Proceedings of Réanimation 2018, the French Intensive Care Society International Congress

Correspondence: Epaillard Nicolas - nicolas.epaillard@aphp.fr

Annals of Intensive Care 2018, 8(Suppl 1):CO-01

Introduction: The role and incidence of viral infections in respiratory failure occurring in ICU patients is more and more investigated. The availability of multiplex PCR assay is actually frequently performed. The aim of this retrospective study was to assess presence of viral species using a multiplex PCR assay in adult patients admitted to ICU for respiratory failure during the November–April period.

Patients and methods: All adult patients admitted to ICU during the November–April period for respiratory failure were eligible. A respiratory sampling was performed according to patient severity—sputum, tracheal aspiration, or per fibroscopic broncho alveolar lavage (LBA) and was analyzed for conventional bacterial strain. In all the patients, a multiplex vial PCR assay was performed on pharyngeal aspirate or LBA fluid.

Results: During the study period 123 patients were analyzed (male 76, SAPS II: 49 ± 17 + age 69 ± 18). 70 of them had at least one respiratory comorbidity. Forty-nine patients required invasive mechanical ventilation. Shock defined by requirement of vasoconstrictor was recorded in 31 patients (23%). In ICU mortality was 23%. Multiplex PCR assay was positive in 23 patients (17%). Co infection by bacteria and virus was noted in 4 patients. Influenzae virus was found in 10 patients and VRS in nine. Four patients with a positive viral sample had a fatal outcome (Influenza virus was found in all these). Mechanical ventilation was required in 45% of patients who were VRS positive but all had a favourable outcome.

Conclusion: Positive viral sample is not infrequent in ICU patients with respiratory failure. VRS is in this study as frequently found than influenzae.

Correspondence: Piton Gaël - gpiton@chu-besancon.fr

Annals of Intensive Care 2018, 8(Suppl 1):CO-02

Introduction: Plasma citrulline concentration, a biomarker of enterocyte function, is frequently low among critically ill patients. Plasma citrulline concentration might be decreased among critically ill patients presenting with sepsis, reflecting an alteration of small bowel function. We aimed to study the link between plasma citrulline concentration measured at ICU admission and variables related to the sepsis in critically ill patients.

Patients and methods: This was a prospective observationnal study performed in a University Hospital. Critically ill patients admitted to the ICU, aged 18 years or more, with ICU duration expected > = 48 h, without chronic renal failure or small bowel disease, were included. Plasma citrulline concentration, plasma I-FABP (biomarker of enterocyte damage), search for sepsis, results of bacterial sample, and usual clinical and biological variables were collected. We compared patients presenting with low plasma citrulline concentration (< = 10 µmol L), suspected of presenting enterocyte dysfunction, and patients with plasma citrulline concentration > 10 µmol L.

Results: Between July and December 2016, 109 patients, aged 64 years, 61% males, SOFA score 8 (6–12), were included in the study. Sepsis was found in 43 patients (40%), mainly of pulmonary (58%) and digestive (21%) origins. A plasma citrulline concentration ≤ 10 µmol L at ICU admission was associated with presence of Gram Negative or Candida infection, and higher plasma concentration of CRP and PCT (all p < 0.05). Plasma citrulline concentration was lower among septic patients than among patients without sepsis (12.0 µmol L [8.6–17.1] vs 20.5 µmol L (11.2–27.1) respectively, p = 0.004). Plasma citrulline concentration was not different among patients presenting with a sepsis of digestive origin and extra-digestive origin (7.8 [7.2–25.2] and 13.4 [9.8–16.3] respectively, p = 0.29). Plasma I-FABP was significantly higher among non-survivors (p = 0.005). Other factors associated with 28-day mortality were the age, plasma creatinine and lactate concentrations, and SOFA score.

Conclusion: This study confirms the link between sepsis and low plasma citrulline concentration in critically ill patients. The gut might be both the culprit, responsible for bacterial translocation, and the victim of sepsis. Indeed, the fact that plasma citrulline was also reduced during extra-digestive sepsis suggests the existence of septic enteropathy. I-FABP confirms its prognostic value among critically ill patients. The prognostic value of plasma citrulline in the critically ill requires further evaluation.

Correspondence: Guerpillon Brice - guerpillonbrice@gmail.com

Annals of Intensive Care 2018, 8(Suppl 1):CO-03

Introduction: Lower respiratory tract infections are very common in intensive care units and require rapid microbiological diagnosis in order to optimize the clinical progression and minimize the ecological impact. It currently takes 2–3 days to obtain an antibiogram using conventional microbiological techniques, compared with just a few hours for some multiplex PCR methods.

This prospective observational study focuses on a technique currently available, the multiplex PCR (MxPCR) Unyvero^® system that targets the 20 bacteria most commonly implicated in these infections as well as 17 of the most widespread antibiotic resistance genes. Results can be obtained within 5 h.

The primary objective of this study was to demonstrate the diagnostic performance of this technique, and the secondary objective was to assess its practical impact.

Patients and methods: Between December 2016 and September 2017, point-of-care testing using MxPCR and conventional microbiological techniques were performed on respiratory specimens from patients with pulmonary infection in 2 intensive care units. Results, delay to obtain it, clinical and therapeutic datas were recorded.

A total of 86 out of 103 samples collected were fully analyzed. The results obtained for 17 samples could not be interpreted due to a handling error (1), technical failures (14) and the absence of conventional antibiograms (2). Samples collected were mostly tracheal aspirates (45). Pathogenic bacteria were detected more often using MxPCR than conventional bacteriology techniques (43 samples compared to 36). Concordance ratio between those two technics is described in Table 1. Antibiotic therapy initiated in the 48 h preceding sample did not increase the mismatching (p = 0.42). Antimicrobial agents administered before microbiological sampling did not alter the concordance between culture and PCRMx (p = 0.42). Despite some inconsistencies, particularly regarding the detection of nonpathogenic saprophytic resistance, clinical and or ecological benefits were observed in 42% of cases with this technique. The benefits were mainly ecological with de-escalation in 24 cases and weaning off antibiotics in 11 cases. The clinical benefits (adjustment of therapy for bacteria not targeted empirically) were more limited as only 7 patients were affected and adjustment was delayed by 2.4 days on average.

Conclusion: Thanks to this technique, the treatment of respiratory infections could have been optimized in 42% (39/86) of cases with a modest clinical gain (7/86) but a significant ecological gain (35/86).

Correspondence: Misset Benoît - bmisset@wanadoo.fr

Annals of Intensive Care 2018, 8(Suppl 1):CO-04

Introduction: Sepsis and non-septic Systemic Inflammatory Response Syndrome (SIRS) are two similar syndromes, which differ by their cause, sepsis being secondary to microbial infection. Microbiological tests are not enough to early affirm the bacterial or fungal origin of the syndrome. More than fifty biomarkers have been proposed for this purpose but none have been repeatedly validated in external populations. Our goal was to address the accuracy of the biomarkers already reported as being able to efficiently discriminate between sepsis and non-septic SIRS.

Patients and methods: Prospective observational multicenter cohort of 363 consecutive patients with criteria of SIRS who were included when the attending physician considered they required antimicrobial therapy. We collected blood at inclusion to measure 53 biomarkers, including 29 plasma compounds, 10 whole blood RNA and 14 leukocyte surface markers. Patients were classified as sepsis or non-septic SIRS a posteriori, blindly to the biomarker results. Performance of each marker was assessed by AU-ROC curve. Multivariate analysis used a lasso regression (Tibshirani 1996; Musoro 2014), including multiple imputation for missing values and bootstrap validations, adapted for high numbers of variables (biomarkers).

Results: Eighty-four patients had exclusion criteria. Ninety-one patients had non-septic SIRS and 188 had sepsis. Median [IQ] SAPS 2 score was 55 [50–61] in both groups (p = 0.81). Time from ICU admission to inclusion was 1 [0–1] day. 69.8% of infections were pulmonary, most frequent strains were E. coli (23%), S. aureus (15%) and P. aeruginosa (13%). Causes of non septic SIRS were circulatory (37%), inflammatory (25%), hypoxemia (22%) and neurological (16%). Only 8 biomarkers had a ROC-AUC over 0.6 with a 95%IC over 0.5. In multivariate lasso analyses, only CRP and HLA-DRA circulating RNA were repeatedly associated with sepsis, and no model performed better than CRP alone (ROC-AUC 0.76 [0.68–0.84]).

Discussion: Strengths of this series are (1) the similar severity of SIRS and sepsis groups, (2) a multicenter design and (3) the use of a multivariate method adapted to the high number of independent variables.

Conclusion: In this prospective cohort, among 53 previously published biomarkers, CRP was found to better discriminate between sepsis and non septic SIRS than the other biomarkers. However, its performance was insufficient to be useful for routine use. HLA-DR circulating RNA was the only independent marker predicting infection. Multivariate models did not improve the performance of CRP alone.

Correspondence: Pouly Olivier - olivier.pouly@hotmail.fr

Annals of Intensive Care 2018, 8(Suppl 1):CO-05

Introduction: Procalcitonine (PCT) is a useful biomarker to reduce antibiotic exposure in intensive care units (ICU), especially in respiratory tract infections. Nevertheless, its usefulness for necrotizing soft tissue infections (NSTI) has not yet been reported. We aimed to assess PCT as a prognostic biomarker and a marker to guide duration of antimicrobial therapy in NSTI.

Patients and methods: We retrospectively included all the patients hospitalised for NSTI between January 1st 2013 and 31 December 2015 in an ICU department of a teaching hospital. Initial PCT value and kinetics, demographic and clinical data were reported. We compared the duration of the antibiotic course with: 1) the IDSA guidelines that recommend to use the decrease of fever and the need for surgical debridement and, 2) a theoretical duration if a PCT-based algorithm has been applied.

Results: We included 196 patients. Median LRINEC score was 6 [5–8.5] and median SOFA score was 6 [4–10]. At D28, mortality was 19.4%, significantly linked with initial PCT value (28.6 ng mL [3.5–58.7] in dead patients versus 3.1 ng mL [0.7–11.8] in survival ones (p = 10^–4). There was no decrease in PCT kinetic for 12.2% of the patients who survived versus 62.5% among those who died at D28 (p < 10^–5). The duration of the antimicrobial therapy was 12.5 days [10–17]. Stopping antimicrobial therapy if PCT levels was inferior to 0.5 ng mL or decreased of more than 80% from the initial value would have reduced the antibiotic treatment by 6.5 days [5–8] (p = 10^–5). It would have been possible wherever the localisation of the infection. But 17.9% of patients could not have had a PCT guided antibiotic duration because of an initial PCT value < 0.5 ng mL, especially in the cervical cellulitis group. Antimicrobial therapy would have been reduced by 4 days if IDSA guidelines had been applied [5–13] (p = 10^–5). The reduction of the antibiotic therapy was superior with the PCT algorithm compared to IDSA guidelines (p = 5.10^–4).

Conclusion: Procalcitonin could be a good prognostic marker as well as a relevant tool to reduce the antibiotic therapy duration in the severe NSTI. We will nevertheless have to prospectively determine the impact of such a practice on patient prognosis, including mortality and need for reintervention.

Correspondence: Benammi Sarah - sara-benammi@hotmail.fr

Annals of Intensive Care 2018, 8(Suppl 1):CO-06

Introduction: This study aimed to identify and describe a set of longitudinal HRQoL trajectories then determine factors associated with trajectory class membership of patients after acute illness.

Patients and methods: This was a prospective cohort study conducted in an acute medical unit of University Hospital, between June–September 2014. Patients aged ≥ 17 years admitted to AMU were included. Demographic, medical history, clinical and paraclinical characteristics were recorded at admission. EQ5D index, EQ-VAS and survival status of patients were collected at admission, 3, 6 and 18 months of follow-up. Latent class growth analysis (LCGA) was applied to identify classes of HRQoL trajectories, while association between baseline covariate and class membership were identified using polynominals logistic regression. Statistical analysis was carried out in STATA version 14.

Results: Two hundred fifty-one patients were included. The mean age was 55.6 ± 18.9 years, women were 54.6%. In-hospital and 540 days follow-up mortality were respectively 11.6 and 34.3%. For 229 patients included in LCGA, three trajectory classes where identified for EQ5D Index + stably low (16.2%), stably moderate (30.6%), and high initially increasing (53.2%). The three trajectory classes of EQ-VAS were low increasing, moderate initially increasing and high initially increasing with, respectively, 29.7, 48.5 and 21.8% of the patients. Concerning EQ5D index, comparing to high initially increasing trajectory, factors associated to—(a) stably low trajectory membership were-age ≥ 70 years (OR 9.3; CI 1.7–49; p = 0.008), intensive care unit transit (OR 5.1; CI 0.01–10; p = 0.01) and low hemoglobinemia (OR 0.8; CI 0.05–1.1; p = 0.05) + (b) stably moderate trajectory membership were—comorbidity (OR 2.8 + CI 0.01–6.4 + p = 0.01) and low hemoglobinemia (OR 0.8; CI 0.001–0.9; p = 0.001). Concerning EQ-VAS, comparing to high initially increasing trajectory, factors associated to—(a) low increasing trajectory membership were—female gender (OR 5; CI 1.7–10; p = 0.006), Km hospital-residence (OR 0.9; CI 0.9–0.9; p = 0.01) and comorbidity (OR 9.6; CI 3.1–29; p < 0.001), (b) moderate initially increasing trajectory membership was comorbidity (OR 9.5; CI 3.7–24; p < 0.001).

Conclusion: Three HRQoL trajectories were identified. Aged patients with low hemoglobinemia and who transited through ICU had the worst EQ5D index. Females, with comorbidities, living far from hospital perceived an amelioration of their, previously low, EQ-VAS.

Correspondence: Pottier Véronique - veroniquepottier@icloud.com

Annals of Intensive Care 2018, 8(Suppl 1):CO-07

Introduction: Hospitalization in ICU is a period of severe stress that can be experienced as a real trauma leading to post traumatic stress disorder (PTSD) after hospital discharge. The objective of this study is to determine the prevalence of PTSD and its impact on the quality of life at 3 months, 6 months and 1 year after ICU discharge.

Patients and methods: A before and after prospective, observational, non-interventional study assessing the impact of a nurse implemented sedation and analgesia algorithm on complications of critical illness and on outcome of patients after their ICU stay, was conducted in surgical ICU, between November 2014 and April 2017. During the “Before” period, sedation and analgesia was managed by the physician, while during the “After” period, it was managed by the nurses. Survivors were followed up by a phone survey at 3 months, 6 months and 1 year after ICU discharge.

Results: 1156 intubated and mechanically ventilated patients were admitted during the study period. Among the 145 included patients, 47 died and 35 were lost or disapproved the follow-up before the 3rd month. Only 62 patients accepted follow-up at 3 months (42.8%), 55 at 6 months (37.9%) and 42 at 1 year (29%). The presence of a new hospitalization after ICU discharge varies from 24.2 to 29.1%, between 3 months and 1 year, and the use of psychotropic drugs from 25.5 to 29%. The prevalence of PTSD (PTSS-10 score ≥ 35) was 12.9% at 3 months, 10.9% at 6 months and 16.7% at 1 year. ICU Stay has a moderate impact on quality of life at 3 months, 6 months and 1 year (62 < WHOQOL-BREF score < 96). Approximately 10% of patients have a severe impact on their quality of life (WHOQOL-BREF > 61) at ICU discharge, up to 1 year. The PTSS-10 score at 3 months is significantly higher in the “Before” period. There is no difference at 6 months and 1 year, nor significant difference on quality of life between the 2 periods.

Conclusion: Improving the quality of care in ICU should also take into account the possible deleterious consequences of ICU stay. Follow-up is paramount in patients at risk, and increased efforts to improve quality of life after ICU discharge are needed.

Correspondence: Charuel Tessa - t.charuel@yahoo.fr

Annals of Intensive Care 2018, 8(Suppl 1):CO-08

Introduction: In intensive-care units, intubated patients, who can’t speak and patients unable to designate because of physical weakness are strongly limited for communicating with relatives and caregivers, despite cognitive function which may be preserved. There is a lack of validated tools to assess oral comprehension, adapted to those patients. Outside of the critical care setting the Montreal-Toulouse aphasia test (MT-86) is a validated tool to assess language and communication capabilities. It includes an evaluation of the lexical and syntaxic comprehension. This preliminary study aimed to adapt the oral MT-86 to overcome physical limitations of intensive-care unit patients using eye-tracking technology.

Patients and methods: A simplified version of the comprehension tasks from the MT-86, assessing the receptive lexicon and syntax was used. 47 item slides (2 or 4 pictures) were digitalized and integrated into an 8 min automated protocol, presented on a computer screen. The patient’s ocular fixation on pictures, when instructed to do so (human voice), was recorded by an eye-tracking device (see Fig. 1, patient instructed to “look at the button”). For patients regaining capability, the MT-86 paper version was performed.

Results: 11 control subjects (21–70 years old, 2 men 9 women) were used for device development which enabled to defining the response validation criterion after 3.5 s of fixation on the instructed picture. 19 non-intubated patients (12 women 7 men, 24–84 years old) were tested with both versions, and 4 intubated patients with the eye-tracking version. No side effect was noted. 4 types of responses were observed: response validation, sporadic fixation time (no choice), shortened fixation time (low awareness), no fixation. Two patients interrupted the test in both versions, at the same point, demonstrating the same resource cost from both modalities. The Fig. 1 presents 2 patients’ ocular fixation (each color corresponds to one patient). Green patient answered successfully and the total fixation duration was the most important for the area of interest “button”. Pink patient ocular fixation answered not successfully (phonological distractor).

Conclusion: This study shows feasibility of passing the MT-86 using eyetracking technology among critically ill patients. The tool appeared easy to use thanks to its brevity and automation. Further development may provide intensive care unit healthcare workers with a new device to assess comprehension and attention capacities of patients. This study may provide a new evaluation tool for patients among whom classical tests cannot be conducted.

Correspondence: Bailly Sebastien - sbailly@chu-grenoble.fr

Annals of Intensive Care 2018, 8(Suppl 1):CO-09

Introduction: Patients admitted in ICU have frequently several chronic diseases, including obstructive sleep apnea syndrome (OSAS). To date, OSAS was not robustly considered as a determinant of ICU patient prognosis. The objective of the study was to assess the impact of OSAS (known at admission) in patient prognosis of ICU patients.

Patients and methods: Data were retrospectively collected between 2006 and 2013 from two centers of a French national prospective database. A nested exposed-unexposed design was used. The OSAS status was checked for exposed patients using the hospitalization files. Adult patients without therapeutic limitation and staying alive in ICU at least one day were included. OSAS patients were matched (1 − n, n > 2) with non-exposed patients on body mass index, age and presence of chronic obstructive pulmonary disease (COPD). The sample size was computed to reach a power upper than 90% to show a one-day increase of ICU stay (± 4 days). The quality of matching was assessed using the standardized mean differences. The following outcomes were considered: length of stay alive in ICU, ICU mortality, in-hospital mortality, ventilator-associated pneumonia (VAP).

Results: From 5146 patients admitted in both ICUs and included in the study, 288 had a OSAS at ICU admission (5.6%). OSAS patients were mainly male (71 vs 60%), were older (64 years ± 12 vs 59 ± 17), had a higher BMI (29 ± 11 kg m⁻² vs 22 ± 4 kg m⁻²), have a more frequent COPD status (19.4 vs 8.2% p < 0.01) and cardiovascular diseases (35.7 vs 20.5%), had a higher SOFA score at ICU admission (6.6 ± 4 vs 5.2 ± 4.4). After matching (289 OSAS exposed vs. 571 OSAS non-exposed patients), there was no significant impact of OSAS on the length of stay alive in ICU (adjusted RR: 1.09–95% confidence interval: [0.88; 1.36], p = 0.42). Sensitivity analyses showed no differences in the results. Using logistic conditional regression models, the impact of OSAS on ICU mortality, in-hospital mortality and VAP were not different between OSAS and unexposed patients. Limitation—under-diagnosed OSAS for non-exposed patients and the presence of unmeasured confounders in the matching process are major limitations of the study.

Conclusion: In a large cohort of ICU patients, obstructive sleep apnea syndrome do not impact patient’s prognosis and outcome.

Correspondence: Goutorbe Philippe - philippegoutorbe@me.com

Annals of Intensive Care 2018, 8(Suppl 1):CO-10

Introduction: Physicians in charge of critically ill patients deal with very high amount of data. Sources of data are as various as biology, bacteriology, monitoring, therapeutic or clinical inputs. ICU dedicated PDMS provides a quick scalable synthesis of this information. Besides computerized physician order entry (CPOE) seems to reduce medications errors in intensive care unit. It enhances traceability of continuous infusion, planed drugs or any items part of the care plan. But in some emergency cases, orally given orders may not be captured in the Medical file. We studied crush induction traceability in Centricity Critical Care (CCC), ICU dedicated solution, from GE Healthcare.

Patients and methods: We performed a retrospective anonymous database analyze of 2016 in 21 French ICU using CCC. We checked: the number of endotracheal intubations done in ICU (intubation 20 min or later after admission), the number of induction doses of rapid onset neuromuscular blockade agents (rocuronium or suxamethonium) at the same time (sampling only non-continuous perfusion). For each ICU, we asked if nurses were allowed to capture medications on behalf of physicians in the CPOE.

Results: During 2016, 19,736 patients were admitted, mean age was 60.68 mean SAPS II was 40.64 and length of stay was 6.69 days. We identified 5516 ICU intubations and found 909 rocuronium and 829 suxamethonium induction doses. The overall traceability ratio of induction was 31.51%. The traceability greatly varies between different ICU: individual ratio range from 1.32 to 83%. Third quartile was 51.55%. Traceability was better in ICU’s where nurses can capture medications in CPOE 48.7 versus 22.49% p < 0.02 (Fig. 1).

Conclusion: Our study may suffer from 2 opposites bias: replacement of tracheal tube without anesthesia or use of bolus of rocuronium for another reason than intubation. Nevertheless, induction drugs seem to be greatly under written in the PDMS. Where nurses support, traceability is better. CCC allowed such database analyzes. Physician in charge of ICU should periodically, among others, check this indicator. Collaborative work with PDMS providers should keep enhancing traceability and safety. In our case, GE agreed in future software releases to add on top of “prescription” a “nurse note option” and when a nurse is adding a device such as endotracheal tube, tracheostomy or gastrostomy to insert a question: did you perform an anesthesia? Clinical PDMS databases should be great sources of knowledge and improvement areas only in the case of serious data collection.

Correspondence: Gibelin Aude - aude.gibelin@gmail.com

Annals of Intensive Care 2018, 8(Suppl 1):CO-11

Introduction: The acute respiratory failure (ARF) of patients with known or de novo small-vessel vasculitis (SVV) admitted to the intensive care unit (ICU) may be due to immune or non-immune causes. The prognosis is poor. Early identification of an underlying immune disorder is essential to initiate appropriate treatment. The aim of the study was to determine the incidence, describe the clinical presentation and assess the prognosis of immune ARF related to SVV.

Patients and methods: This retrospective multicenter study was conducted from January 2007 to June 2017 in 10 ICUs in France. Patients were identified from computerized registers using either keys words research or the International Classification of Diseases, Tenth Revision (CIM10). Inclusion criteria were (1) admission to the ICU for ARF, and (2) known or de novo granulomatosis with polyangiitis, eosinophilic granulomatosis with polyangiitis, microscopic polyangiitis, or anti-glomerular basement membrane antibody disease. Demographics, characteristics of ICU admission, treatments and outcomes were extracted from medical records.

Conclusion: Half of the patients with a known or de novo SVV admitted to the ICU for ARF have an immune cause of ARF. Readily identified predictive factors of an immune cause could allow the introduction of an early and targeted treatment.

Correspondence: Dangers Laurence - laudangers@gmail.com

Annals of Intensive Care 2018, 8(Suppl 1):CO-12

Introduction: Non-invasive ventilation (NIV) is a cornerstone therapy of acute respiratory failure (ARF) and is increasingly used in the intensive care unit (ICU). A recent survey suggests that up to 37% of patients report a certain level of anxiety related to NIV sessions. However, few is known on the actual prevalence, severity, risk factor and prognosis impact of anxiety in patients receiving NIV for an ARF. Objectives—to determine the prevalence and intensity of anxiety in patients receiving NIV for ARF, to identify the factors associated with anxiety, to assess the impact of anxiety on the risk of NIV failure, outcome, quality of life and post—ICU burden.

Patients and methods: Second analysis of a prospective observational cohort study in patients receiving NIV for ARF in 54 ICUs in France and Belgium, in 2010–2011. Anxiety was quantified using a graded ordinal scale from 0 (no anxiety) to 4 (very strong anxiety). A level of anxiety of 0 or 1 defined “mild or no anxiety” and a level of anxiety of 2, 3 or 4 defined a “moderate to severe anxiety”.

Results: Of the 419 patients included, 280 (67%) had mild or no anxiety and 139 (33%) had moderate to severe anxiety. Moderate to severe anxiety was independently associated with de novo ARF (OR 2.14; 95% CI 1.31–3.49; p = 0.002), male gender (OR 1.60; 95% CI 1.01–2.53; p = 0.047), dyspnea (OR 2.03; IC 95% 1.29–3.20; p = 0.002) and NIV intolerance (OR 4.9; 95% CI 3.03–7.93; p < 0.001). Factors of NIV failure were moderate to severe anxiety (OR 1.78; 95% CI 1.07–2.95; p = 0.027), SAPS 2 (OR 1.07; IC 95% 1.05–1.09) and de novo ARF (OR 3.30; IC 95% 1.97–5.55). In patients with moderate to severe anxiety, there was a trend toward a higher in-hospital (24 vs. 16%; p = 0.052) and day-90 mortality (29 vs. 20%; p = 0.05). Anxiety was not associated with higher length of stay, higher post ICU burden and altered quality of life.

Conclusion: Among patients receiving NIV for ARF, anxiety is frequent and potentially severe. Moderate to severe anxiety is associated with NIV failure and a trend toward higher mortality. Future studies should evaluate the benefit strategy aiming at improving anxiety management in ICU patients.

Correspondence: Augy Jean-Loup - augyjeanloup@hotmail.com

Annals of Intensive Care 2018, 8(Suppl 1):CO-13

Introduction: Veno-venous extracorporeal CO₂ removal (ECCO2R) is a promising new therapeutic option in the critical care setting. We conducted a prospective observational study of the use of ECCO2R in selected voluntary centers during 2 years aiming to assess the prevalence of the ECCO2R use mainly among COPD and ARDS patients.

Patients and methods: Two medical devices: Hemolung (Alung Technologies, Pittsburgh, USA) and iLA Activve (Xenios Novalung, Heilbronn, Germany) were selected after literature and medico-economic evaluations. A specific medical and nurses training was provided in each center. Data were collected on a dedicated form and were centralized by the coordinating center. Primary outcome was the number of patients treated per month and per center during the 2-years study period. Secondary outcomes were ICU and hospital-mortality and adverse events complications related to device use.

Conclusion: Our data indicate a preferential use of veno-venous ECCO2R devices in very severe (as illustrated by the overall high mortality) COPD and ARDS patients; with a lower than expected rate of utilization. Safety remains a major concern, indicating the need for further technological improvements as well of for optimization of anticoagulation regimen. Ultimately, RCTs will help to delineate clinical indications in these 2 main settings (COPD and ARDS).

Correspondence: Bouadma Lila - lila.bouadma@bch.aphp.fr

Annals of Intensive Care 2018, 8(Suppl 1):CO-14

Introduction: Maintaining a semirecumbent position (SRP) with a backrest elevation (BE) of 45° is not feasible for patients on mechanical ventilation (MV). Therefore, the effect of a SRP on the development of ventilator-associated pneumonia is uncertain. Our objective was to determine the efficacy of a simple electronic monitoring device to maintain a SRP during MV.

