral computed tomography scan showing one or more filling defects in the pulmonary artery or in its branches, over a period of 1 year from 1 July 2014 to 30 June 2015.
We studied the epidemiological, clinical, biological and radiological these patients. We compare two groups in univariate analysis: survivors and deaths.