Patients and methods: We conducted a prospective, randomized, crossover study in two ICUs. BE of the patients were continuously assessed during a sequence of two consecutive 24 h-periods. Each period was either a control day = CD or a prototype day = PD. During the two periods, the targeted BE was between 40° and 45°. During the CD, BE was managed according to an internal procedure published previously. During the PD, an electronic device developed by our institution was added to the bed. The device was able to monitor continuously the BE and to alarm if the patient is not in the proper position. Adults patient were eligible if they were intubated with an expected duration of MV > 48 h and not eligible if they had a contraindication to SRP or a SAPS II score > 65. The time spent in a SRP defined by a BE between 40° and 50° was compared first in each crossover period by a Wilcoxon Rank-Sum test and then by a linear mixed effect model.

Results: One-hundred-five patients were randomized and a complete set of data were available for 103 patients: 72 men + age 59 ± 16 years, SAPS II score 48 ± 14 + acute respiratory failure in 28% and coma in 27%. BE was recorded for 833 ± 435 min during the CD and 1211 ± 333 min during the PD. During the two periods, patients had a median proportion of time spent in SRP about 30% higher in PD than in CD (IQR [25–75%]). In period 1 and period 2, patients had a higher median proportion of time spent in SRP in a PD than in a CD (93% [86–96%] vs 59% [34–79%], p < 0.0001) and (95% [78–97%] vs 62% [31–92%], p < 0.005) respectively. No period effect was identified. No adjustment covariate was significantly associated with the proportion of time spent in the SRP.

Conclusion: In ICUs with a high proportion of patients in a SRP (about 60%), a simple removable, washable, easily sanitized electronic monitoring device is able to raise this proportion at about 95%. Prospective studies are required to assess its clinical impact.

Correspondence: Haudebourg Anne-Fleur - annefleur.haudebourg@gmail.com

Annals of Intensive Care 2018, 8(Suppl 1):CO-15

Introduction: Ineffective triggering is frequent during pressure support ventilation (PSV). Its occurrence is favored by dynamic hyperinflation that may arise when increasing the pressure support level (PSL). Decreasing the PSL however fails to suppress ineffective triggering in a subgroup of patients that are therefore exposed to refractory ineffective triggering. Proportional assist ventilation with load-adjustable gain factors (PAV +) decreases the incidence of ineffective triggering in unselected patients but its effect on refractory asynchrony during PSV is unknown. The main aim of our study was to assess the effect of PAV + in patients exhibiting refractory ineffective triggering during PSV.

Patients and methods: Refractory ineffective triggering was defined as persisting ineffective triggering at the lowest tolerated PSL. Patients with refractory ineffective triggering during PSV were included. Flow, airway and esophageal pressures were continuously recorded during 10 min under PSV with the minimal tolerated PSL, and then under PAV + with the gain adjusted to target a muscle pressure between 5 and 10 cm H₂O. Primary endpoint was the comparison of asynchrony index between PSV and PAV + . Continuous data are reported as median [25^th–75th percentiles].

Results: Seven patients with refractory ineffective triggering were included so far. The median lowest tolerated PSL was 16 cm H₂O [12–18], with a PEEP level of 8 cm H₂O [5.5–8] and a PaO₂ FiO₂ ratio of 209 mmHg [178–225]. The median gain during PAV + was 70% [63–73]. The asynchrony index was significantly lower with PAV + than PSV (3% [2–8] vs. 24% [13–48] respectively, p = 0.02). Moreover, the asynchrony index decreased in every patient with PAV + (Fig. 1). Noticeably, the tidal volume was already protective in PSV and decreased even more during PAV + (5.71 mL kg [4.6–9.3] vs. 5.71 mL [1.8–7.1] respectively, p = 0.03); and the neural respiratory rate was high in both modes (41 cycles min [26–51] in PSV vs. 40[28–44] in PAV + , p = 0.61). Total esophageal Pressure–Time Product (PTPes) did not significantly differ between the two modes but the PTPes proportion that was wasted in ineffective efforts decreased with PAV + (1% [1–2] vs. 18% [7–37], p = 0.03).

Conclusion: Our preliminary data suggest that: (1) PAV + reduces the incidence of refractory ineffective triggering; (2) patients exposed to refractory ineffective triggering during PSV seem characterized by rapid shallow breathing despite high ventilatory support, questioning the tolerance of both ventilatory modes. Results with further inclusions will be presented.

Correspondence: Bougouin Wulfran - wulfran.bougouin@gmail.com

Annals of Intensive Care 2018, 8(Suppl 1):CO-16

Introduction: An immediate coronary angiogram (CAG) may be associated with better outcome after out-of-hospital cardiac arrest (OHCA) in neurologically preserved patients but could be futile in other cases. We aimed to assess the relationship between an immediate invasive strategy and survival after an out-of-hospital cardiac arrest (OHCA) of presumed cardiac cause, according to prognosis evaluated on hospital arrival.

Patients and methods: From May 2011 to May 2015, we collected data for all patients admitted in hospital after OHCA in Paris and suburbs (France). Risk of in-hospital death was retrospectively calculated using the validated Cardiac Arrest Hospital Prognosis (CAHP) score. Independent predictors of survival at discharge (including immediate CAG) were assessed in multivariate logistic regression in each of the 3 pre-defined subgroups of CAHP score: low (< 150 points), medium (150–200 points) and high (> 200 points) risk for in-hospital death.

Results: 1410 patients were included and overall survival rate at hospital discharge was 32%. Distribution in the low, medium and high-risk CAHP subgroups was 667 (47%), 469 (33%) and 274 patients (20%) respectively. The rate of early CAG was 86, 66 and 47% in the low, medium and high-risk subgroups, respectively (p < 0.001). Early invasive strategy was independently associated with better survival in low-risk patients (OR 2.3, 95% CI 1.4–3.9, p = 0.001), but not in medium (p = 0.55) and high-risk (p = 0.43) patients. Sensitivity analysis found consistent results.

Conclusion: In cardiac arrest patients, our results suggest that the potential benefit of an early coronary angiogram (and subsequent revascularization) is restricted to those with a preserved neurological prognosis at hospital arrival.

Correspondence: Leloup Maxime - leloup.mx@gmail.com

Annals of Intensive Care 2018, 8(Suppl 1):CO-17

Introduction: To our knowledge, no study mapped the epidemiology of unexpected cardiac arrest (i.e. with resuscitation attempt) in French ICUs. The current research primarily aims at describing the demographics, management, and prognosis of patients concerned.

Patients and methods: “ACIR” (in French: Arrêt Cardiaque Inattendu en Réanimation) is a prospective multicenter study that started on January 1st 2017, and is expected to last 1 year. All victims of cardiac arrest during ICU stay are included, whereas patients presenting with cardiac arrest at the time of admission are not. By convention, any resuscitation attempt (chest compression, adrenaline and or electric shock) defines the “unexpected cardiac arrest”. The data collected comprise medical history, circumstances of happening, ongoing treatments, resuscitation maneuvers, and outcome). A 6-month follow up period is planned for survivors.

Results: Forty-four centers (17 University and 27 general hospitals) are participating in the study, including 8 medical, 7 surgical and 29 medical surgical ICUs. Among those, nearly one-third have implemented a specific protocol regarding the management of cardiac arrest. Around half undergo regular resuscitation trainings, and among them 80% do so at least annually. Following unexpected cardiac arrest, most of ICU staffs set up post hoc debriefings (always 32%, often 20%, sometimes 41% or never 7%). On two-thirds of the investigation period, 2842 patients were victim of cardiac arrest in the participating units. Of those, 2426 did not endure any resuscitation attempt: 1687 were previously attributed a “do-not-resuscitate” order, 589 were unresponsive to maximal therapy and 150 for other reasons. Four hundred and sixteen patients (69% men, 68 ± 13 years, mean SAPS2 72 ± 25) were included in the “unexpected cardiac arrest” chart. Half of the events occurred within the first 24 h from admission. Return of spontaneous circulation was achieved in 271 (66%) patients (67% men, 68 ± 13 years, mean SAPS2 70 ± 25), and among them 69 (25%) were discharged alive from hospital (62% men, 64 ± 15 years, mean SAPS2 64 ± 24).

Conclusion: According to the Utstein style, our intermediate data are within the range reported in relevant literature. Although initial resuscitation was successful in two-thirds of cases, only 16.5% of victims of unexpected cardiac arrest in the ICU were discharged alive from the hospital. Among other concerns, this study aims at determining among participating centers what specific interventions are likely to enhance prognosis and quality of life for cardiac arrest patients in the ICU.

Correspondence: Ronflé Romain - romainronfle@hotmail.fr

Annals of Intensive Care 2018, 8(Suppl 1):CO-18

Introduction: Venous-to-arterial carbon dioxide difference (cv-art CO₂ gap) is a marker of tissue perfusion. Despite being in the security zone, high central venous oxygen saturation (ScvO₂) is associated with poor outcome in patients with septic shock. The aim of this study is to assess the ability of the cv-art CO₂ gap to predict clinical worsening in patients with septic shock, according to ScvO₂.

Patients and methods: We performed a multicentric prospective study in 3 ICUs. All consecutive patients with septic shock were included during the study period. Patients were assigned into three groups according to ScvO₂. Clinical worsening was defined as an increase of SOFA score ≥ 1 two days after admission ∆SOFA ≥ 1). Patients characteristics and severity scores were collected during the first 3 days of stay. To assess the ability of the cv-art CO₂ gap to predict a clinical worsening, a ROC curve was produced for the event ∆SOFA ≥ 1. Associations between cv-art CO₂ gap and 28-day mortality along with length-of-stay (LOS) were explored using linear regression and correlations.

Conclusion: cv-art CO₂ gap is a parameter to assess the inadequacy of circulating blood flow in response to metabolic demand. cv-art CO₂ gap could help clinicians to identify septic shock patients with organ dysfunction despite normalization of ScvO₂.

Correspondence: Joffre Jérémie - jeremie.joffre@aphp.fr

Annals of Intensive Care 2018, 8(Suppl 1):CO-19

Introduction: Context—Metabolic acidosis is commonly observed in critically ill patients. Experimental studies suggested that acidosis by itself could impair vascular function but such hemodynamic effect has been poorly investigated in human. Objectives: To assess the relationship between metabolic acidosis severity and endothelial microvascular function.

Patients and methods: Design, settings and patients: Prospective monocenter study. All adult diabetes patients admitted in our medical ICU for ketoacidosis (pH < 7.35) were included. Endothelial microvascular response to acetylcholine iontophoresis was measured at admission (H0) and after correction of metabolic acidosis (H24).

Results: Thirty-nine patients with diabetes ketoacidosis were included, 68% (n = 18) were male, with a median age of 43 [31–57] years. At admission, endothelial function significantly correlated with acidosis severity, the deeper acidosis was the worst endothelial activity (R = 0.53, p < 0.001). Endothelial response was strongly depressed below a pH at 7.20 (AUC 1779 [740–3079] versus 12,944 [4874–21,596], p < 0.0001). At H24, after rehydration and insulin infusion, clinical and biological disorders were fully corrected. After acidosis correction, endothelial response improved more in patients with baseline deep acidosis (pH < 7.20) when compared to patients with mild acidosis (AUC + 453 [213–1470] versus + 121 [79–312]  %, p < 0.01).

Conclusion: We documented an endothelial dysfunction during metabolic acidosis in critically ill patients with diabetic ketoacidosis. Endothelial dysfunction recovered after acidosis correction.

Correspondence: Beurton Alexandra - alex.beurton@gmail.com

Annals of Intensive Care 2018, 8(Suppl 1):CO-20

Introduction: The passive leg raising (PLR) test mimics a volume challenge and is based on the passive transfer of some venous blood from the legs toward the cardiac cavities when moving a patient from the semi-recumbent to a passive leg raised position. Nevertheless, intra-abdominal hypertension (IAH) may increase the resistance to venous return through the inferior vena cava and may impede the PLR-induced increase in cardiac preload. We tested the accuracy of the PLR test to predict fluid responsiveness in case of IAH.

Patients and methods: In patients with an intra-abdominal pressure (IAP) > 12 mmHg, we measured the changes in cardiac index (PiCCO2 device) induced by a PLR test and by a 500-mL saline volume expansion. The IAP (bladder pressure) was measured at different study steps.

Results: Twenty-nine patients were included, 20 being fluid responders (fluid-induced increase in cardiac index 15%) and 9 fluid non-responders. The IAP at baseline was 20 ± 6 mmHg. It significantly decreased during the PLR test in fluid responders (by 34 ± 13%) and in fluid non-responders (by 30 ± 16%). In fluid responders, cardiac index increased by 7 ± 9% during PLR and by 22 ± 6% during volume expansion. The PLR test was negative (PLR-induced increase in cardiac index < 10%) in 15 patients (false negatives) and positive in 5 patients (true positives). In fluid non-responders, cardiac index increased by 6 ± 6% during PLR and 6 ± 3% during VE. The PLR test was negative in all but one of them was positive (false positive). The sensitivity and specificity of the PLR test to detect fluid responsiveness were 40% (95% confidence interval 19–64) and 89% (52–100), respectively. The area under the receiver operating characteristic curve was 0.58 ± 0.11.

Conclusion: Intra-abdominal hypertension is responsible for false negatives to the PLR test. The PLR test significantly reduces IAP.

Correspondence: Nuzzo Alexandre - al.nuzzo@gmail.com

Annals of Intensive Care 2018, 8(Suppl 1):CO-21

Introduction: Acute mesenteric ischemia (AMI) has high mortality and intestinal resection rates. An intestinal stroke center based on a multimodal and multidisciplinary management focusing on intestinal viability was created in our center in January 2016. We aimed to study AMI patients and outcomes.

Patients and methods: Single-center, observational and prospective study was carried out in our intestinal stroke center. AMI was defined as an acute intestinal injury related to a splanchno-mesenteric insufficiency and without alternative diagnosis. All AMI patients underwent a computed tomography angiography. AMI were classified in occlusive (atheroma, thrombosis, embolisms etc.) and non-occlusive (splanchnomesenteric hypoperfusion, vasoconstrictive medications), arterial and venous. Patients with post-AMI short bowel syndrome (SBS), chronic mesenteric ischemia, aorta dissection anevrism and left ischemic colitis were not included. Outcomes studied were mortality at 3 months and at the end of follow-up, intestinal resection, SBS and parenteral nutrition requirement. Predictive factors of intestinal resection and mortality were studied in uni- and multivariate analyses. Quantitative data were reported as medians (range). A p value < 0.05 was considered to be significant.

Results: In 124 included patients (female 38%, 66.1 (26.3–95.6) yo) AMI was occlusive, non-occlusive, arterial, venous in 89, 11, 81 and 18.5%, respectively. Main etiologies of occlusive AMI were atheroma (39.5%), thrombosis (35%), embolism (9.7%). Occlusion concerned one, two or three of the digestive arteries in 54, 23 or 13%, respectively and superior mesenteric artery in 89.1% of arterial AMI. Chronic mesenteric ischemia preceded AMI in 22%. A surgical (n = 23) and or radiological (n = 36) revascularization was performed in 56.4% of arterial AMI. Intestinal resection was necessary in 31.4%. At end of follow-up of 290 (0–614) days, 22 (18%) patients had a SBS and 16 (13%) required parenteral nutrition. Mortality at 3 months and at the end of follow-up were 11.3 and 16.9%, respectively (Figure). In multivariate analysis a stay in ICU was the only factor associated with resection (p < 0.001, OR 7.9 (2.9–21.4)) and mortality (p = 0.02, OR 4.4 (1.2–15.4)).

Conclusion: AMI patients treated in the intestinal stroke center have a survival and small bowel preservation rates of more than 80 and 70%, respectively. Such structure could serve as a model for multicentric studies and to improve at a large territorial scale the prognosis and the knowledge of AMI.

Correspondence: Jacquot Audrey - audreyjacquo@hotmail.com

Annals of Intensive Care 2018, 8(Suppl 1):CO-22

Introduction: The use of prone positioning (PP) for patients presenting moderate to severe Acute Respiratory Distress Syndrome (ARDS) has increased since the PROSEVA trial. Adverse events associated with PP have been well described such as pressure sores. But although patients with moderate to severe ARDS had high mortality, had prolonged length of stay and were particularly exposed to central venous catheters (CVC), there were no data reporting intravascular catheter-related infections (ICRI). In this study, we evaluated the incidence of CVC colonization in patients exposed to PP for ARDS.

Patients and methods: In this retrospective bicentric observational study we compared two groups of adults, assigned 1:1. The “exposed” group was composed of patients treated with PP for moderate to severe ARDS. The “non-exposed” group was composed of non-ARDS patients matched with: centre, year, sex, age, APACHE II score and length of ICU stay. ICRI incidence rate (1/1000 catheter-days) and preventive bundles were similar between the two centres, as well as the practice for PP session. The primary outcome was the incidence of catheter-related colonization in each group. The secondary outcome was the incidence of a composite criterion which evaluated an overall infectious risk associating colonization and/or bloodstream catheter-related infection and/or ICRI.

Results: Between January 1st, 2014 and December 31st, 2015, 173 patients were eligible for the “exposed” group: we matched 101 patients with 101 non-exposed patients. ARDS patients were mainly composed of direct ARDS (Pneumonia), median length of mechanical ventilation was 20 days, the average number of PP session was 2 and its mortality rate was 30%. “Non-exposed” group included a majority of cardiogenic shock. The number and length of catheter did not differ significantly between groups. The incidence of CVC colonization was 23.8% in the exposed group and 8.91% in the non-exposed group (p = 0.007) (Fig. 1). The odds ratio for PP was 3.8 (p = 0.01). The incidence of composite criterion was significantly higher in the exposed than in the non-exposed group (34% versus 19% p = 0.02).

Conclusion: This is the first study to assess the incidence of ICRI in ARDS patients undergoing prone positioning. In our study, PP was associated with a higher rate of colonization. The overall catheter-related infectious risk was affected by the PP. We probably identified a population at high risk of ICRI who may benefit from additional preventive measures.

Correspondence: Levy Bruno - blevy@sfr.fr

Annals of Intensive Care 2018, 8(Suppl 1):CO-23

Introduction: This study was designed to assess the effect of two extracorporeal life support (ECLS) blood-flow strategy in an experimental model of E-CPR in the first six hours of resuscitation on macrocirculatory and microcirculatory parameters, lactate clearance and cytokine storm.

Patients and methods: Cardiac arrest was induced in 18 pigs by surgical ligature of the left coronary artery. ECLS was initiated after 40 min of cardiopulmonary resuscitation and the ECLS blood flow was set on 30–35 ml kg⁻¹ min⁻¹ (low-blood-flow group, LBF) versus 65–70 ml kg⁻¹ min⁻¹ (standard-blood-flow group, SBF) according to the randomized group. Continuous systemic blood pressure and carotid blood flow were continuously monitored. Blood gas analysis and lactate were measured at baseline H0, H3 and H6. Sublingual microcirculatory was assessed by sidestream dark field (SDF) technology and the following parameters were assessed: total and perfused vessels density (TVD, PVD), the proportion of perfused vessels (PPV) and microvascular flow index (MFI). Leg tissue oxygenation (StiO2) was monitored by a Near-Infrared Spectrometer (NIRS) device. Cytokine inflammatory was measured by enzyme-linked immunosorbent assay (ELISA).

Results: There was no differences between groups at baseline and at ECLS initiation (H0). Lactate and sublingual microvascular parameters were significantly impaired after the low-flow period. MAP target (65 mmHg) was reach in each randomized group. Total infused norepinephrine and total infused fluid were similar between the two groups. A significant difference was observed in the six hours evolution concerning carotid blood flow (LBF vs SBF at H6–19 [5–34.45] vs 67.81 [43.5–82] %, p < 0.05). Lactate clearance at H6 was inferior in the LBF compared with the SBF (6.67 [− 10.46 to 18.78] vs 44.72 [19.54–69.07] %, p < 0.05). Concerning the microvascular parameters, the LBF had lower PVD, PPV and MFI at H3 compared with the SBF but no significant difference observed at H6. TNF: was lower (361 [73–778] vs 1164 [177–1848] ng ml⁻¹, p < 0.05) in the LBF at the end of the experiment.

Conclusion: In an experimental porcine model of refractory cardiac arrest treated by ECLS, a low-blood-flow strategy during the first six hours of resuscitation was associated with a decreased cerebral blood flow, lactate clearance and microcirculatory parameters despite a lower inflammatory response.

Correspondence: Geri Guillaume - guillaume.geri@aphp.fr

Annals of Intensive Care 2018, 8(Suppl 1):CO-24

Introduction: After resuscitation of cardiac arrest (CA), an acute circulatory failure occurs in about 50% of cases, which shares many characteristics with septic shock. Most frequently, supportive treatments are unable to control this shock that may provoke multiple organ failure and death. We evaluated whether an early plasma removal of inflammatory mediators using high permeability hemodialysis (HPHD) in addition to conventional treatments could improve hemodynamic status of this patients.

Patients and methods: We performed a randomized open-label trial. Successfully resuscitated comatose CA patients who had a post-resuscitation shock (defined as requirement of norepinephrine or epinephrine infusion > 1 mg h) were included. The experimental group received 2 separated sessions of HPHD during the first 48 h following ICU admission. The control group received continuous veno-venous hemofiltration (CVVH) if needed. Non-parametric tests were used to compare the two groups. The primary outcome was the duration of the shock expressed by the length of catecholamine infusion. Number of vasopressors-free days, 6-h repeated measures of blood pressure, daily fluid balance and mortality (ICU and day-28) have been evaluated as well.

Results: 35 patients were included: 17 (median age 68.4, 59% male) in the HDHP group and 18 (median age 66.3, 83% male) in the control group. Baseline characteristics did not differ between the two groups. Ventricular fibrillation was the first recorded rhythm in 23.5 and 44.4% in the HDHP and control group, respectively (p = 0.289). Day-28 mortality rate was 64.7 and 72.2% in the HDHP and control group, respectively (p = 0.72). Probability of vasopressors discontinuation over time was similar in the two groups (Figure, p for logrank test = 0.67). Number of catecholamine-free days was 25.1 [0, 26.5] and 24.5 [0, 26.2] in the HDHP and control group, respectively (p = 0.65). No difference was observed regarding the daily-dose of vasopressors and the 6-h recorded systolic mean diastolic arterial pressure. No difference in terms of fluid balance was observed either.

Conclusion: In post-cardiac patients with acute circulatory failure, HPHD did not reduced the duration of shock and had no effect on hemodynamic status. Registration: NCT00780299 the study was completely funded by the French Ministry of Health. Baxter provided the Septex membranes that were used in the HPHD group.

Correspondence: Luyt Charles-Edouard - charles-edouard.luyt@aphp.fr

Annals of Intensive Care 2018, 8(Suppl 1):CO-25

Introduction: Reducing the use of broad-spectrum antibiotics in the ICU is a key issue. The P55 pneumonia cartridge of the multiplex-PCR Unyvero (Curetis) system allows identification of 19 bacteria and 1 fungi among the most frequently responsible for ventilator-associated pneumonia (VAP), and 19 of their resistance markers directly in clinical specimens. We aimed to evaluate the concordance between this technique and the conventional microbiological methods (CMM) for the diagnosis of VAP, assuming that it could support a decrease in broad-spectrum antibiotics consumption.

Patients and methods: Prospective, observational, single centre study. All consecutive patients with suspected VAP and a positive direct examination of broncho-alveolar lavage fluid (i.e. intracellular bacteria on direct examination) were included. Fresh BAL fluid was submitted to CMM and tested with the Unyvero system. We compared the results given by the 2 techniques, CMM being considered as the reference.

Conclusion: A strategy based on direct examination of BAL fluid followed by the P55 cartridge when the former is positive allows rapid identification of pathogen in 80% of patients with VAP. Excluding Pa VAP, mechanism of resistance could be correctly detected in 83% of VAP patients. Interventional studies are warranted to test the usefulness of this technique in an antimicrobial stewardship program. Acknowledgment: Curetis GmbH provided the P55 cartridges.

Correspondence: Sellami Walid - drsellamiwalid@yahoo.fr

Annals of Intensive Care 2018, 8(Suppl 1):CO-26

Introduction: Hemophagocytic lymphohistiocytosis (HLH) is a rare yet life-threatening condition characterized by an inappropriate activation of lymphocytes and or histyocytes leading to an abnormal phagocytosis of blood cells. Prognosis and outcomes mainly depend on the precocity of diagnosis and specific treatment implementation. A few studies were interested in HLH occuring in ICU patients. The Purpose of the study was to describe epidemiological, clinical, paraclinical and therapeutic characteristics of HLH in ICU patients.

Patients and methods: It was a retrospective, descriptive and longitudinal study including 30 cases of HLH, assessed during a 5 years period in the intensive care unit. We included all patients who had evidence of hemophagocytosis in bone marrow smears (realized when HLH was suspected) and a H-score superior to 169.

Results: The mean age of our patients was 48.9 ± 17.6 years [17–80] with a female preavalence. Hypertension and diabetes were the most frequent comorbidities. Immunodepression was present in 5 patients. Shock and neurological disorders were the main causes of admission in ICU. Mean APACHE II and SOFA (admission) scores were respectively 22.6 and 8.2. Fever was the most common clinical presentation in SAM. The most common biological disorder was bicytopenia (anemia and thrombopenia). The mean H-score was 209.7 p. Hemodynamic and respiratory distress were the prevalent organ failures. Corticoïds and immunoglobuline were given respectively to 14 and 13 patients. Etoposid was taken by one patient. Infections were the largely predominant etiology of HLH with a clear prevalence of multidrug resistant bacterial infections. Mortality at day 28 was 40%. Septic shock was the leading cause of death. SOFA and APACHE II scores were the only predictive factors of mortality (p = 0.015 and p = 0.042).

Conclusion: Management a patient with HLH is challenging because of its rarity, its variable presentation and its association with a panel of disorders. A multidisciplinary approach is mandatory to determine the best therapeutic option for the patient.

Correspondence: Themelin Nicolas - nicolas.themelin@ulb.ac.be

Annals of Intensive Care 2018, 8(Suppl 1):CO-27

Introduction: Anemia and RBC transfusion as treatment are common in ICU patients. Neverthless, RBC transfusions are associated with increased morbidity and mortality. Restrictive strategy policy, based on a haemoglobin [Hb] level of 7 g dL, is the guideline to transfuse most of ICU patients. The aim of RBC transfusion is to avoid tissue hypoxia by improving oxygen delivery (DO2) and therefore oxygen consumption (VO2). This could suggest to combine [Hb] and some parameters reflecting the sytemic DO2 VO2 balance to decide to transfuse. This study aims to analyse the effects of RBC transfusion on the systemic oxygenation assessed by the (central) venous oxygen saturation (S(c)VO2), the lactate level, the venous-to-arterial carbon dioxyde pressure gradient (Pv-aCO2) and the ratio between cardiac index and O₂ extraction (IC EO2), and to assess their usefulness in the transfusion decision.

Patients and methods: During 9 months, all adult patients transfused in the ICU of CHU-Charleroi Marie Curie were included except patients with active bleeding or without jugular or subclavicular catheter. The systemic oxygenation parameters have been measured before and after transfusion. Patients a priori have been grouped according to their initial S(c)VO2 (< or > 70%), to their cardiac (< or > 50% of a left ventricular ejection fraction) and septic status. The results are expressed in median and interquartile ranges. Comparisons were made by a Wilcoxon test.

Results: 86 RBC transfusions were made on 53 patients. For all patients, mean arterial pressure, [Hb] and S(c)VO2 increased significantly after transfusion ([Hb]: 7.4 [7–7.8] to 8.3 [7.7–8.8] g dL; MAP: 79 [69–90] to 82 [74–92] mmHg and S(c)VO2: 66 [60–73] to 69 [63–75] %; for all p < 0.001). Only patients with an initial S(c)VO2 < 70% improved it after transfusion (62 [58–65] to 66 [62–71] %; p < 0.001). The Pv-aCO2 was significantly higher in patients with an initial S(c)VO2 < 70% (8 [5–10] mmHg) but did not change after transfusion as [lactate] and IC EO2. These results are maintained whatever the cardiac or septic status.

Conclusion: A S(c)VO2 < 70% in anemic ICU patients compared to other systemic oxygenation parameters, seems to be useful to assess the transfusion requirement as indicator of a DO2 VO2 imbalance. It should be interesting to combine it with the [Hb] in the decision to transfuse anemic critically ill patient.

Correspondence: Sirjacques Camille - Camille.Sirjacques@ulb.ac.be

Annals of Intensive Care 2018, 8(Suppl 1):CO-28

Introduction: There is very little data on the survival of elderly patients with cancer requiring admission to intensive care unit (ICU). The aim of the study is to evaluate in our department prognostic factors for hospital mortality and survival after hospital discharge for patients aged more than 65 years.

Patients and methods: This is a retrospective study including all patients aged more than 65 with a solid or hematological tumor admitted for acute complication in an oncological ICU over a 3 years period. In case of multiple admissions only the 1st admission was used for the statistical analyses.

Results: We recorded 311 admissions for 270 patients over the 3 years period. The main reasons for admission were cardiovascular (22%), respiratory (17%) and hemodynamic (13%). ICU and in-hospital mortality rates were respectively 10% (95% CI 6–13%) and 22% (95% CI 17–27%). Median survival after discharge from hospital was 7.8 months (95% CI 5.7–11.3). The identified prognostic factors for higher hospital mortality were non invasive ventilation use (OR 9.8, 95% CI 3.3–28.5), invasive mechanical ventilation use (OR 8.6, 95% CI 3.4–21.9) and the existence of a therapeutic limitation in the first 24 h (OR 3.1, 95% CI 1.5–6.6) (assessment of all covariates was restricted to the period of 24 h following ICU admission). After hospital discharge, prognostic factors for higher risk of death were Charlson’s score bigger than 8 (HR 2.73, 95% CI 1.93–3.86), SAPS II score bigger than 37 (HR 1.50, 95% CI 1.08–2.09), administration of amines in the first 24 h (HR 2.13, 95% CI 1.03–4.38) and the existence of a life-sustaining therapeutic limitation in the first 24 h (HR 2.35, 95% CI 48–3.73). A total of 77% of patients were able to benefit from an antineoplastic treatment after hospital discharge.

Conclusion: 78% of oncologic patients aged more than 65 years were discharged alive from hospitalization after an ICU stay and the large majority were still able to benefit from cancer treatment. Life-sustaining therapeutic limitation is directly related to hospital mortality and post-discharge survival.

Correspondence: Kerhuel Lionel - lionel.kerhuel@aphp.fr

Annals of Intensive Care 2018, 8(Suppl 1):CO-29

Introduction: Studies assessing outcomes of CNS infections in critically ill immunocompromised patients are scarce. We sought to describe clinical features and outcomes in this population.

Patients and methods: We conducted a retrospective observational study over a 13-year period (01, 2004–08, 2017). All patients admitted to the ICU, presenting with meningitis, encephalitis or brain abscess were included. All patients with the following diseases were considered: hematologic malignancy, solid tumor, solid organ transplantation, autoimmune disease and splenectomy. HIV patients were not included. Baseline characteristics and variables related to ICU admission, treatments, ICU outcomes and neurological status (Rankin scale 6 months after ICU discharge) were collected.

Results: Among the 182 patients admitted to our ICU for CNS infection, 64 were included. HIV patients (N = 45) and non-immunocompromised patients (N = 73) were not included. Among the 64 patients, 40 (63%) had hematological malignancies (in whom 6 received an allogeneic stem cell transplant), 8 (13%) solid tumors, 9 (14%) autoimmune diseases, 7 (11%) renal transplantations and 3 (5%) splenectomies. Thirty-four (53%) were male, with a median age of 55 [47; 64] years and a Performance Status of 1 [0; 2]. The main reason for ICU admission was mental confusion or coma (77%). Lumbar puncture was performed in 57 (89%) patients. CNS CT was performed in 51 patients and was abnormal in 14 (27%). Brain MRI was performed in 40 patients and was abnormal in 34 (85%). Meningitis accounted for 48% (n = 31), in whom 74% were microbiologically documented (Streptococcus pneumonia, Streptococcus spp. and Gram negative bacilli in 8, 3 and 3 cases, respectively). Encephalitis accounted for 31% (n = 20) of diagnoses, in whom 80% were microbiologically documented (tuberculosis, Herpes Simplex Virus and Human Herpes Virus 6 in 5, 3 and 3 cases, respectively). Brain abscess accounted for 20% (n = 13) of diagnoses, in whom 69% were microbiologically documented (Aspergillus and Nocardia in 4 and 3 cases, respectively). Median ICU length of stay was 7 [3; 18] days. Mechanical ventilation was performed in 35 (56%) patients. Vasopressors and renal replacement therapy were implemented in 25 (40%) and 2 (11%) patients respectively. ICU mortality was 32% (n = 20), and 25 (48%) patients were alive 6 months after ICU discharge with a good neurological status (Rankin < 4).

Conclusion: This study sheds light on clinical features and outcomes in immuncompromised patients with CNS infections. Determinants of mortality and comparative data across the main groups of CNS infections are being prepared.

Correspondence: Fourmont Aude-Marie - am.fourmont@gmail.com

Annals of Intensive Care 2018, 8(Suppl 1):CO-30

Introduction: Cardiac involvement in patients with TTP is associated with high mortality. It might be undermined by the lack of operational definition and established diagnostic workup for cardiac TTP. The objectives of this study were to describe cardiac involvement in TTP patients and to assess prognostic value of each clinical sign.

Patients and methods: In a single center study, all adult patients admitted to our ICU between 2007 and 2017 with confirmed TTP (ADAMTS13 < 10%) were included and standardized diagnostic workup was performed. Cardiac signs included chest pain, changes in electrocardiogram, increase troponin level, new-onset changes in cardiac echography, cardiogenic shock, cardiac arrest.

Results: Among the 98 TTP patients (67 women, 65% non-Caucasian, age 43(32–53)), 28 had a cardiovascular comorbidity and 12 were HIV-infected. Half the TTP were idiopathic, 25 autoimmune, 5 drug-related and 4 surrounded pregnancies. At initial clinical evaluation organ involvement was present in all patients, including neurological (N = 77), cardiac (N = 76), renal (N = 50), or digestive (N = 32). All patients received urgent plasma therapy with a median number of plasma exchanges of 11(8–18) and all but 3 received steroids, including 28 who received bolus. 52 patients required a rescue therapy (49 rituximab, 21 vincristine, 4 cyclophopshamide, and 6 splenectomy). 68 patients required antihypertensive therapy, 24 mechanical ventilation, 18 vasopressors and 15 renal-replacement therapy. Cardiac involvement was observed in 89 (91%) patients. Namely, 23 patients exerted chest pain, 57 electrocardiogram changes, 70 (71%) increased troponin, and 6 cardiac arrest. Troponin was the only feature of cardiac involvement in 24 patients. In 13 patients, echocardiography identified focal or global hypokinesia. Anticoagulants were prescribed in 86 patients (58 prophylactic and 28 curative) and 88 patients received low dose aspirin. Hospital mortality was 9.4%. All the 9 patients without cardiac involvement survived. Cardiac involvement was associated with prolonged ICU stay (6 (4–7) vs. 8 (6–14), p = 0.04). By univariable analysis, factors associated with mortality included age, platelet count, status epilepticus, troponin level, cardiac arrest. Of striking finding, cardiac involvement of TTP was not associated with mortality or with adverse TTP outcome.

Conclusion: In patients with severe TTP, cardiac involvement is present in 90% of patients. In this cohort of patients receiving a second-line TTP therapy in half the cases, aspirin in 90% and anticoagulants in 88%, cardiac involvement was not associated with mortality suggesting that aggressive TTP management translates into improved outcomes.

Correspondence: Besset Sébastien - sebastien.besset@aphp.fr

Annals of Intensive Care 2018, 8(Suppl 1):CO-42

Introduction: The use of alternatives to carbapenems to treat patients with extended-spectrum beta lactamase-producing Gram negative bacilli (ESBL-GNB) infections remains controversial. Their use in patients with severe infections in the ICU has been poorly studied. The aim of this study was to compare the outcome of ICU patients having received carbapenems to those having received a carbapenem-sparing agent (CSA).

Patients and methods: The charts of patients with ESBL-GNB infection hospitalized in our ICU between 2015 and 2017 were retrospectively reviewed. Patients treated with betalactam betalactam inhibitor (BL BLI), cefepime or quinolones were considered has having received an alternative to carbapenems (CSA). Patients having received such a CSA were compared to those having received a carbapenems. Primary outcome was treatment failure at day 28, defined as ESBL-GNB infection recurrence (relapse with same pathogen) or death, whichever first occurred.

Conclusion: Treatment of patients with ESBL-GNB severe infection in the ICU with a CSA seems to be safe when the pathogen is susceptible to this CSA. However, MIC should be first determined before de-escalating to a CSA. Larger studies are needed.

Correspondence: Petit Laurent - laurent.petit@chu-bordeaux.fr

Annals of Intensive Care 2018, 8(Suppl 1):CO-43

Introduction: This study aimed to assess whether augmented renal clearance (ARC) impacts negatively on beta-lactam antibiotic unbound concentrations and clinical outcome in critically ill patients.

Patients and methods: Over a 9-month period, all critically ill patients treated by beta-lactam antibiotics for a microbiologically documented infection without renal impairment were eligible. During the first 3 days of antimicrobial therapy, every patient underwent 24-h creatinine clearance (CrCL) measurements and therapeutic drug monitoring. The main outcome investigated in this study was the rate of therapeutic failure, defined as an impaired clinical response with a need for escalating antibiotics during treatment and or within 15 days after end-of-treatment. The secondary outcome was the rate of pharmacokinetic pharmacodynamics (PK PD) target non-attainment defined by at least one sample under four times MIC of the known pathogen (subexposure < 4.MIC).

Results: Over the study period, 96 patients were included. The rate of therapeutic failure was 16%. Mean CrCL values over the first three days of antimicrobial therapy were significantly higher in patients with therapeutic failure (203 ± 63 vs. 163 ± 68 ml min, p = 0.04). Subexposure < 4.MIC was independently associated with therapeutic failure with an OR at 5.8 (1.3–25.4), p = 0.02. Patients with mean CrCL > 200 ml min presented higher risk of subexposure < 4.MIC with an OR at 5.1 (1.8–14.4), p = 0.002. The percentages of samplings which attained the PK PD targets for various CrCL are presented Fig. 1.

Conclusion: Our results suggested that (1) mean CrCL values > 200 ml min were associated with higher rates of sub-exposure < 4.MIC and (2) sub-exposure < 4.MIC was associated with higher rates of therapeutic failure. This study emphasizes the need of therapeutic drug monitoring in patients with ARC, especially when targeting less susceptible pathogens or surgical infections with limited penetration of antimicrobial agents. Whether those patients should benefit for increased dosing regimens should be evaluated by randomized controlled studies.

Correspondence: Roux Damien - damien.droux@gmail.com

Annals of Intensive Care 2018, 8(Suppl 1):CO-44

Introduction: Critically-ill patients often receive antimicrobials. Early adequate antibiotic administration, usually including beta-lactams, improves prognosis of septic patients. However, beta-lactam pharmacokinetic as well as effect of renal replacement therapy (RRT) on beta-lactam concentrations have been hardly explored in ICU patients with acute kidney injury (AKI). We aimed to determine factors associated with potential suboptimal beta-lactam concentration in critically-ill patients with AKI treated either with an early or a delayed RRT strategy.

Patients and methods: Ancillary study in a subset of patients with severe AKI (KDIGO3), receiving a beta-lactam antibiotic, in a trial comparing two RRT initiation strategies. In this trial, patients from 31 intensive care units were randomly assigned to either an early (immediate RRT) or a delayed (late or no RRT) RRT initiation strategy. Beta-lactam residual concentrations were sampled at 24 and 48 h after inclusion. The appropriate concentration was defined as a trough of at least 4 times the minimal inhibitory concentration (clinical breakpoint of EUCAST). The primary outcome was an adequate plasma concentration of the beta-lactam during the first 2 days.

Results: Among the 233 patients included in the 11 centers participating to this ancillary study, a beta-lactam trough concentration was evaluated in 112 subjects, 53 in the early and 59 in the delayed groups. Ninety patients (80.4%) had an adequate beta-lactam dosage. RRT initiation strategy had no impact on beta-lactam concentration (p = 0.78). Among the 83 septic shock patients (74% of the sampled patients), 73 (88%) had a correct antibiotic concentration. In contrast, only 10 of the 16 patients without definite sepsis (62.5%) had a correct dosage. Factors associated with an adequate beta-lactam trough concentration in univariate analysis were admission for a septic shock (p = 0.002), a higher plasma creatinine level (p = 0.024), a higher mean arterial pressure (p = 0.018) and a lower serum bicarbonate level (p = 0.045) at randomization. A higher SOFA score was associated with an adequate beta-lactam concentration near to statistical significance (p = 0.053). Multivariate analysis will be presented.

Conclusion: In the context of severe AKI, beta-lactam concentration reached a sufficient level in 88% of septic shock patients. Interestingly, RRT initiation strategy was not associated with beta-lactam trough concentration. Early RRT did not affect trough concentration of beta-lactam. We may hypothesize that physicians were highly vigilant and adapted antibiotic administration adequately in these patients.

Correspondence: Touchard Cyril - cyriltouchard@hotmail.fr

Annals of Intensive Care 2018, 8(Suppl 1):CO-45

Introduction: Amikacin infusion requires to target a peak serum concentration (C_max) 8–10 times the minimal inhibitory concentration, corresponding to a C_max at 60–80 mg L⁻¹ for the least susceptible bacteria. Recent study reported that 33% of critically ill patients do not attain this target with a 25 mg kg dose (1). Membrane sequestration, alteration of the volume of distribution and lack of data in this population make drugs pharmacokinetics (PK) on ECMO challenging. Our study aimed to assess the prevalence of insufficient Amikacin C_max in critically ill patients on ECMO and to identify relative risk factors.

Patients and methods: Prospective, observational, monocentric study of adult patients on venoarterial (VA) or venovenous (VV) ECMO receiving a loading dose of Amikacin for suspected Gram-negative infections. Intravenous Amikacin was administered with a loading dose of 25 mg kg of total body weight and C_max was measured 30 min after the end of the infusion. Independent predicators of C_max < 60 mg L⁻¹ after the first Amikacin infusion were identified by mixed model multivariate analysis.

Results: From January 2015 to February 2016, 106 patients (median SAPS 2 (interquartile range) 68(47–81); age 55(45–62) years) under VA-ECMO (68%) or VV-ECMO (32%) were included. At inclusion, the SOFA score was 15 (12–18) and 54 (51%) patients were on renal replacement therapy. Overall ICU mortality was 54%. C_max was < 60 mg L⁻¹ in 41 (39%) of the patients. Independent risk factors of amikacin under-dosing were body mass index (BMI) < 22 kg m⁻² (Odds Ratio (OR) 6.38, 95% confidence interval 95%CI 1.8–22.8, p = 0.043) and a positive 24 h fluid balance (OR per 500 mL increment: 1.28, 95%CI 1.05–1.65, p = 0.041) (Fig. 1). Our results were comparable to those observed in patients treated with Amikacin without ECMO (1).

Conclusion: This large prospective study suggests that the prevalence and associated risk factors of Amikacin under-dosing are similar in critically-ill patients with or without ECMO. The use of a 30 mg kg dose in low BMI patients and in those with a positive 24-h fluid balance on ECMO is strongly encouraged to obtain adequate therapeutic targets and prevent therapeutic failure.

Correspondence: Van Grunderbeeck Nicolas - nicovgdb9@orange.fr

Annals of Intensive Care 2018, 8(Suppl 1):CO-46

Introduction: Delirium is frequent in the ICU and has impact on morbidity and mortality. Broad-spectrum beta-lactams (BL) are daily used and may have neurological side effects, notably in cases of overdosing. We aimed to assess impact of BL overdosing on confusion through a validated method, the CAM ICU (Confusion Assessment Method for the ICU) as a primary endpoint, and to look for pharmacokinetic factors associated with respect of the BL therapeutic index as a secondary objective.

Patients and methods: Prospective observational study–pilot study in two centers. Patients treated with broad spectrum BL for more than 48 h to treat sepsis or septic shock in the ICU could be included. Delirium was assessed through CAM-ICU, and determined group of inclusion: delirium patients or control group. Exclusion citeria were RASS inferior or equal to: 3, imipenem, and other obvious causes of neurologic failure (stroke, poisoning, previous neurological disorders). BL concentrations were measured at peak and trough for intermittent and extended perfusion, and at plateau for continuous perfusion, by high performance chromatography coupled with mass spectrometry. Data were expressed as median [25–75%, interquartile range]. Statistical were performed by Mann–Whitney U test and Pearson test. Analysis referred to delirium factors and betalactams’ concentrations, and PK factors (obesity, renal failure, hypoalbuminemia) and overdosing.

Results: Fifty patients were included (29 with delirium, 21 controls), at day 4 for controls and day 5 for patients with confusion. Delirium patients were more severely ill SOFA 5 [3; 9] versus 3[1; 5] (p = 0.05); with higher RASS 1[0; 2] versus 0[0; 0] (p = 0.05). They presented with 41% BL overdosing versus 19% in controls (p = 0.09); with 41% of BL in therapeutic index: 71% in controls (p = 0.04). Obesity and renal failure were not associated with BL overdosing but there was a trend with hypoalbuminemia (p = 0.09).

Discussion: Trend in association of BL overdosing with delirium corresponds to previous studies, and would need a larger scale study to be confirmed. Severity differences in groups would need changes in inclusion criteria to obtain homogeneous groups. A possible association of BL underdosing with poor evolution of infection and organ failures would need more precise evaluation. Hypoalbuminemia could have an impact on BL overdosing.

Conclusion: Delirium was not associated with BL overdosing but with therapeutic index. A high variability of BL concentrations warrants therapeutic drug monitoring. A larger scale study should include changes in design.

Correspondence: Schmidt Matthieu - matthieuschmidt@yahoo.fr

Annals of Intensive Care 2018, 8(Suppl 1):CO-47

Introduction: Extracorporeal carbon dioxide removal (ECCO2R) might allow ultraprotective mechanical ventilation with lower tidal volume (VT) (< 6 mL kg ideal body weight), plateau pressure (Pplat) (< 30 cm H₂O), driving pressure, and respiratory rate (RR) to reduce ventilator induced lung injury (VILI). The aim of this study was to assess the feasibility and safety of ECCO2R with a renal replacement platform (RRT) to permit ultra-protective ventilation in patients with mild to moderate acute respiratory distress syndrome (ARDS).

Patients and methods: Twenty patients with mild (n = 8) or moderate ARDS were included. VT was gradually reduced from 6 to 5, 4.5 and 4 mL kg⁻¹ and PEEP adjusted to reach 23 > Pplat > 25 cm H₂O. Stand-alone ECCO2R (no hemofilter associated on the RRT platform) was initiated when arterial PaCO2 increased by more than 20%. Ventilation parameters (VT, RR, PEEP), respiratory compliance, driving pressure, arterial blood gases, and ECCO2R system operational characteristics (blood flow, sweep gas flow, and CO₂ removal rate) were collected during a minimum of 24 h of ultra-protective ventilation. Complications, mortality at day 28, need for adjuvant therapies and data on weaning from both mechanical ventilation and ECCO2R were also collected.

Results: While VT was reduced from 6 to 4 mL kg⁻¹ and Pplat kept below 25 cm H₂O, PEEP was significantly increased from 13.4 ± 3.6 at baseline to 15.0 ± 3.4 cm H₂O at VT = 4 mL kg⁻¹. As a result, the driving pressure was significantly reduced to 7.9 ± 3.2 cm H₂O at VT = 4 mL kg⁻¹ (p < 0.05) (Fig. 1). No significant differences in RR, PaO2 FiO2 ratio, respiratory system compliance were observed after Vt reduction. Mean extracorporeal blood, sweep gas flow and CO₂ removal were 421 ± 40 mL min⁻¹, 10 ± 0.3 L min⁻¹ and 51 mL min⁻¹, respectively. Mean treatment duration was 31 ± 22 h. Main side effects related to ECCO2R were membrane clotting which occurred in 7 patients after 19 ± 9 h.

Conclusion: A low-flow ECCO2R device driven by a RRT platform efficiently removed CO2 while allowing ultra-protective mechanical ventilation settings in patients with mild to moderate ARDS (ClinicalTrials.gov identifier: NCT02606240).

Correspondence: Morimont Philippe - ph.morimont@chu.ulg.ac.be

Annals of Intensive Care 2018, 8(Suppl 1):CO-48

Introduction: Protective lung ventilation (PLV) is recommended in patients with acute respiratory distress syndrome (ARDS) to minimize additional injuries to the lung. However, increased right ventricular (RV) afterload resulting from ARDS could be enhanced by hypercapnic acidosis resulting from ventilation at lower tidal volume. Relative contribution of these factors (ARDS and PLV) in RV afterload is not clearly established. The aim of this study was to compare RV afterload in ARDS combined with PLV to RV afterload in PLV alone.

Patients and methods: This study was performed in an experimental model of severe hypercapnic acidosis performed in 2 series of 6 pigs. In both groups, respiratory tidal volume was decreased by 60%. In the first group (ARDS group), an ARDS (obtained by repeated bronchoalveolar lavage) was performed before reducing ventilation, while in the second group (control group), hypercapnic acidosis was resulting from low tidal volume ventilation alone.

Results: In both groups, systolic pulmonary artery pressure (PAPs) significantly increased during PLV. This increase was significantly higher in ARDS group than in control group (Fig. 1). Severe hypercapnic acidosis occurred in both groups: PaCO2 increased from 41.7 ± 3.6 to 78.6 ± 8.1 (p < 0.01) and arterial pH decreased from 7.44 ± 0.05 to 7.13 ± 0.04 (p < 0.01) in ARDS group while PaCO2 increased from 37.7 ± 9.4 to 93.6 ± 6.4 (p < 0.01) and arterial pH decreased from 7.48 ± 0.05 to 7.11 ± 0.04 (p < 0.01) in control group. PaO2 significantly decreased in ARDS group (178 ± 42 to 54 ± 12.3 mmHg, p < 0.01) but did not significantly changed in control group.

Conclusion: Isolated hypercapnic acidosis resulting from PLV was clearly responsible for increased RV afterload and this effect was significantly enhanced in ARDS. Pulmonary vasoconstriction resulting from hypercapnic acidosis is strongly enhanced by factors like hypoxia, endothelial injuries or inflammatory mediators in ARDS. Extracorporeal CO₂ removal could be the solution to limit afterload burden on the right ventricle when PLV is achieved during ARDS.

Correspondence: Persichini Romain - romain.persichini@hotmail.fr

Annals of Intensive Care 2018, 8(Suppl 1):CO-49

Introduction: Prone positioning has been shown to improve mortality in acute respiratory distress syndrome (ARDS) patients. The respiratory system driving-pressure (DPRS) and the transpulmonary driving-pressure (DPL), measured with esophageal manometry, have been shown to be strongly correlated with mortality. The aim of this study was to investigate the evolution of the DPL during prone positioning and its relationship with evolution of oxygenation in ARDS patients.

Patients and methods: Ten patients with ARDS equipped with esophageal manometry were enrolled. DPRS, DPL and chest wall driving-pressure (DPCW) were measured before and 1 h after prone positioning. Respiratory system, pulmonary and chest wall elastance (ERS, EL, ECW) were calculated at the same time. Finally, we studied the correlation between these respiratory variables and oxygenation indicators. Patients were classified as responders to prone positioning if the change in the ratio of arterial oxygen partial pressure oxygen inspired fraction (Delta.PaO2/FiO2) induced by the manoeuvre was larger than the median value observed in the group.

Results: In the whole population, median value of Delta.PaO2/FiO2 was 53.5 mmHg, and 5 patients were classified as responders and 5 as non-responders. In responders, DPL significantly decreased from 8.8 ± 4.1 cm H₂O to 5.9 ± 5.4 cm H₂O (p = 0.02) and EL decreased from 21.9 ± 5.1 cm H₂O L to 14.9 ± 6.5 cm H₂O L (p = 0.02) after prone positioning. Other respiratory variables did not change. In non-responders, respiratory variables did not change. Between responders and non-responders, there was no significant difference between baseline respiratory variables. After prone positioning, Delta.PaO2/FiO2 was not related to baseline respiratory parameters. On the contrary Delta.PaO2/FiO2 induced by prone positioning was strongly correlated with changes in DPL (r = − 0.70, p = 0.02) and changes in EL (r = − 0.69, p = 0.03). We did not find any correlation between Delta.PaO2/FiO2 and changes in DPCW or changes in ECW. The correlation between Delta.PaO2/FiO2 and changes in DPRS (r = − 0.56, p = 0.09) and changes in ERS (r = − 0.55, p = 0.10) did not reach significance.

Conclusion: In patients who respond to prone positioning by the highest improvement in oxygenation, DPL significantly decrease after prone positioning. The changes in DPL and the changes in EL play a major role in the improvement in oxygenation induced by prone positioning whereas the changes in DPCW and ECW do not.

Correspondence: Guerin Claude - claude.guerin@chu-lyon.fr

Annals of Intensive Care 2018, 8(Suppl 1):CO-50

Introduction: Whereas prone positioning (PP) has been shown to improve patient survival in moderate to severe ARDS patients, its rate of use was 7.4% in Lung Safe study. However, Lung Safe study was not specifically focused on PP. Therefore, present study aimed to determine prevalence of use of PP in ARDS patients (primary end-point), physiologic effects of and reasons for not using PP (secondary end-points).

Patients and methods: The APRONET study was a prospective international one-day prevalence study performed 4 times in April, July, October 2016 and January 2017. At each study day, investigators had to screen every patient staying in ICU from 0 to 24 h and to fill electronic CRF. For patients with ARDS (defined from the Berlin definition criteria) at each study day oxygenation and ventilator settings were recorded. For those receiving PP these variables were recorded before and at the end of PP session. The reasons for not proning were also collected. Values are presented as median (1st–3rd quartiles). Prevalence rates of PP were compared by using Chi square for trend and groups were compared with nonparametric tests.

Results: In the study period 6723 patients were screened in 141 ICUs from 21 countries, of who 735 with ARDS were analyzed. Over the four study days, one-hundred and two ARDS patients received at least one proning session (13.2%). The prevalence of prone positioning in ARDS patients was not significantly different between study days: 13.8% (33 240), 12.6% (18 143), 15.3% (24 157) and 13.8% (27 195) (p = 0.83, Chi square for trend). Over the four study days merged, the rate of PP use was 5.9% (11 187), 10.3% (41 399), 32.9% (49 149) in mild, moderate and severe ARDS, respectively (p = 0.0001). The duration of the first PP session was 18 [16–23] hours. Between supine before PP and end PP, PaO2 FIO2 ratio significantly increased from 101 [76–136] to 171 [118–220] mmHg (p = 0.0001), driving pressure significantly decreased from 14 [11–17] to 13 [10–16] cm H₂O (p = 0.001). The main reason for not proning was not severe enough hypoxemia (472 734 reasons for not proning 633 ARDS patients, 64.3%).

Conclusion: Present study found a higher rate of PP use in severe ARDS patients than in the Lung Safe study, which could reflect change in practice or ICU selection bias.

Correspondence: Hraiech Sami - sami.hraiech@ap-hm.fr

Annals of Intensive Care 2018, 8(Suppl 1):CO-51

Introduction: Although acute respiratory distress syndrome (ARDS) has been largely focused on, few data are available concerning hypoxemia independently of its cause in intensive care unit (ICU) patients. A recent prevalence-point-day (PPD) evaluated the patterns and outcomes of hypoxemia in French speaking ICUs. Here, we describe the main etiologies, management and outcomes of the patients of this cohort presenting with severe hypoxemia.

Patients and methods: A PPD was conducted among 117 French speaking ICUs during spring 2016. Hypoxemia was defined by a PaO2 FiO2 ratio below 300. We analyzed the data from patients with severe hypoxemia (i.e. with a PaO2 FiO2 ratio < 100) and compared their characteristics (causes of hypoxemia, ventilatory and non-ventilatory management) and outcomes to the patients with mild or moderate hypoxemia.

Results: Among the 859 hypoxemic patients the day of the study, 74 (9%) had severe hypoxemia. The main cause of hypoxemia was pneumonia and this diagnosis was more frequent than in mild and moderate hypoxemia. Whereas bilateral radiologic infiltrates were present in 56 (84.9%) patients, ARDS was diagnosed by physicians in only 30 (40.5%) of them. Invasive mechanical ventilation (MV) was used in 55 (74.3%) patients. High flow oxygen was administered in 11 (14.9%) of them and 8 (10.8%) were under non-invasive ventilation (NIV) the day of the study. Median Vt was 6.1 (4.8–6.6) ml kg of IBW. Positive end-expiratory pressure (PEEP) was higher than in mild and moderate hypoxemic patients (10 (8–12) vs. 5 (5–8) and 7 (5–10) cm H₂O respectively, p < 0.001). Median plateau pressure was 25.5 (23–29.5) and was higher than in mild and moderate groups. Median driving pressure was 13 (10–19) cm H₂O with no difference when compared to other groups. Neuromuscular blocking agents were administered in 28 (51.9%) patients, inhaled nitric oxide (iNO) in 7 (13%) patients and only 6 patients (11.1%) were on prone positioning. Fourteen (18.9%) patients were under extracorporeal membrane oxygenation (ECMO). ICU mortality was higher in severe hypoxemic patients as compared to mild and moderate (50.7 vs. 21.3 and 28.5% respectively, p < 0.001). ICU length of stay in ICU survivors was not statistically different between groups.

Conclusion: Severe hypoxemia, independently from ARDS, worsens the prognosis of ICU patients. Even though ARDS might be underdiagnosed, a protective ventilation was respected in severe hypoxemic patients.

Correspondence: Llitjos Jean-François - jllitjos@gmail.com

Annals of Intensive Care 2018, 8(Suppl 1):CO-57

Introduction: Major changes in septic shock management raise the questions of the relevance of the classical risk factors of nosocomial infections in the current era and the link with the primary infectious insult. We herein investigated the risk factors and the outcomes of ICU-acquired infections in a recent cohort of septic shock patients.

Patients and methods: This was a 9-year (2008–2016) monocenter retrospective study. All adult patients diagnosed for septic shock within the first 48 h were included. Septic shock was defined as a microbiologically proven or clinically suspected infection, associated with acute circulatory failure requiring vasopressors. Patients who survived the first three days were eligible for assessment of the risk of the first ICU-acquired infections. The diagnosis of nosocomial infections were based on current international guidelines. Patients were classified according to the development of pulmonary or non-pulmonary ICU-acquired infections. The determinants of ICU-acquired infections were addressed in a multivariate logistic regression analysis.

Results: 938 patients were admitted for septic shock. 788 patients remained alive in the ICU after the first three days and could then be evaluated for the risk of ICU-acquired infections. Hence, 554 patients remained free of secondary infections, 138 patients first developed an episode of nosocomial pneumonia and 96 patients first developed an episode of non-pulmonary infection. The mortality rates of patients with ICU-acquired pneumonia, non-pulmonary ICU-acquired infections and without secondary infections were 49, 49 and 14%, respectively (p = 0.0001). In multivariate analysis, the development of ICU-acquired pneumonia was independently associated with male gender (OR 2.23, CI 95% [1.28–3.48], p = 0.003), renal replacement therapy (OR 2.15, CI 95% [1.35–3.42], p = 0.001), platelet transfusion (OR 2.38, CI 95% [1.4–4.04], p = 0.001) and a primary pulmonary infection (OR 8.06, CI 95% [2.69–24.12], p < 0.001). The development of non-pulmonary infections was independently associated with renal replacement therapy (OR 4.6, CI 95% [2.7–7.85], p < 0.001), fresh frozen plasma transfusion (OR 2.49, CI 95% [1.42–4.37], p = 0.001), healthcare-associated septic shock (OR 1.8, CI 95% [1.1–2.91], p = 0.01).

Conclusion: ICU-acquired pneumonia occurs preferentially in patients with septic shock of pulmonary origin. In addition, we identified the transfusion of blood products as a risk factor for pulmonary and non-pulmonary nosocomial infections.

Correspondence: Hariri Geoffroy -geoffroyhariri@hotmail.com

Annals of Intensive Care 2018, 8(Suppl 1):CO-58

Introduction: Human serum albumin is used for the restoration of blood volume, emergency treatment of septic shock patients. Several experimental studies suggested that albumin could have additional protective effects on the vascular wall and more specifically on endothelial functions. However, the in vivo effect of albumin in human endothelium remains unknown. The aim of this study is to assess the effect of albumin or saline infusion on skin endothelial function in septic shock patients requiring volume expansion.

Patients and methods: We performed a prospective randomized monocentric study in an 18-bed medical intensive care unit. All patients with septic shock who required fluid administration were included between H6 and H24 after vasopressor starting. Patients were randomized to receive either 500 ml of saline solution 0.9% or 100 ml of albumin 20%. Norepinephrine dose was not modified 1 h before and during the procedure. Endothelium-dependant vasodilatation in the skin circulation was assessed by iontophoresis of acetylcholine before and after fluid administration. The improvement of skin blood flow in response to acetylcholine after fluid administration was compared between groups. For each patient, age, sex, SAPS II, site of infection, global hemodynamic parameters and clinical microcirculatory parameters were recorded. Results are expressed as mean ± SD. Qualitative data were compared using Chi-2 or Fisher’s exact test while quantitative data comparisons used Student t Test or Mann–Whitney as appropriate.

Conclusion: In the early stage of septic shock resuscitation, we showed that albumin infusion had protective endothelial effects. This result has to be confirmed in a larger cohort.

Correspondence: Maenhout Christelle - christelle.maenhout@gmail.com

Annals of Intensive Care 2018, 8(Suppl 1):CO-59

Introduction: Encephalopathy occurs in 75% of patients with sepsis and is associated with short-term mortality and long-term cognitive disability among survivors. It is currently unclear what is the gold standard to detect SAE, how it can be prevented, and how evolution to long-term disability can be prevented. The aims of this study were to determine which clinical scales are altered after sepsis and if acute encephalopathy is associated with short-term cognitive impairment.

Patients and methods: We prospectively included adult patients with sepsis between January 2016 and September 2017. Exclusion criteria were encephalopathy from another etiology, psychiatric disorder, recent sepsis or cardiac surgery. We assessed consciousness twice daily (with the Confusion Assessment Method for the ICU [CAM-ICU], Glasgow Coma Scale [GCS], Full Outline of UnResponsiveness [FOUR], Coma Recovery Scale-Revised [CRS-R], and Reaction Level Scale 85 [RLS85]). All patients received continuous EEG monitoring for up to 7 days after admission. We assessed encephalopathy using a modified Synek scale. After discharge from ICU, we assessed cognitive functions (with the Montreal Cognitive Assessment [MOCA], the Frontal Assessment Battery [FAB], and the Language Screening Test [LAST]). A healthy control group of 18 adults was used for comparison.

Conclusion: In this study, EEG was more sensitive than clinical tools to detect SAE but clinical scales correlated with the EEG grade. Encephalopathy was not associated with short-term cognitive function. Further study and a larger cohort are needed to determine which early EEG features can identify patients who will develop short-term and long-term cognitive dysfunction.

Correspondence: Crippa Ilaria Alice - ilaria.alice.crippa@gmail.com

Annals of Intensive Care 2018, 8(Suppl 1):CO-60

Introduction: Sepsis associated encephalopathy (SAE) is associated with increased morbidity and mortality. Its pathophysiology is incompletely elucidated but a possible impairment of cerebral autoregulation (CAR) could result in brain hypoperfusion and neuronal damage (1, 2). We studied CAR in septic patients to test the hypothesis that its alteration is associated with SAE.

Patients and methods: We studied 96 adult patients with sepsis (July 2015–August 2017). Exclusion criteria were—intracranial disease; arrhythmias; extracorporeal membrane oxygenation; and any supra-aortic arteriopathy. Transcranial Doppler (DWL, Germany) was performed, insonating the left middle cerebral artery (LMCA) with a 2 MHz probe. LMCA blood flow velocity (FV) and arterial blood pressure (BP) signals were simultaneously recorded; Pearson´s correlation coefficient between BP and FV (Mxa) was calculated using MATLAB (MathWorks, USA). Impaired CAR was defined as Mxa > 0.3 (3). All data are reported as median [IQR], n (%).

Conclusion: CAR was altered in nearly half of the patients with sepsis. Mxa was higher in patients with than without SAE.

Correspondence: Sivanandamoorthy Sivanthiny - sivanthiny.sivanandamoorthy@aphp.fr

Annals of Intensive Care 2018, 8(Suppl 1):CO-61

Introduction: There is growing evidence that corticotherapy improves survival from septic shock. This observational study aimed at evaluating at bedside resistance to corticosteroids in adults with sepsis.

Patients and methods: Participants—ICU adults with septic shock or without sepsis admitted to the Raymond Poincaré university hospital. We also evaluated 10 healthy controls. Intervention—resistance to corticosteroids was assessed using a skin test. 10 µl of dermocorticoid cream (class III, Betamethasone) was applied on a 3 cm² surface of the skin. At 24 h, two independent physicians scored the blanching of the skin from 0 to 4—0- no blanching + 1- < 50% of surface + 2- 50 to 75% of surface + 3- 75 to 100% of surface, and 4- blanching beyond application area. Cohen’s kappa was used to measure concordance. A mean score of < 2 indicated corticoresistance and a score of 4 indicating normal sensitivity to corticosteroids. We also performed a 250 µg ACTH test.

Results: We enrolled 110 patients, 47 patients with septic shock (22 males, SAPSII 34 [6–114]) and 63 patients without sepsis (28 males, SAPSII 30 [6–100]). Overall, 81 95 (85%) with two measurements patients had concordant evaluation of score by the two physicians + while 14 had a difference of 1-point in scores, resulting in a kappa of 0.78 (95% CI 0.65–0.91). In patients with septic shock, 32 (78%) have corticoresistance, i.e. a mean score < 2, 8 (20%) a score of 2 or 3, and 1 (2%) has normal sensitivity to corticosteroids. In non-septic critically ill, 30 (56%) have corticoresistance, 20 (37%) a mean score of 2–3, and 4 (7%) have normal sensitivity to corticosteroids. Hence, as compared to non-septic patients, patients with septic shock were more likely to have corticoresistance (p = 0.04).

Discussion: Topic application of corticosteroids on the skin results in activation of glucocorticoid receptors present within the vessels. Subsequently, activation of lipocortin 1 may inhibit the activity of phospholipase A2, regulator of prostaglandins, leucotrienes and platelet activating factor. Then, the coupling of alpha adrenoreceptors to their agonists is potentiated, increasing vessels smooth muscles sensitivity to catecholamines. The subsequent local vasocontriction is reflected by skin blanching. Thus, the observed lack of skin blanching in septic patients may reflect altered coupling between gluocorticoids and glucocorticoids receptors.

Conclusion: Roughly one out of two adults with septic shock may develop a resistance to corticosteroids as assessed by a skin blanching test in response to betamethasone.

Correspondence: Daix Thomas - thomas.daix@wanadoo.fr

Annals of Intensive Care 2018, 8(Suppl 1):CO-62

Introduction: Mild therapeutic hypothermia, currently recommended in the management of cardiac arrests with shockable rhythm could promote infectious complications and especially ventilator-associated pneumonia (VAP) (Mongardon et al. Crit Care Med 2011). Despite high incidence of VAP and retrospective trials suggesting a benefit of short-term (48 h) antibiotics in this setting (Davies et al. Resuscitation 2013), systematic use of antibiotic prophylaxis is not recommended in patients treated with mild therapeutic hypothermia after cardiac arrest. The primary objective was to demonstrate that systematic short-term antibiotic prophylaxis with amoxicillin-clavulanic acid can reduce incidence of early VAP (< 7 days) in patients treated with mild therapeutic hypothermia after out-of-hospital cardiac arrest. Secondary objectives were its impact on incidence of late VAP and on Day 28 mortality.

Patients and methods: Multicenter two parallel-group double-blinded randomized trial. Adult patients hospitalized in ICU, mechanically ventilated after out-of-hospital resuscitated cardiac arrest related to initial shockable rhythm and treated with mild therapeutic hypothermia were eligible. Exclusion criteria were pregnancy, need for extracorporeal life support, ongoing antibiotic therapy or pneumonia, known chronic colonization with multiresistant bacteria, known allergy to beta-lactam antibiotics and moribund patients. Patients received either intravenous injection of amoxicillin-clavulanic acid (1 g 200 mg) or placebo three times a day for 2 days. The primary endpoint was the onset of early VAP. All suspected pulmonary infections were adjudicated by a blinded independent committee.

Results: Out of 198 patients included, 196 were finally analyzed, 99 in treatment group and 97 in placebo group (mean age 60.5 ± 14.4 years, sex ratio = 4, SOFA score 8.7 ± 3.1). Characteristics of cardiac arrest were similar in both groups (no flow = 3.5 ± 4.8 min vs 3.8 ± 4.0 min, low-flow = 21.8 ± 13.7 min vs 18.2 ± 10.1 min). 51 early VAP were confirmed, 19 in treatment group vs 32 in placebo group, with an incidence of 19.2 vs 33.0%, respectively (HR = 0.546 + IC 95% = [0.315 + 0.946], p = 0.031) (Fig. 1). The procedure did not affect occurrence of late VAP (> 7 days), respectively 4 vs 5. Day 28 mortality was similar in both arms (41.4 vs 37.5%, p = 0.58) and no adverse event was related to study treatment.

Conclusion: Short-term antibiotic prophylaxis with amoxicillin-clavulanic acid significantly decreases incidence of early VAP in patients treated with mild therapeutic hypothermia after out-of-hospital cardiac arrest related to shockable rhythm.

Correspondence: Grimaldi David - david_grimaldi2001@yahoo.fr

Annals of Intensive Care 2018, 8(Suppl 1):CO-63

Introduction: Immunosuppressed (IS) patients are prone to develop respiratory failure and to need ventilatory support. Invasive ventilation shared a grim prognosis in the past and non-invasive ventilation had been recommended in these patients, however NIV efficacy has been recently challenged and the advent of high flow oxygen therapy had brought even more complexity in the management of such patients. Using the data from a recent point-prevalence-day of hypoxemia in ICU, we compare the frequency, management and outcomes of hypoxemia in IS and immuncompetent (IC) patients.

Patients and methods: The SPECTRUM study was conducted in 117 French-speaking ICUs in 7 countries during spring 2016. IS was retained in case of malignant hemopathy, HIV positivity, immunosuppressive drugs, recent chemotherapy, neutrophil count < 0.5 G L. Hypoxemia was defined as a PaO₂ FiO₂ ratio > 300 and separate into severe (> 100), moderate (> 200) and mild (> 300). We focused on the causes of hypoxemia, the ventilatory management and the outcome.

Results: Among the 1604 patients included, 187 (12%) were IS out of whom 99 (53%) were hypoxemic, proportion similar to the IC patients. Mean age and IGS-2 of hypoxemic patients were similar in IS and IC patients. Hypoxemia was mild in 46 (46%), moderate in 43 (43%) and severe in 10 (10%) IS patients with a similar distribution compared to hypoxemic IC patients. The causes of hypoxemia were also similar pneumonia being the leading cause. 24 (24%) hypoxemic IS patients fulfilled the Berlin criteria for ARDS in a similar proportion to IC patients. Respiratory support used in hypoxemic IS patients was ambient air in 3, low flow oxygen in 24, high flow in 11, NIV in 5 and invasive ventilation in 56 patients, with a different distribution from the IC patients (more patients on high flow therapy and less invasively ventilated). The day of the study, thoracic CT scan and echocardiography were performed in a similar proportion in IS and IC patients whereas broncho-alveolar lavage was more frequently performed in IS patients (14 vs 6%, p < 0.001). Finally, as expected, ICU mortality was higher in hypoxemic IS patients (38 vs 25%, p < 0.01).

Conclusion: Immunosuppression in the ICU seems not to be associated with hypoxemia, severity of hypoxemia or ARDS. Oxygenation management is slightly different from immunocompetent patients with more frequent use of high flow therapy.

Correspondence: Milesi Christophe - c-milesi@chu-montpellier.fr

Annals of Intensive Care 2018, 8(Suppl 1):CO-64

Introduction: HFNC is currently proposed as first-line respiratory support in moderate to severe acute viral bronchiolitis (AVB) in infants. However, the flow setting remains empiric, 2 L/kg/min being used by most teams. Considering the failure rate observed with this device, as high as 50% in some studies, we hypothesized that a higher flow rate may improve this issue. The purpose of the present study was to compare the failure rates with two flow regimen-2 L/kg/min versus 3 L/kg/min.

Patients and methods: A randomized controlled study was performed in 16 French Pediatric Intensive Care Units (PICUs). Infants younger than 6 months-old with moderate to severe AVB, defined by Wood-modified Clinical Asthma score (mWCAs) > 3, were randomly allocated to HFNC treatment with a flow rate of 2 L/kg/min or 3 L/kg/min for 48 h. The primary endpoint was the percentage of failure, defined as the occurrence of one or more of the following—increase in mWCAs or RR, increase in discomfort (EDIN score), and severe apnea episodes.

Results: 287 infants with mean (SD) age and weight of 47 (58) days and 4460 (1130) g were included from November 2016 to March 2017. At baseline, RR was 58 (16) rpm, mWCAs 4.5(1), FiO2 32 (13) %, PCO2 59 (13) mmHg, pH 7.26 (0.1). 142 were included in the 2 L/kg/min group and 145 in the 3 L/kg/min group. No difference was observed between groups for baseline characteristics. Failure rate was not different between groups—38.7 vs 39.3% + p = 0.92. No center effect was observed for failure. Discomfort was more frequent in the 3 L kg min group—7 vs 17% + p = 0.006. The length of stay was shorter in the 2 L kg min group—5.3 (2.8) vs 6.4 (5) days + p = 0.048. Intubation occurred in 4 142 patients in the 2 L/kg/min group vs 10 145 patients in the 3 L kg min group (p = 0.12).

Conclusion: HFNC with a flow rate of 3 L/kg/min did not reduce the risk of failure compared to 2 L/kg/min at the initial respiratory management of AVB in young infants.

Correspondence: Levy Bruno - blevy@sfr.fr

Annals of Intensive Care 2018, 8(Suppl 1):CO-65

Introduction: Despite the frequent use of vasopressors which are administered in 90% of patients in cardiogenic shock (CS), there is only limited evidence from randomized trials comparing vasopressor in CS. Hence, the OPTIMA CC study was designed to compare epinephrine and norepinephrine in cardiogenic shock following myocardial infarction.

Patients and methods: Multicenter, double-blind, randomized trial in 8 french ICU. Cardiogenic shock patients due to myocardial infarction treated by PCI were randomized to receive epinephrine or norepinephrine to maintain MAP at 70 mmHg. Dobutamine was introduced at the physician discretion according to a combination of parameters—echocardiographic parameters, cardiac index, lactate clearance, SVO2 and Swan-Ganz derived parameters.

Results: 56/57 patients were ventilated (98%). There were no differences in the duration nor in the maximal dose or cumulated dose of epinephrine or norepinephrine. Dobutamine was used in 18/27 (67%) in the epinephrine group and in 20/30 (67%) in the norepinephrine group. There were no differences in the duration, in the maximal or cumulated dose. Arterial pressure evolution was similar. Heart rate increased significantly in epinephrine group and did not change in norepinephrine group. Cardiac index and cardiac power index increased significantly more in the epinephrine group than in the norepinephrine group. Cardiac double product, a surrogate of myocardial oxygen consumption increased in epinephrine group and did not change in norepinephrine group. Epinephrine use was associated with a lactic acidosis from H2 to H24 while arterial pH increased and lactate level decreased in norepinephrine groupEpinephrine was significantly associated with an higher incidence of refractory shock—10/27 (37%) versus 2/30 (7%) p = 0.008). The incidence of arrhythmia was identical (epinephrine—41% versus norepinephrine—33%, p = 0.59). ECMO was used in 3/27 (11%) in the epinephrine group and in 1 30 (3%) in the norepinephrine group (p = 0.34) Mortality was 11/30 (37%) in the norepinephrine group and 14/27 (52%) in the epinephrine group (p = 0.25) Epinephrine use was associated with a trend to an increased risk of death (p = 0.08) and an increased risk of death plus ECMO (p = 0.031) at 7 days. There was a trend for an increased risk of death plus ECMO at J28 (p = 0.064).

Conclusion: In patients with cardiogenic shock following myocardial infarction, epinephrine use was associated with a lactic acidosis, an higher incidence of refractory shock and an increased risk of death plus ECMO at J7.

Correspondence: Javouhey Etienne - etienne.javouhey@chu-lyon.fr

Annals of Intensive Care 2018, 8(Suppl 1):CO-67

Introduction: Superantigen toxins synthesized by S. aureus or by S. pyogenes are responsible for toxic shock syndromes (TSS) which lethality can reach 28%. High dose intravenous immunoglobulins (IVIG), able to neutralize these toxins, are frequently used even tough evidence of its efficacy is not supported by randomized controlled study (RCT). Moreover, IVIG are expensive and possibly harmful. Before conducting a RCT, a pilot study was first designed to assess the feasibility in the context of pediatric critical care.

Patients and methods: A double blinded RCT was performed comparing 2 g kg of IVIG to isovolumic 4% albumin perfusion within the first 12 h of TSS in children aged between 1 month to 17 years. A priori criteria to determine the feasibility were defined as a rate of inclusion among eligible patients > 50%, a rate of protocol’s deviations < 30% (treatment delivery, non-respect of blinding, premature stop), and by the practical and financial aspects of the protocol. Secondary objectives were to assess the efficacy of IVIG on organ dysfunction (using Pelod-2 score), on mortality at day 60 and their safety. The study was promoted by the Hospices Civils of Lyon, approved by the CPP Sud-Est and registered at clinical trial (NCT02219165). Inform consent from both parents was required before randomization. This study was funded by CSL-Behring company.

Results: During the 30 months study period, 21 patients were included in 9 centers. The inclusion rate was of 71% (5 parent’s refusals, 3 parents were absent at admission). Two patients were wrongly included (pneumococcal shocks), one patient didn’t receive the treatment because he was transferred for ECMO in a non-investigator center, three patients were treated after 12H, and in two patients one bottle of treatment was missing. The blinding was well respected. Missing data on the Pelod2 score and mortality was lower than 10%, and no premature stop was reported. The eCRF completion was judged easy by investigators. The inclusion of children within the first 12 h was judged challenging. The treatment delivery had to be improved, requiring the help of research assistants. Seven serious and one severe adverse events were registered, all patients recovered and no death was reported.

Conclusion: This pilot study suggested that a RCT is feasible. It provides crucial information to improve the recruitment, the respect of the protocol and the correct measure of organ failure. However, inclusion of international centers is necessary to attain the sample size required.

Correspondence: Mortamet Guillaume - mortam@hotmail.fr

Annals of Intensive Care 2018, 8(Suppl 1):CO-68

Introduction: Indirect calorimetry is a non invasive tool to measure oxygen consumption (VO₂) and resting energy expenditure at bedside. The aim of the present study was to assess the validity of the indirect calorimetry-based method for the work of breathing assessment when compared to esophageal pressure (Pes) measurement and Electrical Activity of the Diaphragm (EAdi) during a spontaneous breathing trial in continuous positive airway pressure.

Patients and methods: A prospective single center study. All intubated and mechanically ventilated children > 1 months and < 18 years old, hospitalized in the pediatric intensive care unit were eligible. Patients considered as ready to extubate were included. Simultaneous recordings of VO2, Pes and EAdi were performed during 3 steps: before, during and after the spontaneous breathing test in continuous positive airway pressure.

Results: Twenty patients, median 5.5 months, were included. Half of the patients were admitted for a respiratory reason. Predicted resting energy expenditure was overestimated as compared to measured resting energy expenditure (51 [47–55] vs 23 [21–29] kcal kg day, p < 0.001). Spontaneous breathing test was associated with an increase in Esophageal Pressure–Time Product from 23 [5–89] to 83 [24–110] cm H₂O s min. The same trend was observed in respiratory drive, assessed by EAdi which increased from 7.5 [3.2–16.3] to 15.9 [5.2–22.1]. Oxygen consumption obtained by IC was higher during spontaneous breathing test as compared to conventional ventilation (3.8 [3.0–5.2] vs 3.6 [3.1–4.6] ml kg min) but non significantly. Changes in work of breathing as assessed by VO₂ was poorly correlated with measurements from Pes and EAdi whereas we found a moderate correlation between Pes and EAdi values. Spontaneous breathing test and extubation were successful in 20 (100%) and 18 (90%) patients, respectively.

Conclusion: During weaning from mechanical ventilation, spontaneous breathing test in continuous positive airway pressure induced an increase in work of breathing, both in respiratory drive, as measured by EAdi and in respiratory mechanics, as measured by Pes. Oxygen consumption measured by Indirect Calorimetry does not seem to be a reliable tool to assess work of breathing in mechanically ventilated children.

Correspondence: Ben Gheriba Khalil - bg.khalil@gmail.com

Annals of Intensive Care 2018, 8(Suppl 1):CO-69

Introduction: During the winter season 2015–2016 we had evaluated breastfeeding disruption after hospitalization for bronchiolitis in our hospital in infants under 6 month (n = 84). We observed 51% of mothers whose breastfeeding was stopped of modified. Clinical severity had no impact on breastfeeding but 63% of mothers stated that lack of support and advice was the first cause of breastfeeding disturbance. We conducted this second phase to evaluate the potential impact of actions to promote breastfeeding on unwanted weaning during hospitalization for bronchiolitis.

Patients and methods: This is a cross sectional study during two epidemic seasons of bronchiolitis in a tertiary care hospital. All patients aged 6 months or younger hospitalized with acute bronchiolitis and receiving at least partial breastfeeding were eligible for the study. Patients discharged at home whose parents accepted to be contacted by phone were included. A bundle of actions to promote breastfeeding in patients with bronchiolitis was implemented (posters, flyers, staff training, equipment with breast pumps) between the two epidemic seasons. The data was extracted from the charts and from a phone survey two weeks after discharge to evaluate breastfeeding in eligible patients in our hospital. Phase I (before action) had included 84 patients hospitalized between December 2015 and March 2016 in all wards hosting patients with bronchiolitis. Phase II (after action) included 50 patients hospitalized from October 2016 to December 2016. The data from phase II was compared with data from phase I.

Results: Fifty patients could be included during the second step of the study, with a mean age of 37 days. Breastfeeding was exclusive for 78% of mothers (vs 77% in phase I). The median length of stay was 6 days (vs 3 days in phase I). Twenty-one (42%) patients spent time in PICU vs. 32% in phase I, 3 21 needed intubation, 18 21 received non invasive ventilation for a median length of 4 days (vs. 3 days in phase I). The number of patients needing nutritional support was 35 50 (70%) during phase II vs. 44 84(52%) during phase I. After implementation of our actions, 40 50 (81%) mothers kept breastfeeding as before (vs. 49% in the previous epidemic season, p < 0.05), 4 mothers (8%) stopped, 4 (8%) switched to partial breastfeeding and 2 (4%) reduced without stopping.

Conclusion: Bronchiolitis is a high risk event for breastfeeding disruption but staff training and correct advices and support for mothers during hospitalization seems to diminish that risk.

Correspondence: Nardi Nicolas - nicolas.nardi@hotmail.fr

Annals of Intensive Care 2018, 8(Suppl 1):CO-70

Introduction: Pediatric acute respiratory distress syndrome (PARDS) is frequent in pediatric intensive care units (PICU), potentially lethal and the diagnosis is often missed or delayed (PALICC 2015). In PICU, data are usually recorded between 5 to 60 min which leads to only a minority of the arterial partial pressure of oxygen (PaO2) that are usable to calculate a valid oxygenation index (OI). If not available, PaO2 should be replaced by the SpO2 if < 98% to calculate the oxygen saturation index (OSI). Using a high temporal resolution (HTR) database that records data every 5–30 s, we aim to develop a relevant clinical algorithm of mass data aggregation to improve PARDS screening with the automatic OI and OSI calculation.

Patients and methods: All the patients admitted to our Pediatric ICU between May 2015 and August 2017 were included. The HTR and the electronic medical records (EMR) were queried through Structured Query Language (SQL) following these steps—(1) Data selection (2) Extraction to a linear format (3) Date and time synchronization (4) Data pivoting (5) Aggregation through a 10-min moving average (6) Hypoxemia calculation. Statistical analysis included proportions, correlations and Bland–Altman analysis.

Results: Between May 2015 and August 2017, 1793 patients (2210 stays) were admitted to the PICU. Approximately 46 million rows were retrieved from the databases including 19,189 PaO2 values. The algorithm was able to calculate 11,320 (59% of the PaO2) OI and 5204 OSI. The comparison between OI and OSI showed that 97.4% of the results were between the limits of agreements (− 17.2 + 10.6), a bias of − 3.3 and a correlation R2 = 0.643. The comparison between the OIs from the HTR and EMR databases showed that 94.6% of the results were between the limits of agreements (− 5.47 + 5.39), a bias of − 0.04 and R2 = 0.904.

Conclusion: Using a mass data aggregation algorithm on a HTR database allows more PaO2 to be used to calculate an OI than the usual EMR. The OI results differ slightly between the HTR and the EMR. The accuracy is probably in favor of the HTR because of the shorter time-lapse between the OI parameters. The OSI is possibly a biased OI surrogate and should be interpreted with caution. Our next step will be to measure the impact of the algorithm on the PARDS real-time diagnosis and PARDS severity categories.

Correspondence: Chosidow Anaïs - anaischosidow@gmail.com

Annals of Intensive Care 2018, 8(Suppl 1):CO-71

Introduction: Early administration of appropriate antibiotic therapy with adequate concentration is the cornerstone of the severe sepsis and septic shock’s treatment. Adult studies showed alteration of distribution and elimination which can lead to insufficient drug concentration in septic patients. In children, studies are lacking and antibiotic dosing may be suboptimal. We aim to describe the plasma concentration of the most used beta-lactam in critically ill children, to describe the rate of patients with suboptimal exposure and associating clinical and biological factors.

Patients and methods: This was a prospective, single center, observational study designed in 32 beds Pediatric Intensive Care Unit (PICU) and high dependency care at the Necker Hospital (Paris, France) from January 2016 to May 2017. Were included, children with severe sepsis or septic shock, aged less than 18 years and weighing more than 2.5 kg, and receiving one or more of the following antibiotics—amoxicillin, cefotaxime, cefazolin, ceftazidime, piperacillin-tazobactam, meropenem and imipenem for suspected or proven infection. Beta-lactam plasma concentrations were analysed using High Performance Liquid Chromatography.

Results: We enrolled 37 children (severe sepsis, n = 22 (59.5%) + septic shock, n = 15 (40.5%)) with a median age of 19 months (4–64. Bacteria were identified in 26 patients (70.3%). A total of 132 blood samples were analysed at a median of 2 days (1–12) following the onset of sepsis. Twenty-four patients (64.8%) had insufficient concentration (cefotaxime 7 14 (43%) + piperacillin-tazobactam, 10 13 (77%) + amoxicillin 6 7 (86%) + meropenem 3 6 (50%), cefazoline 1 4 (25%), imipenem 0 2 (0%) + ceftazidime 0 1 (0%)). Insufficient concentrations were associated with early measurements (< 72 h from the sepsis’ onset) (p = 0.035) and creatinine clearance increase (p = 0.01). Adequate concentrations were associated with small age (p = 0.048).

Conclusion: In conclusion, current standard beta-lactam dosing in children with severe sepsis or septic shock could be inadequate to reach the target concentrations. That could lead to the risk of clinical and bacteriological failures as well as the emergence of bacterial resistance. Further pharmacokinetic studies are mandatory to improve antibiotic therapy in this vulnerable population.

Correspondence: Bigé Naïke - naikebige@gmail.com

Annals of Intensive Care 2018, 8(Suppl 1):CO-72

Introduction: Intermittent hemodialysis is a key support therapy in ICU. Despite protocol-based optimization, intradialytic hemodynamic instability (IHI) remains a common complication and could account for mortality and delayed renal recovery. The identification of patients at high risk for IHI is crucial but remains poorly explored. Our objective was to test whether tissue perfusion parameters assessed at the bedside (mottling, index capillary refill time (iCRT), and lactate) predict IHI and to develop and to validate a predictive score of IHI.

Patients and methods: Prospective observational study in a 18-bed medical ICU in a tertiary university hospital including hemodialysis sessions performed for acute kidney injury. Exclusion criteria were patients with dark skin and dialysis performed in extreme emergency. Mean arterial pressure (MAP), mottling score, iCRT, and lactate were recorded just before starting hemodialysis. First episode of IHI requiring therapeutic intervention was recorded 60, 120, and 240 min after hemodialysis starting.

Results: Ninety-six hemodialysis sessions performed in 43 patients were recorded. Half of the patients received vasopressors (n = 43, 45%). IHI occurred in 22 (23%) sessions and was more frequent among patients receiving vasopressors (42 vs 7%, p < 0.0001). Mottling were more frequent (59 vs 26%, p = 0.005), lactate levels higher (2.8 [1.4–6.9] vs 1.1 [0.8–1.5] mmol L, p < 0.0001) and iCRT longer (3.3 [1.6–4.6] vs 1.1 [0.8–2.0] s, p < 0.0001) before sessions with IHI compared to sessions without, independently of MAP (p < 0.0001). The incidence of IHI increased with the number of tissue perfusion alterations (3, 11, 41, and 55% for 0, 1, 2, and 3 alterations, respectively, p < 0.0001). A tissue hypoperfusion score, defined as iCRT (seconds) + lactate level (mmol L) + 1 if mottling presence was predictive of IHI independently of MAP (OR 1.18 [1.05–1.32], p < 0.0001) with an AUROC of 86 [77–94] %. A threshold of 4.1 predicted IHI with a sensitivity of 82 [60–95]  %, and a specificity of 82 [72–90] %. The accuracy of this score was validated in a second prospective cohort including 115 hemodialysis sessions performed in 45 patients.

Conclusion: The incidence of IHI increases with the number of tissue perfusion alterations independently of MAP. At the bedside, a combined tissue hypoperfusion score including mottling score, iCRT and lactate level is helpful to identify patients at high risk of IHI.

Correspondence: Ait Hamou Zakaria - aithamou-zakaria@hotmail.fr

Annals of Intensive Care 2018, 8(Suppl 1):CO-73

Introduction: Epidemiological data suggest an increased risk of long-term chronic kidney disease after acute kidney injury (AKI). In survivors of out-of-hospital cardiac arrest (OHCA), AKI is frequent and is associated with numerous factors of definitive renal injury. We made the hypothesis that AKI after OHCA was a strong risk factor of long-term chronic kidney disease (CKD). We aimed to evaluate renal outcome of OHCA survivors according the occurrence of AKI in ICU.

Patients and methods: We used the cohort of consecutive OHCA patients admitted between 2007 and 2012 in a tertiary medical ICU previously described (Geri et al. ICM. 2015). AKI was defined by Kidney Disease Improving Global Outcomes (KDIGO) criteria. Long-term creatinine level was the last blood creatinine assessment we were able to retrieve. The main outcome was the occurrence of CKD, defined by an estimated glomerular filtration rate (eGFR) lower than 60 mL min 1.73 m² according to the MDRD equation. Long-term mortality was evaluated as well. Factors associated with CKD occurrence were evaluated by competing risk survival analysis (Fine Gray and Cox cause specific models providing sub-hazard ratio (sHR) and Cox sub-hazard (CSH)).

Results: Among the 246 OHCA patients who were discharged alive, we were able to retrieve the outcome of 133 patients (median age 55 [iqr 46, 68], 75.2% of male) who were included in the analysis. During a median follow-up time of 1.8 [0.8–2.5] years, CKD occurred in 17 (12.7%) patients and 24 (18%) patients died. A previous history of arterial hypertension (sHR = 3.28[1.15 + 9.39], p = 0.027 + CSH = 4.83 [1.57 + 14.9], p = 0.006), AKI during ICU stay (sHR = 3.72[1.40 + 9.84], p = 0.008 + CSH = 5.41[1.79 + 16.3], p = 0.003) and an age higher than 55 (sHR = 6.13[1.55 + 24.3], p = 0.009 + CSH = 2.16[1.72 + 43.8], p = 0.006) were independently associated with CKD occurrence. AKI was not associated with long-term mortality (sHR = 0.73 [0.27 + 1.99], p = 0.55 + CSH = 0.75 [0.28 + 2.01], p = 0.57).

Conclusion: In OHCA survivors resuscitated from an OHCA, CKD was a frequent long-term complication. AKI during ICU stay was a strong determinant of long-term CKD occurrence.

Correspondence: Garnier Fanny - garnier.fanny26@gmail.com

Annals of Intensive Care 2018, 8(Suppl 1):CO-74

Introduction: Acute kidney injury requiring dialysis (AKID) is associated with high mortality. In several countries, epidemiology of AKID is lacking. Objectives of the study was to assess changes in standardized incidence, patients’ characteristics and hospital mortality in France.

Patients and methods: The study was conducted using the French database on hospitalizations (Program of Medicalization of Information Systems). It focused on adults hospitalized in metropolitan France between 2009 and 2014 and diagnosed with AKID according to the codes of Common Classification of Medical Procedures. We described and statistically compared crude and standardized incidence of AKID on gender and age, as well as comorbidity and mortality rates, principal diagnosis and dialysis modality. We also analyzed variables associated with mortality using logistic regression.

Results: Crude incidence of AKID increased from 459per million people (pmp) in 2009 (16,500 cases) to 512pmp (25,142 cases) in 2014. Standardized incidence increased from 475[468–482] in 2009 to 524 pmp[517–532] in 2013, an annual increase of 2%. AKID’s incidence was twofold higher for men. Patients’ median age was 68 years. Standardized incidence of AKID increased by 18, 8, 15 and 50% for the 60–69, 70–79, 80–89 and over 90 age groups respectively. The most common comorbidities were cardiocerebrovascular (67.3%), pulmonary (42%), chronic kidney disease (CKD) (32.5%), diabetes (27.9%), all of which increased significantly (excluding CKD). The five most frequent principal diagnoses were heart failure (17.7%), sepsis (13.4%), acute kidney injury (12.7%), digestive diseases (11.6%) and shock (8.7%), with significant increase in heart failure and digestive diseases. Number of stays in intensive care units increased from 80.3 to 83.9%. Continuous dialysis was the most widely used dialysis modality, with a utilization rate up from 56.9 to 61.8%. Intra-hospital mortality was stable at 47%. Factors associated with higher mortality were: advanced age, intensive care admission (OR 3.33), introduction to dialysis exceeding 8 days (OR 1.46), cardiocerebrovascular comorbidities (OR 1.12), pulmonary comorbidities (OR 1.39), cancer (OR 1.33), hepatic comorbidities (OR 1.85). Diagnoses associated with higher mortality, compared with CKD, were: shock (OR 5.99), respiratory diseases (OR 4.81), sepsis (OR 4.44), digestive diseases (OR 4.10) and heart failure (OR 3.33).

Conclusion: It is the first study in France showing a significant increase in AKID’s standardized incidence among adults and a significant change in patients’ characteristics.

Correspondence: Faguer Stanislas - stanislas.faguer@gmail.com

Annals of Intensive Care 2018, 8(Suppl 1):CO-75

Introduction: Many critically ill patients have a moderate to high risk of bleeding but they also require prolonged intermittent dialysis to ensure a negative water balance without hemodynamic adverse events. Thus, a heparin-free easy-to-use anticoagulation within the dialysis circuit is needed but, to date, usual protocols (iterative saline flushes, heparin grafted membranes) lead to 20–50% of premature clotting and sessions that last greater than 240 min are rarely achievable. We assessed the safety and efficiency of heparin-free regional citrate anticoagulation of the dialysis circuit using a calcium-free citrate-containing dialysate, with calcium reinjected according to ionic dialysance (an online measure of the instantaneous clearance of small molecules available in most of dialyzers).

Patients and methods: We prospectively reported the clotting events that occurred during all the heparin-free dialysis sessions that were performed with a regional anticoagulation based on calcium-zero citrate-containing dialysate (Citrasate, Hemotech, France) between January 2016 and August 2017 in a 28-beds ICU.

Results: A total of 186 dialysis sessions were performed in 70 patients (mechanical ventilation n = 93 + norepinephrine n = 20). Median duration of dialysis was 240 min (IQR, 240–300 + maximum 360 min), and median ultrafiltration volume was 2 L (IQR 1.3–2.6). When assessed, urea and Beta2-microglobulin reduction rates were 64.5% ± 0.4% and 48% ± 0.13%, respectively. Postfilter ionized calcium was 0.35 ± 0.17 and 0.38 ± 0.14 mmol L at 1 and 3 h, respectively, within the extracorporeal circuit. A major clotting event that led to premature termination of the session occurred in only 5186 sessions (2.7%). In these five cases, major catheter dysfunction occurred before clotting within the circuit. Prefilter ionized calcium remained within narrow ranges (before after change +0.07 ± 0.006 mmol L), and total-to-ionized calcium ratio, a surrogate marker for citratemia, was unchanged and always below 2.5. In 85 sessions, no ionized calcium measurement was required.

Conclusion: Dialysis anticoagulation with calcium-free citrate containing dialysate is an easy-to-use, efficient, and inexpensive form of heparin-free regional anticoagulation. Calcium reinjection according to ionic dialysance allows prolonged hemodialysis sessions in critically ill patients without the need to systemically monitor ionized calcium. Sessions can be safely extended according to the hemodynamic tolerance to ensure an adequate dose of dialysis and a negative water balance, a major point in patients with severe AKI.

Correspondence: Gaudry Stéphane - stephanegaudry@gmail.com

Annals of Intensive Care 2018, 8(Suppl 1):CO-76

Introduction: Acute kidney injury (AKI) is frequent and associated with poor intensive care unit (ICU) outcome in critically ill patients. Superimposed AKI on chronic kidney disease (CKD) is associated with long-term risk of death and end-stage renal disease. We aimed to investigate the effect of an early compared to a delayed RRT initiation strategy in patients with severe AKI on CKD.

Patients and methods: Post-hoc analysis of the AKIKI trial. In this multicenter pragmatic RCT, patients with severe AKI (stage 3 of KDIGO classification) were randomly assigned to either an early or a delayed RRT initiation strategy. The subgroup of patients with CKD was a priori defined as a preexisting creatinine clearance < 60 ml min in stable condition. The primary endpoint was 60-day mortality. Secondary endpoints included—number of patients who actually received RRT, RRT-free days, mechanical ventilation-free days and vasopressor-free days, ICU and hospital length of stay, time to renal function recovery, day-28 and day-60 dependence on RRT.

Results: Of the 619 patients, 60 (10%) had CKD. We found significant treatment effect heterogeneity according to the CKD status (test for interaction p = 0.006). Patients with CKD had an increased risk of death (HR 2.50 (95% CI 1.23–5.00, p = 0.009) with the early strategy (Fig. 1). Delayed strategy allowed avoidance of RRT in 45% of patients with CKD. Mechanical ventilation-free days, ICU and hospital length of stay, dependence on RRT at day-28 and day-60 did not significantly differ between RRT strategies. The number of vasopressor-free days was significantly higher in patients with CKD assigned to the delayed strategy (5.5 [0–20] vs 22.0 [3–26], p = 0.02). We found a trend for an earlier renal function recovery with the delayed RRT strategy but this difference did not reach significance (p = 0.06).

Conclusion: Early renal replacement therapy initiation strategy was associated with a significant increase in 60-day mortality in the subgroup of patients with CKD. Delaying RRT initiation may benefit patients with severe AKI on CKD.

Correspondence: Benghanem Sarah - benghanem.sarah89@gmail.com

Annals of Intensive Care 2018, 8(Suppl 1):CO-82

Introduction: Brain injury is the first cause of death after cardiac arrest (CA) and multimodal neuroprognostication is a cornerstone of post-resuscitation care. Among the different usable information provide by electroencephalogram (EEG), the aim of this study was to evaluate the predictive value of EEG reactivity regarding neurological outcome at discharge.

Patients and methods: Using our prospective registry of successfully resuscitated patients admitted to a cardiac arrest center between January 2007 and 2016, we studied all consecutive comatose patients still alive at 48 h and in whom at least one EEG was performed during coma. In addition to usual clinical findings, we collected EEG (patterns and reactivity, status epilepticus) and somatosensory evoked potentials characteristics. The EEG reactivity was evaluated by a blinded neurophysiologist and was defined as a reproducible change of the tracing (in amplitude or frequency) provoked by an auditory and a nociceptive standardized stimulation. We evaluated the predictive values of persistent lack EEG reactivity and other indicators regarding their respective ability to predict a favorable or unfavorable outcome. Recovery of a level 1 or 2 on the Cerebral Performance Category (CPC) scale at discharge was considered as a favorable outcome, as opposed to recovery of a CPC level 3–5 (unfavorable outcome).

Conclusion: In this population of post-cardiac arrest patients, the presence of EEG reactivity was poorly predictive of a favorable neurologic outcome. The absence of reactivity was highly predictive of unfavorable outcome. In combination with other indicators, searching for EEG reactivity may have important implications in the neuroprognostication process.

Correspondence: Heming Nicholas - nicholas.heming@aphp.fr

Annals of Intensive Care 2018, 8(Suppl 1):CO-83

Introduction: The safest resuscitation fluid for surgical patients remains uncertain. We sought to compare the safety and efficacy of either crystalloids or colloids in surgical critically ill patients.

Patients and methods: Subgroup analysis of the multicenter randomized trial. All critically ill surgical patients were included in the current analysis. 741 patients were randomized to receive crystalloids or colloids. Patients were stratified as suffering from hypovolemia due to sepsis, trauma or other causes. Primary outcome measure was death by day 28. Secondary outcome measure included death by day 90.

Results: Crystalloids (n = 356) or colloids (n = 385) were administered for all fluid interventions other than fluid maintenance throughout the ICU stay. 484 (65.3%) patients were male, median [IQR] age was 66 [52 + 76] years. Surgery was unscheduled in 543 (73.3%) cases. Mortality by day 28 did not significantly differ for crystalloids 84 (23.6%) vs. colloids 100 (26%) (Adjusted odds ratio 0.86 [IC 95%, 0.61 to 1.21] + p = .49). Mortality by day 28 did not significantly differ among all three predefined strata (p = .52). Death by day 90 (111 (31.2%) vs. 122 (31.7%) (Adjusted odds ratio 0.97 [IC 95%, 0.70 to 1.33] + p = .86) did not significantly differ between groups.

Conclusion: This subgroup analyses of a randomized controlled trial, found no survival benefit when comparing crystalloids to colloids in critically ill surgical patients.

Correspondence: Roger Claire - claire.roger@chu-nimes.fr

Annals of Intensive Care 2018, 8(Suppl 1):CO-84

Introduction: Goal of a fluid challenge (FC) is in fine to increase the stroke volume (SV) or the cardiac index (CI) when an episode of hypovolemia or a preload dependence status are suspected. FC is one of the most common practices in ICUs, however, the way to assess the response to FC is not standardized. The present study aimed to evaluate whether the trans-thoracic echocardiographic (TTE) assessment of the response to FC immediately at the end of the infusion or delayed 20 min later could affect the results of the FC.

Patients and methods: Prospective, observational, multicentre study including all ICU patients in septic shock requiring a FC. Were excluded patients with—arrhythmias, poor echogenicity and severe mitral or aortic regurgitation. FC was performed administering 500 mL of crystalloids over 10 min. Fluid responsiveness was defined as a > 15% increase in stroke volume (SV). The following echocardiographic parameters were recorded—E wave, A wave, E A ratio, velocity–time integral (VTI), Ea wave and Sa wave. MAP, HR and TTE variables were collected at baseline (T0), at the end of fluid challenge (T10) and 10 (T20) and 20 min (T30) after the end of fluid challenge. Quantitative data are expressed as mean and standard deviation (SD) or median and interquartile (IQR), according to their distribution. Qualitative data are expressed as absolute number and frequency (%).

Results: From May 20th 2014 to January 7th 2016, a total of 143 patients were enrolled in 11 French ICUs (mean age—64 ± 14 years, median IGS II—53 [43–63], median SOFA score—10 [8–12]). Among the 76 143 (53%) patients responders to FC at T10, 37 patients were transient responders (TR), i.e. became non-responders at T30 (49%, 95% CI = [37–60]) and 39 (51%, 95% CI = [38–62])) patients were persistent responders (PR), i.e. remained responders at T30. Among the 67 non-responders (NR) at T10, 4 became responders at T30, (6%, 95% CI = [1.9–15.3]). In the subgroup analysis, no statistical difference in haemodynamic and echocardiographic parameters was found between non-responders, transient responders and persistent responders (Fig. 1).

Conclusion: The present study shows that, after a 15% VTI increase at the end of the FC, VTI returns to baseline at 30 min in half of the responders. Blood volume status (normo or hypovolemia) before initiating the fluid infusion could explain the transient or persistent response to FC observed in septic patients.

Correspondence: Dumas Guillaume - dumas.guillaume1@hotmail.fr

Annals of Intensive Care 2018, 8(Suppl 1):CO-85

Introduction: Sepsis is a frequent critical condition. Mottling score, an hypoperfusion parameter, is well correlated with outcome. However, uncertainties persist regarding its value not only as a marker of patient severity but also as an independent predictor of mortality and treatment efficacy.

Patients and methods: We performed a post hoc analysis of four published prospective studies including sepsis patients with or without shock. We analyzed the relationship between the mottling score (from 0 to 5) and Day-14 mortality according to other prognosis covariates such as catecholamine dosing, urine output and plasma lactate levels. First, factors associated with outcome were determined by multivariate analysis. Second, mottling score-by-covariate interaction was studied to better understand its effect on mortality. Finally, effect of mottling score variation at different time point (H0–H6–H12–H24) was assessed.

Results: 274 patients were included. SAPSII at admission was 54 [41.2–70.0]. Six hours after ICU admission, 210 (76.6%) patients received catecholamine (norepinephrine (73.2%)), median dose of 0.3 [0.05–0.8] µg Kg min. At H6, the mottling score distribution was—grade 0 in 126 patients (48.6%), grade 1 in 36 (13.9%), grade 2 in 29 (11.2%), grade 3 in 29 (11.2%), grade 4 in 14 (5.4%) and grade 5 in 25 (9.7%). Day-14 mortality rate was 37.2%. Factors associated with Day-14 mortality were—mottling score (OR 2.16 [1.63–2.86]), lactate level > 5 mmol l (OR 3.74 [1.35–10.35]), urine output < 0.5 ml Kg h (OR 2.5 [1.09–5.7]) and SAPSII (OR 1.04 [1.02–1.06]). The predictive value of mottling score was not affected by catecholamine dosing (p = 0.45). OR of mottling score ranged from 2.38 [1.37–4.12] for patients without catecholamine to 3.84 [1.98–7.43] for patients receiving > 0.8 μg/kg/kg min. We did not find any significant interaction between mottling score prognosis value and urine output (p = 0.62) or lactate concentration (p = 0.09). A decrease of mottling score between the first 6 h of resuscitation was associated with outcome improvement (adjusted on SAPSII, p = 0.001).

Conclusion: Our results support the high prognostic value of mottling score for 14-day mortality, independently of catecholamine dosing and other hypoperfusion parameters. Mottling score variation during resuscitation is a strong predictor of mortality.

Correspondence: Chommeloux Juliette - juliette.chommeloux@gmail.com

Annals of Intensive Care 2018, 8(Suppl 1):CO-86

Introduction: Numerous studies have identified a relationship between microcirculation disorders and poor outcomes in cardiogenic shock patients. Despite the increasing interest for Veno-Arterial Extra Corporeal Membrane Oxygenation (VA-ECMO), data on microcirculation in this context are scarce. Thus, we aimed to describe the evolution of microcirculation and its relationship with the outcomes of patients on VA-ECMO for refractory cardiogenic shock.

Patients and methods: Consecutive patients with refractory cardiogenic shock (cardiac arrest excluded) who required VA-ECMO were included. Microcirculation evaluation by video microscopy, global haemodynamic and Doppler echocardiography variables were obtained before and 2, 4, 12, 24 and 48 h after ECMO initiation. Patients who survived had additional evaluations six hours before and after VA-ECMO removal.

Results: Fourteen patients (median age 58 (interquartile range 56–62) years + SOFA score 14 (12–18) were included. Myocardial infarction (33%) was the main cause for shock. Mortality on VA-ECMO was 57%, whereas ECMO was successfully weaned in 6 (33%) patients. Proportion of Perfused Vessel (PPV), Perfused Vessel Density (PVD), Micro Flow Index (MFI) and Heterogeneity Index (HI) were severely impaired before ECMO. Re-establishing high and stable peripheral blood flow with VA-ECMO led to a rapid decrease in heart rate and vasoactive inotropic support and significantly improved all microcirculation parameters within 12 h. Total vessel density and PVD, measured before and after ECMO initiation, were better in patients successfully weaned from ECMO (p < 0.05) (Fig. 1).

Conclusion: Cardiovascular support with ECMO-VA rapidly improved macro and microcirculation in refractory cardiogenic shock patients. Total vessel density and perfused vessel density were significantly better in survivors 12 h after ECMO initiation and might therefore help to predict outcomes. Further studies are now needed to better define the utility of this technology in larger groups of VA-ECMO patients.

Correspondence: Bourcier Simon - simon_bourcier@hotmail.com

Annals of Intensive Care 2018, 8(Suppl 1):CO-97

Introduction: Thyroid storm is a rare but life-threatening disease related to thyrotoxicosis. It can lead to multiple organ failure including cardiovascular disorders or neurological impairment. To date, data on this disease in ICU patients are scarce and limited to case reports. We therefore aimed to describe clinical presentation, outcomes and management of thyroid storm in ICU patients.

Patients and methods: Local diagnoses coding database (from January 2000 to July 2017) from 22 French ICU were interrogated for main and secondary diagnoses codes including thyrotoxicosis based on the International Classification of Disease 10th revision. Thereafter two investigators reviewed all the medical records selected. Inclusion criteria were thyroid storm based on the diagnostic criteria of the Japan Thyroid Association (T. Satoh, Endocrine Journal 2016). It combines thyrotoxicosis with elevated levels of free triiodothyronine (FT3) or free thyroxine (FT4) with at least two of the following symptoms—central nervous system manifestation, fever, tachycardia > 130 bpm, congestive heart failure, or total bilirubin level more than 50 micromol/l. Clinical presentation, therapy used, and outcome were recorded.

Results: Sixty-two patients (median age 57 years (interquartile range 44–67) + SAPS II 45 (32–66) were included. Thyroid storm was the first manifestation of thyrotoxicosis in 18 (29%) patients. Graves’ disease (27%), amiodarone induced thyroiditis (26%), autoimmune thyroiditis (11%), and toxic multinodular goitre (10%) were the main causes of hyperthyroidism. Amiodarone, thyroid hormone toxicity, antithyroid drugs withdrawal or infectious trigger were identified in 59 (95%) patients. Organ support including mechanical ventilation, catecholamine infusion, renal replacement therapy and veno-arterial ECMO were used in 34, 31, 11, and 12 patients, respectively. Main thyroid storm treatments included antithyroid drugs (81%), betablockers (69%), corticosteroids (45%), and plasmapheresis (13%). Lastly, ICU-mortality was 16%, with multiple organ failure responsible of death in all patients.

Conclusion: Although its incidence appears low, ICU physicians should be aware of the multiple clinical features of thyroid storm. Our preliminary data reported various specific therapeutic management of this potentially fatal disease.

Correspondence: Joseph Adrien - adrien.joseph@hotmail.fr

Annals of Intensive Care 2018, 8(Suppl 1):CO-98

Introduction: Thrombotic microangiopathies (TMAs) are severe diseases which often require admission in an intensive care unit (ICU). Prompt initiation of targeted therapies is required for atypical hemolytic uremic syndrome (aHUS) and thrombotic thrombocytopenic purpura (TTP), but no specific therapy is consensual for Shiga toxin-associated hemolytic uremic syndrome (STEC-HUS). Thus, rapid differentiation of STEC-HUS is mandatory to tailor the initial treatment. Furthermore, apart from large outbreaks, characteristic features of this syndrome in adults have not been described.

Patients and methods: In this study, we retrospectively compared the characteristics of STEC-HUS, aHUS and TTP patients at admission in two expert ICUs. Patient were included if they presented with the triad of mechanical hemolytic anemia, thrombocytopenia and organ damage, and TMAs were classified using international criteria. Other causes than STEC-HUS, aHUS and TTP were excluded.

Results: Amongst 236 TMAs admitted between September 2003 and January 2017, 12 STEC-HUS, 21 aHUS and 91 TTP were included. STEC-HUS patients were older (64) than aHUS (36, p = 0.02) and TTP patients (43, p < 0.01). They presented with more frequent digestive symptoms (92 versus 62 and 37% for aHUS and TTP, p = 0.01 and < 0.01), but bloody diarrhea was rare (17%) and non-statistically different from other TMAs. Confusion was more frequent in STEC-HUS (33%) than aHUS patients (5%, p = 0.05). Biologically, STEC-HUS patients displayed elevated fibrinogen levels (5.1 vs 3.2 and 3.6 for a HUS and TTP, both p < 0.01) and severe renal failure. Forty-two percent required renal replacement therapy and 83% were treated with plasma exchange before the distinction from other TMAs could be made. Only 1 (8%) STEC-HUS patient died in the ICU (Fig. 1).

Conclusion: Characteristics supposed to identify STEC-HUS are largely shared with other TMAs. In particular, the differential diagnosis between aHUS and STEC-HUS appears to be more difficult than the stereotypical description derived from pediatric studies.

Correspondence: Ben Hamou Adrien - adrien.benhamou@hotmail.fr

Annals of Intensive Care 2018, 8(Suppl 1):CO-99

Introduction: Stress hyperglycemia frequently occurs in critically ill patients independently of preexisting diabetes. Some studies suggest an association with poor prognosis. The aim was to evaluate 28 and 90-day mortality after admission to the intensive care unit and the incidence of diabetes during follow-up at 6 and 12 months in patients who presented a severe hyperglycemia in their first 3 days in the ICU (glycemia > 200 mg dL, > 2 samples).

Patients and methods: We conducted a prospective multicentric study between April 2012 and August 2016 in French ICUs. Patient’ characteristics, medical history, blood tests, aetiology of ICU admission and organ dysfunctions were collected. Patients were divided into 4 groups—diabetic patients with severe hyperglycemia (HD), non-diabetic with severe hyperglycemia (HND), diabetic without severe hyperglycemia (NHD) and non-diabetic without severe hyperglycemia (NHND). Kaplan–Meier estimator was used to analyze survival adjusted to disease severity (Logistic Organ Dysfunction, Simplified Acute Physiology Score, Knaus, McCabe scores) and Kaplan–Meier survival curves were compared by Cox proportional hazards test. Fasting glycaemia above 126 mg dL was used to diagnose diabetes at 6 or 12 months.

Adrien BEN HAMOU The Kaplan–Meier survival curve Abstract SRLF 2018.

Results: 991 out of the 1000 enrolled patients were analyzed (62% men). Median age was 61 years old, median Body Mass Index was 27 kg m² + 236 patients (24%) were diabetic. Among the 991 patients, 413 (42%) presented severe hyperglycemia and had a significant increase in mortality at day 28 (HR = 1.87, IC 95% [1.36–2.56], p = 0.0001), and day 90 (HR = 1.72, IC 95% [1.33–2.22], p < 0.0001), independently of the cause of death. Severe hyperglycemia was associated with increased mortality risk at days 28, when patients were admitted for shock, except septic shock, (HR = 1.70, IC 95% [1.01–2.87], p = 0.04) but not when admitted for coma, sepsis or cardiac arrest. Mortality rate was significantly higher in patients with severe hyperglycemia compared to those without, regardless of preexisting diabetes (HND HD vs. NHND NHD groups + p < 0.05). Patients with severe hyperglycemia had a higher incidence of type 2 diabetes at 6 (16 vs. 8% + p = 0.01) and 12 months (14 vs. 9% + p = 0.08) compared to those who did not.

Conclusion: Severe hyperglycemia occurring in the first 3 days of ICU admission was associated with higher mortality rate and an increased risk of diabetes in the following months regardless of preexisting diabetes.

Correspondence: Khaldi Amina - amina.khaldi@erasme.ulb.ac.be

Annals of Intensive Care 2018, 8(Suppl 1):CO-100

Introduction: Vitamin D deficiency is frequent in Northwestern countries and could represent a modifiable risk factor for critically ill patients, in relation with its pleiotropic effects (1). Some studies reported an association between 25OH vitamin D (25OH) deficiency, chronic health status and ICU- and hospital-related outcomes. However, a large supplementation study have not been found to improve outcome of patients with moderate 25OH deficiency (< 20 ng ml) (2). The aim of the study is to analyze the relationship between the severity of 25OH deficiency at ICU admission, severity of illness and outcomes and ultimately to identify subgroups of patients in whom the likelihood of benefit of supplementation is larger.

Patients and methods: Consecutive patients admitted over a 4-month period who stayed at least 48 h in a medical surgical 32-bed ICU were included. In these patients, demographic data, Charlson co-morbidity score, severity scores (SAPS 3 and SOFA) and 25-OH (chemiluminescence, DiaSorin) were collected at admission. ICU and hospital length of stay (LOS) and mortality were recorded. Correlations were searched between 25OH and the different scores, and vital outcomes by non-parametric tests for continuous value and by cross dynamic board for categorical variables.

Results: A total of 185 patients have been analyzed. The mean (SD) age was 66 (15) years, and 50% were men, most often admitted for a medical reason (57.8%), Charlson score 2.5 ± 2.9, SAPS III 56.2 ± 20.5, SOFA 5.5 ± 4.5, ICU (LOS) 8.2 ± 9.2 days, Hosp LOS 28.4 ± 27.5 days, ICU mortality 17.8%, hospital mortality 23.8%. The mean 25OH at admission was 16.9 ± 11.5 ng ml. 25OH deficiency (> 20 ng ml) in 69.7%, severe deficiency (> 12 ng ml) in 47.5%, and a moderate deficiency (12–20 ng ml) in 22.2%. However, there was no correlation between vitamin D level at ICU admission, severity of illness, mortality and LOS (Fig. 1).

Conclusion: These findings confirm the high incidence of severe 25OH deficiency and a possible relation with a higher mortality in the subset of long-stayers. It reflects a marker of poorer health. Specific subsets of patients could benefit from vitamin D supplementation.

Correspondence: Bourcier Simon - simon_bourcier@hotmail.com

Annals of Intensive Care 2018, 8(Suppl 1):CO-101

Introduction: Severe hypothyroidism manifestations including myxedema coma and tamponnade frequently lead to intensive care unit (ICU) admission. To date, data on this severe disease in ICU are scarce and limited to case reports. We aimed to describe clinical presentation, management and outcomes of severe hypothyroidism admitted in ICU.

Patients and methods: Local diagnoses coding database (from January 2000 to July 2017) from 15 French ICU were interrogated for main and secondary related to hypothyroidism and myxedema coma in the International Classification of Disease 10th revision. Two investigators therefore reviewed all medical records selected. Inclusion criteria were severe patients whom severe clinical hypothyroidism manifestations were the main reason for ICU admission. Clinical presentation, therapy used, and outcome were recorded.

Results: We report the preliminary data of 40 patients (80% female + median age 70 years (65–76) + SAPSII 58 (41–74)). Hypothyroidism was unknown before ICU admission in 68% patients. Median SOFA score at ICU admission was 6 (4–10). Myxedema coma, circulatory failure, respiratory failure, and severe hypothermia were respectively the main admission reason in 60, 18, 10, and 5% patients. A precipitating factor such as drugs thyroid toxicity, thyroid hormone withdrawal or infection was found out in only 13 (32%) patients. Main causes of hypothyroidism were thyroiditis and thyroidectomy. Thirty-two (80%) patients had alteration of consciousness with a median Glasgow score at 10 (6–13). In addition, median heart rate at ICU admission was 50 (35–65) bpm while hypothermia < 35 °C was noted in 24 (60%) patients. Median TSH level at admission was 78 (43–135) mUI l, T3 and T4 levels respectively 0 (0–0.45) pmol l and 2 (0–5.07) pmol l. Rhabdomyolysis was frequent with median CPK level 539 (192–1500) UI l. Organ support including mechanical ventilation, catecholamine infusion and, renal replacement therapy were respectively used in 58, 45, and 20% patients. Lastly, 70% patients received oral levothyroxine whereas the intravenous form was used in others. Overall ICU-mortality was 22%.

Conclusion: Our preliminary data showed that severe manifestations of hypothyroidism leading to ICU admission represent de novo hypothyroidism in two-thirds of patients, leading to a high mortality.

Correspondence: Le Borgne Pierrick - pierrick_med@yahoo.fr

Annals of Intensive Care 2018, 8(Suppl 1):CO-102

Introduction: When it comes to infections of the central nervous system (CNS), the greatest challenge in the Emergency Department (ED) is to identify patients that have a rare life-threatening diagnosis. Alone or in combination, fever, headache, altered mental status encompass a broad differential diagnosis. Antibiotics or antiviral therapy should be given as soon as possible, ideally after both blood and cerebrospinal fluid (CSF) have been obtained. Early treatment is associated with a lower mortality.

Patients and methods: We present here, a four-year (2012–2015) retrospective and monocentric study. During the period of the study, we included all adult patients with the diagnosis of CNS infection (positive CSF culture). We collected and analyzed all clinical, biological, imaging, treatments and evolution datas during the stay. A total of 67 patients with CNS infection have been included for statistical analysis. We analyzed a second group (n = 25) with suspected CNS infection (negative CSF) as a control group.

Results: In the study population, mean age was 43 ± 21.5 years old and the sex-ratio was 1.2. There were no difference between the two groups in terms of clinical signs except for more altered mental status in the control group (p = 0.02). All patients of the study (n = 92) benefited of lumbar puncture (LP) in the ED with an average time of 381 ± 370 min after admission. This delay was the same between the two groups (p = 0.76) but was significantly higher in the encephalitis subgroup (n = 13, p = 0.03). Patients who had imaging (CT or MRI) during the ED stay had more likely a delay in LP realization (450 vs 193 min, p = 0.0005). Patients where the CNS infection diagnosis was firstly evoke by the triage nurse had LP more quickly (p = 0.03). The median door to-antibiotic-time was 339 min (IQR 198–534) with no difference between the two groups of the study (p = 0.7). 94% of all patients were hospitalized for an average length of stay of 11.7 ± 15.5 days and 10% of them were admitted in the ICU. The in-hospital mortality was 14% in the study population.

Conclusion: Our study highlights the poor adherence to guidelines. The management of CNS infections remains slow and heterogeneous. Management practices, such as the timing of LP realization or antibiotics in relation to the return of CT scan or CSF results are the most important factors associated with antibiotic timeliness.

Correspondence: Elabbadi Alexandre - alexandre.elabbadi@gmail.com

Annals of Intensive Care 2018, 8(Suppl 1):CO-103

Introduction: There are numerous causes of acute exacerbations of COPD (AECOPD), the most common of which are bronchial and or pulmonary infections. Viral etiologies may account for 30% of AECOPD, but this rate is likely underestimated because of the limited performance of the conventional diagnostic tests. Multiplex molecular diagnostic tests may identify several pathogens including viruses and bacteria, from a single respiratory tract sample, with high sensitivity. Using these tests, respiratory viruses are identified in 30 to 60% of cases, according to the series. The objective of this work was to describe the microbial epidemiology, the management and the outcome of patients admitted to the intensive care unit (ICU) with moderate to severe AECOPD, in the era of multiplex testing.

Patients and methods: A prospective non interventional multicenter study conducted in two university-teaching hospitals. In addition to the usual samplings, a nasopharyngeal swab was performed for multiplex polymerase chain reaction (PCR), using respiratory panels FilmArray Biomérieux (17 viruses and 3 bacteria) or ePlex automaton (22 viruses and 4 bacteria) depending on the center.

Results: The preliminary results involve the 37 patients (29 males + 70 years (55–87)) included in Tenon hospital over a 6-month period. The mean FEV1 was 37% (21–60) median 35% [31–44]. Drug therapies included anticholinergics (n = 21 + 62%) and beta-2-mimetics (n = 24 + 71%), inhaled (n = 16 + 47%) or oral (n = 2 + 6%) steroids, and azithromycin (n = 5 + 23%). A respiratory virus was identified in 11 patients (30%), alone or in combination with a bacterium (n = 6). A bacterial pathogen was identified alone 9 times (21%). Therapeutic interventions did not differ depending on whether a virus was detected or not-exposure to antibiotics (10 ± 1.8 vs. 12.5 ± 2 d + p = 0.4), administration of oseltamivir (5/6 vs. 8/18 + p = 0.39), steroids (1/7 vs. 3/15 + p = 0.12) and mechanical ventilation (11/11 vs. 20/26 + p = .08). The ICU length of stay (5.8 ± 0.8 vs. 7.8 ± 1.1 d + p = 0.29) was similar. The ICU and 90d-mortality rates were 5.4 and 13.5%, respectively.

Conclusion: Respiratory viruses are frequently involved in moderate to severe AECOPD. The respiratory multiplex PCR should be performed in this setting and the results should be taken into account to more adequately use the anti-microbial treatments.

Correspondence: Roujansky Ariane - ariane.roujansky@gmail.com

Annals of Intensive Care 2018, 8(Suppl 1):CO-104

Introduction: Co-infections during severe malaria have been widely studied among children in the endemic setting and there is evidence they result in worse outcome. Little is known about the epidemiology and impact of these co-infections (community-acquired or nosocomial) in the course of imported malaria in adults. We sought to describe community-acquired co-infections in severe imported malaria (SIM) in adults and to identify associated factors.

Patients and methods: We conducted a multicenter observational study in 52 French intensive care units (ICU), derived from a retrospective (2000–2006) and a prospective (2007–2010) study. We identified patients admitted for severe imported falciparum malaria according to 2000 World Health Organization (WHO) criteria adapted for imported malaria. Community-acquired co-infection was defined as an infection diagnosed within the first 48 h following ICU admission. Clinical and demographic data were collected from medical charts using standardized case-report forms. Factors associated with community acquired co-infection were identified by univariate and multivariate analysis.

Conclusion: Our study shows that in a large population of adults admitted to ICU for SIM, 7.5% have a community-acquired co-infection, mainly bacteraemias and pulmonary infections. These results emphasize the need to search for co-infection in these patients and make therapeutic adaptations accordingly. Male sex, respiratory distress and acidosis at admission may help to identify patients most at risk.

Correspondence: Miailhe Arnaud-Félix - flexmiailhe@hotmail.com

Annals of Intensive Care 2018, 8(Suppl 1):CO-105

Introduction: Leptospirosis is a worldwide zoonosis, and is more common in tropical area. Epidemiological data reported increased incidence during the past 10 years in Metropolitan France. Most of the cases are asymptomatic or sparsely symptomatic, but some patients develop severe leptospirosis, requiring hospitalisation in intensive care units (ICU). Only few studies exist on severe leptospirosis and data on severe leptospirosis in Europe are very scarce. We performed a retrospective multicentre study in metropolitan France in order to identify the characteristics, treatments and prognostic factors of severe leptospirosis in this context.

Patients and methods: LeptoRea was a retrospective multicentre study performed in 73 French ICUs. Each unit was requested to include all the patients admitted in ICU, between 2012 and September 2016, over than 18 years old, and with leptospirosis diagnosed with at least one positive test (PCR, MAT, ELISA). For each patient, a case report form was completed and hospitals report was sent to the study coordinator (AFM). A descriptive analysis was performed to assess the patients characteristics, treatments received and mortality rate. A univariate analysis was performed to identify factors associated with mortality.

Conclusion: This study is the first large multicenter study on severe leptospirosis in Metropolitan France and the world’s largest series on leptospirosis in ICU. Our data show that severe leptospirosis is uncommon in ICUs of Metropolitan France. Despite high IGS2 and SOFA score, mortality rate was lower than expected and previously reported.

Correspondence: Jaquet Pierre - pierre.jaquet@yahoo.fr

Annals of Intensive Care 2018, 8(Suppl 1):CO-106

Introduction: Herpes simplex encephalitis (HSE) remains the most common infectious cause of encephalitis in Europe and is associated with a poor prognosis. Little is known about management and functional outcomes of severe HSE cases requiring ICU admission. This study aimed to identify factors associated with poor neurological outcome in HSE patients admitted to the ICU.

Patients and methods: We conducted a large retrospective multicenter study on consecutive patients diagnosed with HSE in 46 ICUs in France, between January 2006 and December 2016. Patients were included if they met the International Encephalitis criteria for confirmed HSE, i.e. an acute encephalitis syndrome and a positive cerebrospinal fluid PCR for Herpes simplex virus. Multivariate logistic regression analysis was used to identify factors associated with poor outcome at 90 days, defined by a score of 3–6 (indicating severe disability or death) on the modified Rankin Scale (mRS). Data are presented as number (percentage) or median [interquartile range].

Conclusion: In adults with HSE requiring ICU admission, only 28% patients had functional independence at 90 days (mRS 0–2). Immediate ICU admission, early brain MRI and elevated body temperature represent modifiable factors that may impact patients’ outcome.

Correspondence: Zarrouki Youssef - zarroukiyoussef@hotmail.fr

Annals of Intensive Care 2018, 8(Suppl 1):F-01

Introduction: Prophylactic non-invasive ventilation (NIV) is a well established method for prevention of post-extubation acute respiratory failure in hypercapnic patients. However, its role in the post-extubation period, in traumatic brain injury patients, is uncertain. Especially, because of effects of the brain injury, on respiration and airway control. We perform a study to assess the impact of prophylactic NIV after extubation among patients with severe traumatic brain injury.

Patients and methods: Over a period of 1 year, adult patients with isolated severe traumatic brain injury, who were under invasive mechanical ventilation for more than 48 h were eligible for inclusion in the study. They were randomized, after decision of extubation, to receive conventional therapy or conventional therapy associated with NIV. Conventional therapy consisted of oxygen delivery by facial mask, semi-recumbent position, mucus suctioning and nebulization therapy. The main objective of the study is to assess the impact on reintubation rate. Extubation succes was defined by the absence of need for reintubation within the 7 days. The secondary objective is to evaluate the effect on ICU length of stay after extubation.

Results: Fifty-three patients ranging in age from 17 to 61 years, were included in the study. They have no history of respiratory disease. 26 patients to receive NIV and 27 to receive conventional therapy. NIV was delivered with mean levels of inspiratory positive-airway pressure of 11.1 ± 1.8 cm H₂O and expiratory positive-airway pressure of 5 ± 1.00 cm H₂O for a mean period of 19.85 ± 10 h for 6 to 11 h day. Reintubation rates among NIV group (34.61%) and conventional group (44.44%) were similar (p = 0.39). The duration of intensive care unit stay after extubation (8.5 ± 2.55 vs 9.35 ± 1.72 days, p = 0.83) in both groups were comparable.

Conclusion: NIV is safe to use in severe brain injured patients soon after extubation, although clear benefit is not documented in this study. In order to establish a potential benefit of NIV in this type of patients, it is probably necessary to study it in the subgroup of hypercapnic patients in the post-extubation period.

Correspondence: Cayrol Elsa - elsa.cayrol@free.fr

Annals of Intensive Care 2018, 8(Suppl 1):F-02

Introduction: The clinical benefit of non-invasive ventilation (NIV) in patients with acute hypoxemic respiratory failure (ARF) is being called into question. Indeed, in a multicenter randomized trial recently conducted in hypoxemic ARF patients (Pa02 Fi02 < 300), intubation rate in the NIV group was 50% and intensive care unit (ICU) mortality rate was 25%, numbers higher than in the standard-oxygen group (1). An excessive tidal volume under NIV is a hypothesis to explain these bad outcomes (2). Our experience does not seem to support these data. Therefore we wanted to—investigate the rate of NIV success in hypoxemic ARF and global in-ICU mortality. Estimate the average expired tidal volume and identify predictive factors of NIV failure.

Patients and methods: Observational cohort study conducted in a medical ICU, including all patients treated with NIV for hypoxemic ARF over a 3 years period.

Results: During the study period, 57 patients were included (mean age 57.9 ± 15.1 years, median SAPS II 43 (29–54), median Pa02 Fi02 151 (111–186)) and 23 had acute respiratory distress syndrome (ARDS). Intubation rate was 28.1% and in-ICU mortality was 14%. Patients with a Pa02 Fi02 ratio < 150 during the first NIV session had a mortality rate five times higher than those with a Pa02 Fi02 ≥ 150 (37.5 vs 7.3% p = 0.047). More than 2000 h of NIV were analysed. Median expired tidal volume was 10 ml/kg of predicted body weight (PBT) (8–11 ml/kg PBT), whether the patients were intubated (NIF failure) or not (NIV success). In multivariate analysis, only the presence of ARDS was independently associated with NIV failure.

Conclusion: With a success rate of 71.9% and a mortality rate of 14% it seems legitimate not to exclude VNI of the management of patients with hypoxemic ARF. Moreover, the hypothesis of a deleterious effect related to an excessive tidal volume has not been confirmed in our study.

Correspondence: Kudela Agathe - kudela.agathe@gmail.com

Annals of Intensive Care 2018, 8(Suppl 1):F-03

Introduction: Unplanned extubation (UE) occurs in up to 33% of ICU patients. Reintubation is required in 28.5 to 74.7% because of respiratory failure, or other hemodynamic, neurologic or laryngeal matter. Very few data is available regarding the effect of non-invasive ventilation (NIV) performed after UE.

Patients and methods: Retrospective analysis of prospectively collected data on a 4 years period. A NIV local procedure, based on current guidelines, is available in the investigating centre, but UE is not mentioned. The decision to perform NIV in this setting was only based on the attending physician’s judgment. All patients experiencing an UE were included. We registered demographic data, patient’s characteristics and evolution after UE according to the occurrence of NIV. Prophylactic NIV denoted NIV performed after extubation without sign of respiratory failure, whereas curative NIV was a mean to treat post extubation respiratory failure.

Results: 106 UE occurred during the study period. Age was 59.4 [41.1–74.4] years, SAPS II 45 [36–54]. Reintubation was needed in 30 cases, within 1.8 [0.33–21.5] hours after UE, mostly for acute respiratory failure (n = 25). The death was associated with the need for reintubation (10 30 vs 0 76, p < 0.0001). NIV was performed in 21 cases. In all but 8 patients, NIV was used to treat an acute respiratory failure, the remaining indications being described as preventive of reintubation without evidence of respiratory failure. Respiratory failure as an indication for NIV was more often associated with reintubation (10/13 vs. 1/8, p = 0.008). NIV was also associated with death (5/21 vs. 5/85, p = 0.02). As a comparison, during the same period, 536 patients ventilated at least 48 h had a planned extubation (age 67 [54–80] years, SAPS II 49 [35–62]). Regarding NIV in this subgroup, data were available in 438 patients. NIV was used 141 times, with prophylactic NIV being more frequent (n = 118) as compared to rescue NIV (n = 23). Reintubation rates were similar as compared to UE patients for prophylactic NIV (16 118 vs. 1 8, p = 0.94) and rescue NIV (9 23 vs. 10 13, p = 0.26).

Conclusion: Though limited by its design, our study seems to show a similar efficacy of NIV following UE as compared to planned extubation, with a safety concern for rescue NIV and a potential interest for “prophylactic” NIV. Further data is warranted.

Correspondence: Lahmar Manel - firassmal4@gmail.com

Annals of Intensive Care 2018, 8(Suppl 1):F-04

Introduction: Noninvasive ventilation (NIV) is the gold standard in the treatment of COPD exacerbation. NIV in acute respiratory failure (ARF) due to bronchiectasis exacerbation remain less recommended. Objectives—to evaluate the utility and factors predicting failure of NIV during bronchiectasis exacerbation.

Patients and methods: we conducted a retrospective study between 2001 and 2016 in the Medical intensive care unit in the hospital of Taher Sfar in Mahdia. All patients with bronchiectasis exacerbation and requiring NIV were enrolled. Only the first episode of exacerbation was recorded and analyzed. At admission we collected the following parameters—demographic details, etiology of exacerbation, comorbidities, the SAPSII score, arterial blood gases, respiratory, hemodynamic and neurologic parameters, NIV failure, nosocomial infection, duration of NIV, length of stay and mortality.

Results: 4900 patients were hospitalized in our medical intensive and 89 patients met the inclusion criteria. Bronchiectasis was associated with other pulmonary diseases in 16 patients. Bronchitis was the most common etiology of exacerbation (56%). three patients pseumononas aeruginosa in sputum cultures. NIV was successful in 71% of patients. The mortality was 19%. Length of stay was 12 ± 10 daysThe NIV failure group was older 67 ± 14 vs 59 ± 18, had a lower pH (7.25 ± 0.96 vs 7.31 ± 0.068) and elevated SPAS II score (33 ± 15 vs 24 ± 10). Factors associated with NIV failure were the SAPSII score (OR 2, IC 95% (1.5–3.02), p = 0.038) and the pH at admission (OR 6, IC 95% (2.04–8.06), p = 0.012). Nosocomial infection (54 vs 5%) and mortality (30 vs 2%) were significantly lower in NIV success group. However, Duration of NIV and length of stay were similar between the two groups.

Conclusion: NIV in bronchiectasis exacerbation reduce mortality and nosocomial infections. The rate of NIV failure was high. The SAPSII score and pH were the predictive factors of NIV failure.

Correspondence: Denault Marie-Hélène - marie-helene.denault.1@ulaval.ca

Annals of Intensive Care 2018, 8(Suppl 1):F-05

Introduction: Noxious effects of hyperoxia have been described for more than 60 years. Recent studies on stable obesity-hypoventilation syndrome (OHS) patients have risen concern about hyperoxia-induced hypercapnia in this population, but many questions remain unanswered. How important is this risk in acute care? Is there a risk in non-OHS obese patients as well? This study aims to evaluate the risk of hyperoxia-induced hypercapnia in post-op obese cardiac surgery patients.

Patients and methods: 30 obese patients (BMI > 30) having coronary artery bypass grafting (CABG) are currently being recruited. Informed written consent is obtained before inclusion. With a crossover design, we are comparing two oxygenation strategies immediately after extubation, in terms of their effect on arterial partial pressure of carbon dioxide (PaCO2)—a peripheral oxygen saturation (SpO2) target of > 95% achieved with manual titration (control period) and a SpO2 target of 90% achieved with automatic titration by a closed-loop system (FreeO2 period). Every patient is delivered oxygen according to both strategies for 30 min each, in a randomized order. FreeO2 (Oxynov, Québec, Canada) is a closed-loop oxygen delivery system which titrates oxygen flow according to the patient’s real-time SpO2 and a target set by the physician. SpO2, oxygen flow and heart rate are continuously recorded by FreeO2 during both periods. Hemodynamic parameters are checked at 10 min intervals and arterial blood gases are collected at the end of each period. The primary outcome is the change in PaCO2 between periods.

Results: 17 patients have been included so far and data for 16 patients are presented here. 13 are men, mean age is 60 ± 8 years and mean BMI is 34 ± 4 kg m². 9 patients are responders, as shown by an elevation in PaCO2 in control period compared to FreeO2 period, and 7 patients are non-responders. Two distinct oxygenation profiles are being studied (see Fig. 1). The mean SpO2 for FreeO2 period is 91.1 ± 1.3% while it is 97.2 ± 1.4% for control period. Mean oxygen flow is 0.7 ± 0.7 L min for FreeO2 period (equivalent to < 24% FiO2) while mean FiO2 is 51 ± 10% for control period.

Conclusion: FreeO2 is helpful in maintaining oxygenation within target. High SpO2 values may cause an elevation in PaCO2 in obese patients after CABG, but more analyses are required for this outcome. Characteristics of responders and non-responders will be analyzed after study completion.

Correspondence: L’Her Erwan - erwan.lher@chu-brest.fr

Annals of Intensive Care 2018, 8(Suppl 1):F-06

Introduction: Optimal patient-ventilator synchronisation improves patients’ comfort and avoids unnecessary WOB. Asynchronisms are currently diagnosed by direct bedside ventilatory curves analysis, which is yet operator dependent and time-consuming, or by invasive methods including esophageal pressure or diaphragmatic electromyogram measurements. The main purpose of this study was to assess the relevance of Curvex as a noninvasive diagnostic and classification tool for asynchronism management.

Patients and methods: This project is based on a prospective physiological tracing data-warehousing program (Rea STOC, clinicaltrials.gov # NCT02893462) that aims to record 1500 consecutive ICU patients, over 3-years. All consecutive patients were recorded for a 2-hours period during 24-h following ICU admission. All measurements were recorded with the patient laying supine, with a 30° bed angulation. Raw ventilatory pressure and flow curves were transferred to a centralized server using a dedicated network. The physician in charge of the study was informed of the online analysis on a routine basis. Physiological recordings were associated with metadata collection. Asynchronisms detection is based on a non-parametric hypothesis testing (Random Distortion Testing), that requires no prior information on the signal distribution. Beside Asynchrony Index monitoring (AI), five asynchronism’s types were qualified—ineffective efforts (IE), short cycles (SC), multiple cycles (MC), prolonged inspiration (PI) and premature cycling (PC).

Results: 72 patients were included (age 61.2 ± 13.8 yr + 52 male 20 female + SAPS II 53.6 ± 2.4), of whom 15 with chronic respiratory diseases. Reason for admission was respiratory distress (47%), hemodynamic failure (29%), and neurological failure (22%). 92% patients where under invasive mechanical ventilation (12.8 ± 18.7 days) and 32% under sedation. Asynchronies were frequent (less 10% patients with AI < 10%) and mean AI was 34 ± 18%. IE were the most frequent asynchronisms (39.7 ± 28.6%). Concordance between visual and automatic evaluation was 97.2%. An attempt to reduce AI according to a standardized protocol was performed in 56/72 patients. Significant AI reduction (≥ 10%) was observed for only 10 patients (mean AI reduction = 15.5 ± 3.9%). PC were reduced for 28/56, SC for 27/56, PI for 25/56, MC for 25/56, and IE for 20/56 patients. No correlation was depicted between AI and patients status.

Conclusion: Asynchronies are frequent during MV and can be efficiently and noninvasively monitored using Curvex. The impact of asynchronism on patients’ prognosis and the availability of correction measures require further investigation.

Correspondence: Melone Marie-Anne - m_a.melone@icloud.com

Annals of Intensive Care 2018, 8(Suppl 1):F-07

Introduction: About 30% of patients with Guillain-Barré syndrome (GBS) need mechanical ventilation (MV), of whom roughly 75% may develop pneumonia, in particular those with swallowing impairment. This trial aimed at assessing the benefit of early mechanical ventilation (EMV) in the prevention of onset of pneumonia in GBS patients.

Patients and methods: Design—this was a single centre, open-label, randomized controlled trial performed on two parallel groups. Patients—50 ICU adults admitted for GBS and risk factors for endotracheal intubation, i.e. time from onset to admission of < 7, inability to lift the head, and forced vital capacity of < 60%. Intervention—patients were randomized to receive EMV via face mask or endotracheal intubation owing to the presence or absence of impaired swallowing (experimental arm), or to receive conventional care that may include invasive mechanical ventilation whenever acute respiratory failure occurred (control arm). Outcomes—the primary outcome was the time to onset of pneumonia up to ICU discharge (or 90 days, pending of which occurred first). Secondary outcomes included the total number of episodes of pneumonia, time to and length of mechanical ventilation, length of hospital stay, mortality and any serious adverse events.

Results: Twenty-five patients were randomized in each group. In EMV, 44% required invasive mechanical ventilation. There was no evidence for clinically relevant differences in baseline characteristics between groups (Vital capacity—47% [40 + 54], PaCO2—5.3 kPa [4.6 + 5.6], PaO2—11.8 kPa [9.6 + 12.3]). 48% of patient in experimental group and 52% in control group had swallowing impairment. There was no significant difference between groups for the time to pneumonia (p = 0.50, Gray test). There were 16/25 (64%) and 15/25 (60%) of patients that presented at least one episode of pneumonia (p = 1.00), in the experimental and control groups, respectively. During follow-up, 16/25 (64%) patients in the control group were eventually mechanically ventilated (all invasively) and 25/25 (100%) in the experimental arm (p < 000.1). The time on ventilator was non-significantly shorter in the experimental arm (14 [7 +29] versus 21.5 [17.3 +35.5], p = 0.10). There were no significant differences between groups for length of hospital stay, neurological scores, the proportion of patients who needed tracheostomy, in-hospital death, or serious adverse events.

Conclusion: In Guillain-Barré syndrome patients at high risk of respiratory failure, early mechanical ventilation did not prevent the onset of pneumonia.

Correspondence: Messika Jonathan - jonathan.messika@aphp.fr

Annals of Intensive Care 2018, 8(Suppl 1):F-08

Introduction: French hospital fellows’ missions are threefold—teaching, research and care. Although some diplomas are welcome (Master 2, for instance), prerequisites to become a fellow are limited to graduation in a specialized medical degree, and a medical thesis. Hence, fellows in the ICU may have a very varied background. We therefore decided to perform a survey of the scientific, clinical and educational training of intensive care medicine fellows, at the dawn of a new degree in intensive care medicine, and analyse heterogeneity causes.

Patients and methods: We conducted a national survey among the current and former fellows (2014–2017) of French medical and mixed critical care units, with an electronic questionnaire.

Results: One hundred and fifty-seven current and former fellows were contacted, among whom 123 (78%) responded (age—33 years [32–35], men 60.2%)—56.9% completed a complementary degree in intensive care, while 37% were on their way to graduate. Their background speciality was anaesthesiology (47.2%), nephrology (13%), cardiology and pneumology (12.2% each), internal medicine (11.4%), or another medical speciality (4%). One hundred and eighteen respondents (87.8%) had at least one university diploma. Seventy-one respondents had graduated with a Master 2 degree and 16 of them had a Ph.D. while 29 anticipated its completion. Medical education training was mainly provided by the Universities, when the fellows entered their position (n = 38 + 30.9%). The scientific production of the respondents was valorised by at least one presentation in a congress (83.7%) or by a scientific article published or to be published as first author (51.2%). There were heterogeneities—more Master 2 degree holders in Ile-de-France and older fellows were more trained in pedagogy.

Discussion: The increasing demand for fellow positions and the stability of their number calls for a clear definition of criteria for position attribution. Although a Master 2 degree is usually required to perform experimental research, no counterpart is necessary for clinical research.

Conclusion: This survey shows a diverse, but solid scientific and clinical training of French critical care fellows, attested by a high rate of Master degree and Ph.D. This inventory must be addressed in other medical specialities.

Correspondence: Maury Eric - eric.maury@aphp.fr

Annals of Intensive Care 2018, 8(Suppl 1):F-09

Introduction: International guidelines recommend ultrasound (US) guidance for central venous catheter (CVC) insertion. However, evidence is lacking for several aspects of guidance such as probe shape or whether the needle has to be in plane (IP) or out-of-plane (OOP). We assessed these issues in a randomized trial.

Patients and methods: 53 Health Care Workers [mean age—27 [22–31], male (28), senior (n = 18), resident (n = 13), student (n = 22), having previously inserted > 10 US guided CVC (n = 29)] were proposed to take part to a pilot study. Naive HCWs received a training (course and hands-on). Right Internal Jugular Vein (RIJV) and Right Axillary Vein (RAV) with either linear probe (LP) [HFL38, with the M-Turbo^®, Sonosite, Bothewell, MA], or micro-convex probe (MCP) [GE8C, with a Vivid 7^®, General Electric, 78457 Velizy, France] using IP and OOP approaches were performed on an inanimate manikin (Blue Phantom II, CAE Healthcare St. Louis, MO). Procedure was stopped at 180 s. The order of the 8 punctures—site probe approach—was randomized. Success at first pass, number of attempts (needle passes), success, times between skin contact and needle skin penetration and between needle skin penetration and liquid back flow in the syringe were recorded. Qualitative and quantitative values are expressed as number (percentage), and median (range), and were compared using the Wilcoxon matched pairs test and the Fisher exact test, respectively.

Conclusion: This study suggests no benefit of MCP over LP for CVC insertion and supports IP approach for axillary puncture.

Correspondence: Satre Buisson Lea - leasatrebuisson@gmail.com

Annals of Intensive Care 2018, 8(Suppl 1):F-10

Introduction: Simulation training is an effective teaching tool, preeminently in high-risk situations such as advanced cardiac life support (ACLS). Sudden cardiac arrest (CA) in clinical conditions or high fidelity simulation (HFS) represents a highly stressful medical situation. Stress measurement and its correlation to technical and non-technical performances could be a tool to improve teaching methods. Heart rate variability (HRV) is mediated by the autonomic nervous system and a lower HRV reflects parasympathetic inhibition therefore it could reflects stress. The aim of this study was to validate the feasibility of objective stress measurement by heart HRV monitoring during CA in HFS. Analyzing the relation between ANI, self-reported stress and performance was the second objective.

Patients and methods: An open monocentric prospective study was conducted from December 2015 to December 2016 in Lille’s medical simulation center PRESAGE. Intensive care residents from Lille Medical University played team leader in a CA scenario. Objective stress was measured from recordings of three leads electrocardiogram portable device (PhysioDoloris^® monitor, MDoloris Lille, France). HRV was obtained by analyzing time interval between successive R waves and transformed in the Analgesia Nociception Index (ANI). ANI gives both a qualitative and quantitative measurement of HRV. Self-reported stress was assessed with a visual analog scale (VAS). Results were expressed in numbers (%), mean (± Standard Derivation). Correlations were tested using the non-parametric Spearman rank test (Rho). p < 0.05 was considered significant.

Results: Sixty-four intensive care residents were monitored for HRV in simulation training. 36 signals (56.3%) were fully interpretable. Team leader’s self-reported pre-HFS stress VAS was 43.3 (± 22.4) and maximal stress VAS was 58.7 (± 21.7). Team leaders’ maximal heart rate was 157 bpm (± 20). Minimal ANI, reflecting intense stress was 33.7 (± 9.9). Objective and subjective stress of each team leader is shown in Fig. 1. There was a significant negative linear correlation between minimal ANI and maximal HR (Rho = − 0.52, p = 0.001). There was no significant correlation between self-reported stress VAS (neither pre HFS or maximal stress) and minimal ANI.

Conclusion: HRV monitoring is a feasible method to evaluate continuous physiological stress for team leaders in highly stressful simulation-teaching. Upgrading signal connection by Bluetooth 4.0 or Wi-fi could improve the method. Focusing on specific stressful time points might improve stress assessment and its correlation with performance.

Correspondence: Ouanes Islem - ouanes.islem@gmail.com

Annals of Intensive Care 2018, 8(Suppl 1):F-11

Introduction: Simulation training has become available in Health sciences faculties and proposed in many specialties. Intensive Care is one of the fields of development of simulation based training. The aim of the present study was to report the experience of the Faculty of Medicine of Monastir Simulation Center in training medical students and residents in Intensive Care and to compare their respective perceptions.

Patients and methods: This was a descriptive study including students (5th year of the medical curriculum) and residents who received training during the last academic year (2016–2017), in the simulation center during their ICU traineeship. Simulation training was based on high-fidelity mannequins for students and seminars with high fidelity and procedural simulation training for residents. Three sessions per group were organized for students and a total of five sessions for residents. We collected participant characteristics and used Likert scale (from 1 to 5) to assess participant satisfaction, simulation fidelity, impact on clinical practice, stress level and instructor behaviors. Chi 2 test was used to compare students’ and residents’ perception of the simulation based-training.

Results: During the study period 91 students (of the 269 students’ whole promotion) and 88 residents actively participated at least in one of the simulation-based training sessions. Median students’ and residents’ ages were respectively 24 years (24–24) and 29 years (27–30). Participants were mainly females in 78 and 73.8% respectively in students and residents. Prior to the study period all the students received simulation-based training, whereas 48.9% of residents assisted for the first time in a simulation based training. For all participants, satisfaction (levels 4 and 5 of the Likert scale) was reported in 87.2%, simulation fidelity in 60.3%, possible impact on clinical practice in 77.1%, 40.8% perceived training as stressful and 85.5% noted favorably instructors’ behaviors. Compared to ICU residents, students were significantly less satisfied (80.2 vs 94.3% + p = 0.005), perceived training more stressful (59.3 vs 21.6% + p < 0.001) and noted less favorably instructors’ behaviors (79.1 vs 92%; p = 0.014), whereas the other items were comparable (Fig. 1).

Conclusion: We observed globally a favorable appreciation of ICU students and residents of the simulation based training differences in perceptions’ items could be mainly explained by different previous experiences with the real life in the hospital of the two study groups.

Correspondence: Dachraoui Fahmi - dachraoui.fahmi@gmail.com

Annals of Intensive Care 2018, 8(Suppl 1):F-12

Introduction: Social media (SM) can be used to improve or enhance professional networking and education, organizational promotion, patient care, patient education, and public health programs. Little data are available about perceptions among young medical practitioners although unconstrained online social activity is the norm for them. The aim of the present study is to describe the actual use and perception of the role of SM by Tunisian residents in intensive care medicine.

Patients and methods: This was a prospective observational study involving intensive care residents at the university hospitals of Sousse, Monastir and Mahdia (Tunisia). Residents were invited to participate and answer an anonymous questionnaire about social media in clinical medicine, accessibility to the internet, their personal and academic use of SM and their perception of the use of SM. The questionnaire comprised 22 different types of questions—Likert style, multiple choices, yes no, ranking and short and open response questions.

Results: Overall, 52 residents (36 females, age = 27.5 ± 3.9 years) responded to the questionnaire (66%) + 84.6 and 65.5% of them had smartphones and laptops, respectively. Sixty-five percent had mobile Internet access or a home WIFI network, while 13.5% said they had access to the Internet only through their faculty or hospital where they were exercising. Facebook and YouTube were the most frequently daily used SM for both personal and academic use, 94–32.7 and 61.6–34.6% respectively. In contrast, 94.2% said they had never used Twitter. The daily use of medical applications on Smartphones and weekly e-learning platforms were reported by 65.4 and 63.5% of residents respectively. Eighty-eight percent of the residents thought that YouTube should play a key role in their training. The use of SM was supposed to improve the collaboration between residents and with their supervisors respectively in 88.4 and 86.5%. Nearly half of the residents (42.3%) were opposed to communication with patients and or their families by SM. Respondents in the survey suggested that medical schools should use SM to keep them informed on training cycles, and that they should adopt a policy that would govern the use of SM in 94.2 and 88, 6% respectively.

Conclusion: Facebook and YouTube are the most used social media by Tunisian residents in intensive care medicine. They remain however, little contributors in their training course.

Correspondence: Allary Chloé - chloe.allary@gmail.com

Annals of Intensive Care 2018, 8(Suppl 1):F-13

Introduction: Hospitals are encouraged to edit local antibiotic therapy guidelines. ANTIBIOGARDE^® is an electronic antibiotic prescription referential developed by a multidisciplinary team of French physicians, regularly updated, and locally customizable, which has been purchased by more than 170 French hospitals. We compared adequacy of initial antibiotic prescription by ICU clinicians, ANTIBIOGARDE^® proposal and national or international guidelines.

Patients and methods: Between January and June 2016, initial antibiotic prescriptions in an ICU were retrospectively analyzed when microbiologically documented. ANTIBIOGARDE^® and guidelines proposals were simulated, considering data available at the time of initial prescription. Adequacy was defined when all bacteria responsible for infection were sensitive to at least one prescribed proposed antibiotic. National guidelines were used when published after 2011. Otherwise, most recent international guidelines were used.

Results: 120 initial prescriptions were analyzed (45 monotherapy) in 97 patients (median age 63y, median SAPS II 53, median SOFA on prescription 7, ICU mortality 41%, 17% immunocompromised). Main sources of infection were lung (n = 75) and intra-abdominal (n = 23). Leading isolated bacteria were Enterobacteriaceae (n = 67, antibiotic resistance in 14), Streptococci (n = 55), non-fermenting Gram negative bacilli (n = 27, antibiotic resistance in 12) and Staphylococci (n = 25, resistance to methicillin in 5). In the 120 clinical settings analyzed, there was a proposal by ANTIOGARDE^® in 97 (81%) and a guideline available in 102 (85%) (p = 0.39). Initial antibiotic regimen prescribed by clinicians was adequate in 107 120 settings (89%, 95% CI [82–94%]). Considering only settings with a proposal, ANTIBIOGARDE^® was adequate in 90 97 (93%, 95% CI [86–96%]) (p = 0.36 vs clinicians), and guidelines in 102 102 (100%, 95% CI [96–100%] (p < 0.001 vs clinicians and p < 0.001 vs ANTIBIOGARDE^®). Considering all settings and no proposal as inadequacy, ANTIBIOGARDE^® was adequate in 90 120 (75%, 95% CI [67–82%]) (p = 0.004 vs clinicians), and guidelines in 102 120 (85%, 95% CI [78–90%] (p = 0.34 vs clinicians and p = 0.05 vs ANTIBIOGARDE^®). The antibiotic regimen prescribed by clinicians had a narrower spectrum against Gram negative bacilli than ANTIBIOGARDE^® and guidelines proposals in 70 and 68% of cases respectively. Proportions of prescriptions, ANTIBIOGARDE^® proposal and guidelines proposal including antibiotics active against methicillin resistant staphylococci were 12, 25 and 37% respectively (p < 0.00005).

Conclusion: ANTIBIOGARDE^® electronic referential could allow even non expert ICU physicians to adequately (93%) prescribe initial antibiotic therapy in most clinical settings (81%).

Correspondence: Tabatchnik Xavier - xavier.tabatchnik@gmail.com

Annals of Intensive Care 2018, 8(Suppl 1):F-14

Introduction: Intubation is plagued with a high morbimortality, especially in emergency situations. It is now acknowledged that a seated position allows for optimized preoxygenation (1). However, there is no guideline concerning the patient’s position for intubation. The patient is most often laid in a supine position, leading to a higher risk of aspiration (2). Face-to-face intubation in sitting position (FtFi) would allow for an easier intubation and a lower morbidity. We focused on learning the FtFi technique using the Macintosh laryngoscope and the Airtraq videolaryngoscope in simulated difficult intubation situation and comparing the performance of the FtFi with the classic technique.

Patients and methods: The participants would intubate a high-fidelity manikin (SimMan3G, Leardal, Norway) configured with a tongue edema (Cormack 2b-3). For each trial, time to intubate (TTI), success and complication rate, intubation difficulty and glottis exposure were noted. In classic position, three trials were performed with the Airtraq followed by the laryngoscope in order to obtain baseline parameters. In FtFi, at least 30 intubations were performed by each participant for each device. The utilization order was randomized.

Results: Thirty physicians, with an experience of at least 200 intubations each, were included. Figure shows the learning curves of the FtFi based on the evolution of the TTI measured for the Airtraq and the laryngoscope. In classic position, the mean TTI with the Airtraq was 30.5 ± 19.1 s versus 27.3 ± 16.7 s with the laryngoscope (p = ns). In FtFi, once the technique mastered, the TTI was 14 ± 2.9 s with the Airtraq versus 17.5 ± 2.4 s with the laryngoscope (p < 0.05). Success rate, TTI, complication rate, intubation difficulty and glottis exposure were better using FtFi versus classic intubation (p < 0.05). These parameters were even better with the Airtraq than with the laryngoscope (p < 0.05).

Discussion: The learning profile of FtFi is different between the Airtraq and the laryngoscope. It could be due to the participants’ lesser familiarity with the Airtraq. The better performances in FtFi could be due to better ergonomics allowing easier glottis exposure and learning (3).

Conclusion: Face-to-face intubation in sitting position is easy to learn. It provides better performances and fewer complications than the classic intubation technique which might result in a lower morbidity. The Airtraq provides even better results than Macintosh laryngoscope. All participants recommend their colleagues to be trained in face-to-face intubation.

Correspondence: Heilbronner Claire - claire.heilbronner@aphp.fr

Annals of Intensive Care 2018, 8(Suppl 1):F-15

Introduction: Acute chest syndrome (ACS) is the leading reason of admission in Pediatric Intensive Care unit (PICU) for patients with Sickle Cell Disease (SCD). Non invasive ventilation (NIV) is nowadays unavoidable for the management of respiratory failure in PICU. Few studies assessed NIV’s impact on ACS outcome. We sought to investigate the clinical practice and particularly the use of NIV for ACS management in PICU or high dependency care unit (HDU) in France and French speaking countries.

Patients and methods: Participants were sought via e-mail through membership in the French Group for Paediatric Intensive Care Emergencies (GFRUP) to complete a 78-questions online survey for children admitted to PICU during the 1 year (2015). Demographics, clinical data and management’s description of ACS were collected.

Results: Seventeen PICUs or HDU with a total of 145 and 95 beds, respectively answered to the survey from June 2016, 10th to November 2016, 10th. Fifteen centres (88%) had a local transfusion center and 14 (82%) had a local blood centre and 14 (82%) were in connection with a SCD reference centre. During the 1 year 2015, 360 SCD patients were managed (median 7 SCD per centre, range 0–127) of which 137 (38%) for an ACS (median 4ACS per centre, range 0–34). Median length of PICU stay for ACS was 5 days, range 2–11 days. Three centres (totalling 62 beds) in the Parisian agglomeration hosted 146 SCD patients (40% of all SCD patients reported) with a total of 83 ACS (65% of ACS reported). Simple blood transfusion was performed in 73 patients with an ACS (53%). Sixteen patients (12%) received exchange transfusion. For patients who observed an ACS—bronchodilatators nebulisations were administrated in 47 (34%) of them, incentive spirometry was performed by 94 patients and 64 (47%) received standard oxygen therapy. Among non-invasive respiratory support, NIV with bilevel pressure was the most frequent (n = 68, 50%) before CPAP (n = 23, 17%) and high flow oxygen (n = 21, 15%). The proportion of patients on NIV was up to 71% in the centres hosting more ACS patients.

Conclusion: Despite the absence of evidence from randomized controlled trials NIV is nowadays commonly used in PICU and HDU for SCD patients with ACS, especially in centres taking in charge a high number of SCD patients. Future physiological studies and randomized controlled trials might help to choose between the different ventilatory support options for ACS.

Correspondence: Emeriaud Guillaume - guillaume.emeriaud@umontreal.ca

Annals of Intensive Care 2018, 8(Suppl 1):F-16

Introduction: Restrictive transfusion practices are increasingly used in pediatric intensive care unit (PICU), but evidence is lacking for deciding transfusion in children with severe acute respiratory distress syndrome (PARDS). This study aims to describe the red blood cells (RBC) transfusion practices during PARDS, to identify the main transfusion determinants in this population, and to describe the evolution of the patients after the transfusion.

Patients and methods: Patients with PARDS were prospectively screened in the PICU of a university affiliated pediatric hospital, from June 2015 to October 2016. Patients aged between 3 days and 18 years with a PARDS lasting at least 6 h were included. Baseline patient characteristics, PARDS data (for a period of 72 h), RBC transfusion characteristics, and main outcome data in the PICU were retrospectively collected in the electronic medical records. PARDS severity was classified based on PALICC definition. Results are reported as median (25^th–75th percentile), and comparisons were made using Mann–Whitney test and Friedman test.

Conclusion: In our PICU, a relatively restrictive policy of RBC transfusion was observed even in patients with severe PARDS. Decision to transfuse seemed associated with the general severity status of the patient and with the hemoglobin level. Further studies are needed to explore the generalizability of these findings, and to investigate the impact of transfusion on oxygen transport consumption balance in pediatric acute respiratory distress.

Correspondence: Nardi Nicolas - nicolas.nardi@hotmail.fr

Annals of Intensive Care 2018, 8(Suppl 1):F-17

Introduction: Pediatric acute respiratory distress syndrome (PARDS) is associated with a high mortality rate of 30%. Lung protective strategies in ARDS advocate using low tidal volume and high positive end expiratory pressure (PEEP) while limiting the inspiratory plateau pressure in the recent international consensus conference on PARDS (PALICC)¹. Our objectives were to evaluate the adherence to the guidelines for lung protective ventilation proposed by the PALICC and determine whether variations in tidal volume were related to the severity of oxygenation impairment.

Patients and methods: All patients, admitted to our tertiary pediatric ICU from January 2013 to May 2015, who received conventional ventilation for severe PARDS using PALICC definition and who had data to calculate ideal body weight (IBW) were retrospectively included. We excluded children with pulmonary bronchodysplasia and those admitted after cardiac surgery. Expiratory tidal volume (VTe), PEEP, peak inspiratory pressure, mean airway pressure, FiO2 and PaO2 were collected from electronic medical record (ICCA, Philips), using SQL Server 2008 (Microsoft, USA) + data were collected for 10 days following onset of conventional mechanical ventilation. Adherence proportions were analyzed with the non-parametric Cochran’s Q test, using cut-off values of VTe ≤ 8.0 mL kg. Mixed model analysis was performed to test the hypothesis that variations in tidal volume were related to the severity of oxygenation impairment.

Results: Twenty-one patients reached inclusion exclusion criteria, with median age of 5 years (interquartile [IQR] + 2.5–12 years) and a median weight of 17 kg ([IQR] + 12–30). Mortality rate was 38% (8/21) with a median duration of mechanical ventilation of 10 days ([IQR] + 7–32 days). On 1780 measurements of respiratory parameters, 72% of VTe measurements were ≤ 8.0 mL kg with median VTe of 6.7 mL kg ([IQR] + 5.6–8.3). During the first 10 days of conventional mechanical ventilation (159 patient-days), there were a significant variations in the daily rates of adherence to low tidal volume ventilation (p < 0.001) and a significant decrease in VTe with the highest oxygenation index (p = 0.02).

Conclusion: In our centre, adherence to PALICC guidelines for low tidal volume ventilation in PARDS is not perfect. However the adherence to protective ventilation strategy was higher in the most severe patients, in line with the recommendations. We plan to test if the implementation of a clinical decision support system can improve the adherence to guidelines at the bedside.

Correspondence: Tencer Jérémie - jeremie.tencer@gmail.com

Annals of Intensive Care 2018, 8(Suppl 1):F-18

Introduction: Children with polyhandicap are a complex medical and ethical situation. There are few studies about their admission in intensive care units, despite them having a strong risk of acute diseases. Our objetive was to describe the hospitalization of children with polyhandicap in intensive pediatric care units.

Patients and methods: This is a retrospective study in 4 pediatric intensive care units in Paris. Children with polyhandicap hospitalized in emergency from January 1st 2013 to December 31st 2015 were included. We looked at epidemiological datas, medical supports and medications at admission and treatment level (respiratory, circulatory, neurologic, invasive procedures), and limitations of treatments.

Results: 96 patients were included, representing 143 hospitalizations. At admission average age was 11 ± 5 years, mean number of medical support per patient was 1 ± 0.6 and mean number of neurological medication per patient was 2.5 ± 1.8. The main admission criteria were—respiratory failures (59%), neurological failures (20%) and circulatory failures (10%). Respiratory support was needed in 56%, circulatory support in 31% and nutritional support in 20% of patients. During these hospitalizations 25 consultation meetings have been held. Forty-two percent of patients did not have to be submitted to any invasive procedures. Overall mortality was 16%.

Conclusion: children with polyhandicap admitted in intensive care units are at school age and are mostly having respiratory or neurological failures. There is a strong rate of rehospitalisation. There are few therapeutic limitations that have been found. Anticipating a crisis is a major point to provide the best care for these children.

Correspondence: Béranger Agathe - agathe.beranger@gmail.com

Annals of Intensive Care 2018, 8(Suppl 1):F-19

Introduction: Pharmacokinetic parameters are altered in critically ill patients. For instance, in adult patients, it has been well demonstrated that augmented renal clearance results in subtherapeutic antibiotic concentrations. Our objectives were to build a pediatric population pharmacokinetic model for Piperacillin, in order to optimize individual dosing regimen.

Patients and methods: All children admitted in pediatric intensive care unit, aged less than 18 years, weighing more than 2.5 kg, and receiving intermittent Piperacillin infusions were included. Piperacillin was quantified by high performance liquid chromatography. Pharmacokinetics were described using the non-linear mixed effect modelling software MONOLIX. Monte Carlo simulations were used to optimize dosing regimen, in order to maintain plasma concentration above the minimum inhibitory concentration (16 mg L⁻¹ for Pseudomonas aeruginosa) throughout the dosing interval (100% fT > MIC).

Results: We included 50 children with a median (range) post natal age of 27.2 (1.1–222.9) months, median (range) body weight of 11.9 (2.7–50) kg, median (range) PELOD-2 score of 4 (0–16) and median (range) estimated creatinine clearance of 142 (29–675) mL.min^-1.m^-2. A one compartment model with first-order elimination adequately described the data. Median (range) values for Piperacillin clearance and volume of distribution were respectively 3 (0.71–10) L h⁻¹ and 3.8 (0.72–25.8) L. Body weight (allometric relationship), estimated creatinine clearance and PELOD-2 severity score were the covariates explaining the estimated between subject variability. A third of the cohort attained the target, according to our dosing regimen and to the European guidelines. To reach the target and according to the simulated dosing regimens, children with acute kidney injury should receive intermittent infusion every 6 h, administered on 30 min. Those with augmented renal clearance should receive a continuous infusion.

Conclusion: To reach the target, standard intermittent Piperacillin dosing regimen in critically ill children is not appropriate. In addition to body weight, dosing regimens should take into account the creatinine clearance. Continuous infusion is adequate for children with augmented renal clearance. Piperacillin individualized dosing regimens and therapeutic drug monitoring are mandatory in pediatric intensive care unit.

Correspondence: Ros Barbara - rosbarbara33@gmail.com

Annals of Intensive Care 2018, 8(Suppl 1):F-20

Introduction: All data support the need for early recognition, evaluation of pain in the NICU. Multiparametric analysis including physiological parameters could be useful to have a more objective evaluation of pain in the NICU compared to scales built on external-evaluation. The Newborn Infant Parasympathetic Evaluation (NIPE^®) was developed to assess pain in newborns and infant, from preterm to the age of 2 years.

Patients and methods: We conducted a monocentric, prospective study to compare the instantaneous NIPE^® index value (NIPEi^®) to the DAN scale during acute procedural pain (PICC line insertion) in preterm infants (under 37 GW). The operators and the nurse were blinded to the continuous recording of NIPEi^® during the entire procedure. DAN scale was assessed every 5 min by a third person, trained to this scale and blinded to NIPEi^®. A direct correlation assessment between the DAN scale and the NIPEi^® was performed by calculating the Pearson’s linear correlation coefficient. The differences between the NIPEi^® of non-painful (DAN < 3) and painful (DAN ≥ 3) infants were estimated by the Wilcoxon–Mann–Whitney test. The usefulness of NIPEi^® as a new tool for pain assessment in neonates was estimated by the corresponding ROC curve. Our study was approved by our local ethic institutional review board.

Results: Thirty-five preterm infants were included, NIPEi^® data were incomplete in 3 infants. Fifty percent of newborns were born before 31 GW, and 66% had non-invasive respiratory support (continuous positive airway pressure CPAP). At the time of the procedure, newborns had a median post-natal age of 3 days and a median weight of 1180 grams. There was a moderate correlation between the NIPEi^® index and the DAN scores (r = 0.605 + p < 0.001). The median NIPEi^® index was 59 for non-painful events vs. 43 for painful events, p < 0.001. The area under the ROC curve was 0.92. For a threshold of NIPEi^® < 50, the sensitivity was 91.3%, the specificity was 83%. Positive likelihood ratio was 5.36 and the negative likelihood ratio was 0.10 (Fig. 1).

Conclusion: We showed a correlation between the DAN scale and the NIPEi^® index for pain assessment in preterm infants. The NIPE^® monitor could be a useful and non-invasive tool for pain assessment in neonates. Further studies are needed to confirm our results and to define more precisely the place of such monitors for pain evaluation in daily clinical practice in the NICU.

Correspondence: Javouhey Etienne - etienne.javouhey@chu-lyon.fr

Annals of Intensive Care 2018, 8(Suppl 1):F-21

Introduction: The aim was to identify factors associated with the occurrence of acute pituitary hormone dysfunction in children with moderate to severe TBI and to describe the impact of this dysfunction on the stability of the children.

Patients and methods: Prospective bicenter study including all children aged between 1 month to 17 years, admitted to PICU for a moderate-severe TBI and with an expected stay > 2 days. Setting—Pediatric intensive care units of Grenoble and Lyon, from 2010 to 2017. Endocrine explorations at the second morning following admission and 24 h before discharge were performed—cortisol 24 h cycle with free cortisol and ACTH dosages every 4 h (or 6 h if no central line) + free 24H urinary cortisol + TSH and T4L, 24H urinary LH and FSH, blood level of testosterone or estradiol for children aged > 10 years, and IGF1. Patients were classified as having cortisol insufficiency if all the cortisol dosages were < 200 nmol l and all ACTH were < 12 pg l. TSH deficiency was defined as T4 l < 11.5 pmol l and TSH < 4.2 mUI l. Gonadotropin defciency was defined as urinary LH < 0.14 UI 24 h and urinary FSH < 1.52 UI 24 h for males + urinary LH < 0.018 UI 24 h and urinary FSH < 2.24 UI 24H in female. Patients with deficiency (ACTH and any deficiency) were compared to those without deficiency in terms of hemodynamic instability, respiratory instability, neurological and infectious complications For continuous variables means and 95% confidence interval were calculated and compared by t student test. Chi-2 test was used to compare proportions.

Conclusion: We didn’t find any predictive factors of pituitary hormone deficiency in children with moderate-severe TBI justifying a systematic screening of those patients.

Correspondence: Fathallah Ines - ines822004@yahoo.fr

Annals of Intensive Care 2018, 8(Suppl 1):F-22

Introduction: Most intensive care unit (ICU) patients cannot make decisions themselves. Familiy members are actively involved in the care process as surrogate decision-makers and judges of care quality. However, family satisfaction with care is complex and is not clearly defined. The aim of this study is to evaluate the different procedures (reception book and staff education for aid and relationship) used in a new ICU to improve the family care.

Patients and methods: We included in our study patients who had spent more than 48 h in our department. A questionnairy, adapted to our population, was performed by our staff and validated by the Hygiene and Quality Care departement. We proceded by phone calls, 10 months after the inauguration of our ICU.

Results: Sixty-five questionnaires were included (Fig. 1). The average of age was 52 ± 20 with a sex ratio of 2. The average of the simplified acute physiology score (SAPSII) was 36 ± 17. The median stay was 7 days [4–16] with a total mortality rate of 25%. Mostly, we interrogated first-degree parents (n = 52). Only three families recieved reception book at admission. Visit in patient room was autorised only for 27% (n = 17) of family members. Only four persons said they were disturbed in visit hours for architectural reasons (tight space). Disponibility was found excellent in 40% (n = 24) of cases for medical staff, 45% (n = 26) for paramedicals. Informations provided by physicians were clear in 71.5% (n = 53) of cases. Fifteen of the family members (23%) asked psychology support. Only 15% of parents felt involved in the management care decisions. Most parents (87%) opted for making all dicisions related to doctors. Patients management was judged excellent in 74% (n = 48) of cases. There was no difference according to satisfaction between families of dead patients and families of surviving ones.

Conclusion: To optimise patients care and families satisfaction specific psychological support is necessary.

Correspondence: Ben Saida Imen - imen.bensaida@yahoo.com

Annals of Intensive Care 2018, 8(Suppl 1):F-23

Introduction: Acute exacerbations of COPD (AE COPD) is one of the most common causes of admission in the intensive care unit (ICU). The prognosis of COPD requiring ICU seems to be poor with high in-hospital and long-term mortality. Predictors of post ICU mortality for those patients are still not well defined. This study was carried out to define the predictors of 1 year mortality in patients admitted to the ICU with AE COPD.

Patients and methods: A prospective cohort study was performed in a Tunisian medical ICU between 2014 and 2016. A follow-up was performed on ICU survivors admitted for AE COPD during 1-year using phone interviews. Univariate and multivariate analyses were used to predict 1 year mortality. Stepwise backward elimination was performed to identify independent factors, with retention of predictors with p < 0.2.

Results: Among 119 patients admitted for AE COPD in ICU during the study period, 76(63.8%) patients were discharged alive. The mean age was 67.24 ± 10.9 years. 85% were male. Exacerbations were mainly caused respectively by—tracheobronchitis, 49(81.7%), community acquired pneumonia, 5(8.3%), lung edema, 4(6%) and bronchial hyperresponsiveness, 2(3.3%). 36(60%) required intubation, and the median duration of mechanical ventilation (MV) was 2 days with extremes ranging from 0 to 21 days. 30(39.5%) required inotropic agents. The most common ICU adverse events were respectively acute renal failure 20(26.3%), cardiac arrhythmia 21(35%), nosocomial infection in 9(15%) and difficulty in weaning 10(13.2%). The in-hospital mortality rate for the study group was 36%. The follow up was performed for 60(79%). The mortality rate at 1 year was 33.33%. Univariate analysis identified three factors associated with 1 year fatal outcome—invasive mechanical ventilation (p = 0.0015), WHO performance status (p = 0.0001) and decline in functional status (p = 0.000). Multivariate logistic regression analysis showed only two independent variables associated with long term mortality—SAPS II (OR, 1.14 + 95% CI [1.04–1.25] + p = 0.004) and performance status at discharge (OR, 3.11 + 95% CI [1.55–6.23] + p = 0.001).

Conclusion: In the present study, severity of illness and physiological reserve at discharge were identified as independent predictors of long term mortality after ICU discharge in AE COPD. Therefore, caregivers should allocate adequate strategies for the prevention and control of COPD patients.

Correspondence: Ben Saida Imen - imen.bensaida@yahoo.com

Annals of Intensive Care 2018, 8(Suppl 1):F-24

Introduction: Randomized controlled trials have confirmed the evidence and helped to define when and where NIV should be the first line treatment of acute exacerbation of COPD (AE COPD). Its effectiveness in preventing endotracheal intubation and reducing intensive care mortality is undeniable. Few studies have evaluated the relationship between ventilation modalities (noninvasive ventilation [NIV] versus invasive mechanical ventilation [IMV] and long-term survival in COPD patients treated in the intensive care unit (ICU). The aim of this study was to investigate one-year survival stratified by mechanical ventilation modality in COPD patients treated in the ICU.

Patients and methods: An observational prospective cohort study including all patients admitted to a Tunisian Tertiary ICU for AE COPD and discharged alive between January 2014 and December 2015. Patients were followed up via phone calls during 1 year after discharge. Characteristics on admission and outcomes after discharge were analyzed stratified by ventilation modality NIV vs IMV. The overall survival was analyzed on the basis of the Kaplan–Meier curves.

Results: During the predetermined period of data collection, the follow-up involved 60 patients. 24 patients were treated by NIV (group1) and 36 patients needed IMV (group2). There was no difference between the 2 groups in age (p = 0.69), severity of COPD (p = 0.39), physiological reserve at discharge (p = 0.14) and ICU readmission (p = 1). Short term outcomes were not different between the 2 groups—1-month readmission (16.7 vs 25% respectively in NIV and IMV, p = 0.52) and 1-month mortality (4.2 vs 19.4%, p = 0.18). However, one-year Mortality rate was lower in NIV group (16.7 vs 44.4%, p = 0.02). Long-term survival analyzed with the Kaplan–Meier method was significantly greater in the NIV group (p = 0.016) (Fig. 1).

Conclusion: In the present study while the characteristic of AE COPD patients’ prior to ICU admission and at discharge were similar, long term survival was higher in the NIV managed patients. Invasive mechanical ventilation seems to be a negative salient feature in the natural course of the COPD. Avoiding IMV by early treatment with NIV in AE COPD needing critical care is associated with better long-term survival.

Correspondence: Ben Saida Imen - imen.bensaida@yahoo.com

Annals of Intensive Care 2018, 8(Suppl 1):F-25

Introduction: The major challenge of intensivists is to reduce intensive care mortality rate. However, less attention is paid to outcomes beyond ICU discharge. A predictive triage model may influence decision making at ICU discharge and hence help caregivers to decrease short term post ICU mortality. The aim of this study was to develop a predictive triage model for discharge to identify patients at risk of 30 days post ICU mortality.

Patients and methods: An observational prospective cohort study was performed in a Tunisian medical ICU between January 2014 and December 2015. The study included all survivors after ICU admission. The mortality was assessed by telephone interviews at thirty days after discharge. Univariate and multivariate analysis were performed to identify candidate variables for the model.

Results: Among 573 patients admitted during the study period, 215 were included. The 30 days mortality rate was 16%. In univariate analysis, predictors of 30-days mortality of ICU survivors were Charlson comorbidity index (p = 0.001), baseline functional impairment ≥ 3 (p = 0.003), SAPS II ≥ 30 (p = 0.0001), Tachycardia (Heart rate ≥ 90b min) at discharge (p = 0.011), performance status ≥ 3 (p = 0.0001) and decline in functional status (p = 0.0001). Multivariate regression analysis identified the following independent risk factors, SAPS II ≥ 30 (OR, 3.258 + 95% CI [1.1–9.6] + p < 0.032), tachycardia at discharge (OR, 3.024 + 95% CI [1.01–9.11] + p < 0.049), decline in functional status (OR, 15.868 + 95% CI [15.18–48.56] + p < 0.000) and performance status ≥ 3 (OR, 6.57 + 95% CI [2.03–21.25] + p < 0.002). Based on these results, the following variables were used to develop the model—SAPSII ≥ 30, tachycardia, decline in functional status and performance status ≥ 3. These four items were weighted by their respective OR SAPSII ≥ 30, 3 points + tachycardia, 3 points + decline in functional status, 16 points + performance status ≥ 3, 6 points. The area under the ROC curve value of this predictive model obtained from the development cohort was 0.914 (95% CI [0.86–0.96]), a level deemed to be good discrimination (Fig. 1).

Conclusion: This discharge triage model is a simple, didactic, not time-consuming, costless and minimally invasive tool. It may afford the opportunity to critical care practitioners to identify patients at risk of early post-ICU mortality.

Correspondence: Merhabene Takoua - takouamg@yahoo.fr

Annals of Intensive Care 2018, 8(Suppl 1):F-26

Introduction: Background—intensive care physicians rarely have to manage children less than 18 years of age. In Tunisia, lack of pediatric intensive care units (ICU) requires their hospitalization in adult units. Purpose—To assess the effectiveness of child care in a polyvalent adult ICU and determine the relevant mortality risk factors.

Patients and methods: Methods—Retrospective study including children less than 18 years old at admission in the adult polyvalent intensive care unit of Abderrahman Mami Hospital in Ariana between January 1st 2008 and August 31th 2017. Data were analyzed using Student test or logistic regression as appropriate.

Results: Out of 4994 admissions, 80 patients (incidence = 1.6 1000 admission) were enrolled according to established criteria. 10.1% were asthmatic, 3.9% were diabetic, 2.6% have a neuromuscular disease and 83.4% have no previous comorbidities. Mean age was 14.9 ± 3 years [5–18]. Sex ratio was 1.42. Acute respiratory failure (69.7%) complicating community acquired pneumonia, neurologic disorders (15.6%) and acute poisoning (12.8%) were the main causes of admission. References from our emergency room represent 53.7% of cases, wards of our hospital 19.1%, pediatric units 5.2% and onco-hematologic units 5.2%. Mean PRISM during the first 24 h was 6 ± 6.6. Septic shock was observed in 3.9% of cases, 36.6% were invasively ventilated, 22.5% were non-invasively ventilated, 8.75% had a combination of invasive and non-invasive ventilation (NIV) and 41.2% did not receive ventilatory support. 20% of patient developed complications during their hospital stay—it was predominately nosocomial infection (16.6%) complicated by septic shock in 22.5%, acute renal failure (13.8%) and barotrauma (5.6%). Average ICU stay was 6.28 ± 9.72 days with a mortality rate of 21.3%. Multivariate analysis showed that independent factors predictors of mortality were—organ dysfunction at admission (acute respiratory failure, coma and hemodynamic failure), use of mechanical ventilation, occurrence of complications during hospital stay and PRISM ≥ 18.32.

Conclusion: Pediatric mortality remains high in our adult intensive care unit. Presence of organ failure at admission and occurrence of complications were the major factors for worse prognosis.

Correspondence: Prevedello Danielle - danielle.prevedello@ulb.ac.be

Annals of Intensive Care 2018, 8(Suppl 1):F-27

Introduction: Post-intensive care syndrome (PICS) has been recently described as a combination of physical, cognitive and mental impairments appearing during a stay in an intensive care unit (ICU). The prevention and detection of PICS require the participation of each category of healthcare workers. However, the level of knowledge is unknown. We sought to assess the awareness among our ICU staff in preparation for a follow-up consultation.

Patients and methods: The study used a short multiple-choice survey filled on a voluntary basis. All members of the staff were asked to fill the questionnaire over a one-week period. The assessment was composed by seven structured questions which aimed measure basic knowledge of post-intensive care syndrome and general strategies to diagnose that syndrome and the tests used.

Discussion: In spite of the sample size and of the voluntary character of this survey, these data indicate an ample room for improvement of the awareness of PICS in our ICU. The knowledge of the specific tests used for the assessment of PICS is mostly determined by the professional background. Hence, the implementation of a follow-up clinic is challenging as it requires educational training. Large similar projects have been successful after intensive training and awareness campaigns.

Conclusion: Knowledge of PICS is poor in the absence of training and education. Post-ICU follow-up clinics require skill and motivation from a large variety of ICU healthcare workers.

Correspondence: Wiem Ben Amar - wiembenamar@yahoo.fr

Annals of Intensive Care 2018, 8(Suppl 1):F-28

Introduction: Drafting a death certificate (DC) is a procedure considered as a part of doctor’s daily practice, especially in emergency and intensive care departments. This certificate represents a civil, social, epidemiological and medico-legal act. It can engage the liability of the certifying doctor. The objectives of our study were to examine the content of DC drafted in emergency and intensive care departments, assess the quality of writing, and analyze drafting errors.

Patients and methods: A prospective study extended over a period of 12 months from January to December 2015, including all DC emanating from emergency and intensive care departments and received in the forensic department of Habib Bourguiba hospital in Sfax.

Results: During the study period, 120 DC meeting the inclusion criteria were collected. Although confidential, the medical part of the DC was sealed by the doctor in onlyone third of cases. In the administrative section, nine socio-demographic parameters were studied. In 10% of the cases, less than four of the nine criteria were found. In the section concerning the certifying doctor data, 7 parameters were screened. 67.1% of the certifying doctors met at least six criteria. The most frequently missing parameter in this section was the identity of the person to whom the certificate was issued. The identity of the doctor was not mentioned in 10% of the cases. Forensic data (4 items) was complete in over three quarters of the certificates. Nevertheless, in 26.7% of cases, the medicolegal obstacle to burial box was left empty (8.3%) or not ticked even if judicial investigation was required (18.3%). The section on causes of death was the source of almost all of the drafting errors. We have classified these errors into six major ones, according the classifications reported in the literature. The percentage of certificate without faults was 13%. The most common major error was insufficient cause of death found in 40.4% of cases followed by incorrect sequence of causes of death (28.8%), medicolegal obstacle to burial not ticked although required (26.6%), several causes of death mentioned simultaneously (16.7%), unacceptable cause of death (10.8%) and mechanism of death mentioned instead of the cause of death (5.8%).

Conclusion: Our study showed that the quality of drafting of DC suffered from several insufficiencies, which encourages us to provide more effort in training doctors and to review the current official model of DC.

Correspondence: Jouffroy Romain - romain.jouffroy@gmail.com

Annals of Intensive Care 2018, 8(Suppl 1):F-29

Introduction: Septic shock is defined as a sepsis with hyperlactaemia greater than 2 mM after correction of hypovolemia requiring vasopressors to maintain MBP > 65 mmHg [1]. It can be observed in pre-hospital emergency medicine (PHEM). The use of a reliable portable device for measuring lactate in PHEM would allow a better evaluation of septic patient facilitating their orientation towards intensive care unit (ICU) or emergency department (ED). This portable delocalized biology device must be validated against the laboratory reference method (NFEN ISO 22870) [2]. The aim of this study was to clarify the validity of a delocalized measure of lactatemia.

Patients and methods: We performed a prospective study including 47 patients admitted into ICU for septic shock (CPP number 2015-08-03 SC). Lactate was measured in parallel on 2 samples—one capillary with the portable device (Lactate StatStrip Xpress, Nova Biomedical) and the other venous on a centrifuge tube for plasma analysis (Architect C16000 Abbott Diagnostics). We evaluated the analytical performance (coefficients of variation (CV) for repeatability and reproducibility evaluated at 2 levels of quality control (QC)—1.6 and 3.6 mM) and then the concordance between lactate levels measured by the devices and lactate levels measured by laboratory analyzer.

Results: At the QC concentrations tested, the CVs were in agreement with the limits set by the French Society of Clinical Biology—CV < 3% for repeatability and < 5% for reproducibility. An excellent correlation was observed between the 2 measurements—correlation coefficient R2 = 0.98, slope = 0.95 and ordered at the origin = 0.1. The latter suggested a low positive bias of the device not confirmed by Bland–Altmann graph analysis and graph of the differences.

Conclusion: We verified the analytical performance of the device and showed an excellent correlation with the laboratory measurement. The delocalized measure can be used in PHEM in patients with suspected sepsis syndrome. This measure should allow a more accurate and early assessment of their severity in order to improve triage and hospital orientation between ED and ICU.

Correspondence: Halter Maryline - haltermaryline@orange.fr

Annals of Intensive Care 2018, 8(Suppl 1):F-30

Introduction: We examine whether values of PaO2 taken immediately after starting Extra Corporeal Pulmonary Resuscitation (eCPR) inserted in a intensive care unit (ICU) is associated with death in patients treated by eCPR after refractory Out Hospital Cardiac Arrest (OHCA).

Patients and methods: We performed a retrospective observational study. Patients were divided into 3 groups defined on oxygen partial arterial pressure (PaO2) on the first arterial blood gas analysis after starting eCPR. Hyperoxia was defined as PaO2 greater than 300 mmHg + hypoxia as a PaO2 of less than 60 mmHg and normoxia, not classified as hyperoxia or hypoxia. The main outcome was mortality at day 28 (D28).

Results: Sixty-six consecutive patients, 77% male, with a mean age of 51 ± 14 years, were admitted to the ICU after refractory OHCA requiring eCPR. Aetiologies of OHCA were mainly due to acute coronary syndrome (67%), hypertrophic cardiomyopathy (8%) and cardiotoxic overdose (8%). Mortality at D28 reached 61%. Mean PaO2 at admission was 227 ± 124 mmHg. We observed an association between D28 mortality and PaO2—OR 1.010 [1.005–1.015], (p = 0.0005). Area Under Curve (AUC) for PaO2 after starting eCPR was 0.77 [0.65–0.89]. In a model controlling for potential confounders (age, no flow and lo flow duration, lactate level, pH and carbon dioxide partial pressure), hyperoxia exposure had an odds ratio for D28 mortality of 3.92 (95% CI 1.14–16.50) (p = 0.04) whereas in normoxia had an odds ratio for D28 mortality of 0.25 (95% CI 0.06–0.91) (p = 0.04).

Conclusion: There is an association between mortality at D28 and hyperoxia in patients admitted in ICU for refractory OHCA requiring eCPR. These data underline the potential toxicity of high dose of oxygen and suggest that control of oxygen administration in such patients is an important part of the treatment. A value of PaO2 between 100 and 300 mmHg after starting eCPR seems to be a target during treatment of OHCA treated by eCPR.

Correspondence: Lafon Thomas - thomas.lafon@chu-limoges.fr

Annals of Intensive Care 2018, 8(Suppl 1):F-31

Introduction: Sepsis has been defined as a dysregulated host response to infection leading to life-threatening organ dysfunction (Singer M et al., JAMA 2016). A qSOFA score relying on 3 simple clinical criteria (respiratory rate, mental status and systolic blood pressure) has been proposed to better identify septic patients with associated higher mortality outside the intensive care unit (Seymour CW et al., JAMA 2016). The study aim was to evaluate the ability of qSOFA to predict the development of organ failure and increased 28-day mortality in patients admitted for suspected sepsis in the Emergency Department (ED).

Patients and methods: Prospective study conducted over a period of 6 months comparing the prevalence of organ failure and 28-day mortality according to the value of qSOFA at admission to the ED between group A (qSOFA > = 2) and group B (qSOFA < 2). As part of routine care, an electronic sepsis form was specifically created to identify prospectively and exhaustively all eligible patients on-line. For the purpose of the study, sepsis diagnosis was independently validated off-line by an adjudication committee which included three physicians who reviewed clinical, biological and microbiological data. For each patient, demographic data, source of infection, qSOFA and SOFA score, biological data and 28-day mortality were recorded.

Conclusion: The present study confirmed that the qSOFA score is a reliable and practical tool to predict the development of organ failure and higher 28-day mortality in patients with suspected sepsis in the ED.

Correspondence: Ramin Séverin - severin.ramin@gmail.com

Annals of Intensive Care 2018, 8(Suppl 1):F-32

Introduction: Intravascular contrast extravasation (ICE) on CT imaging is commonly considered as the marker of pelvic active arterial bleeding, which justifies immediate performing of an pelvic angiography and trans-arterial angioembolization (TAE). However, ICEs lack of specificity in detecting arterial injuries especially in cases of bones or venous bleeding. The objective of present study was to determine the predictive performance of different ICEs characteristics for the need for Trans-arterial embolization (TAE) and the risk factors for false positive for arterial bleeding.

Patients and methods: A retrospective study was performed in our trauma center between 2010 and 2015. All severe trauma patients with pelvic fracture were included. Pelvic ICE characteristics on CT scan were specifically studied—arterial (aS2ICE), portal surface (pS2ICE) and extension (exS2ICE), anatomical relationships (ICE contact with complex bone fractures, direct relation between ICE and large retroperitoneal Hematoma. The overall predictive performance of ICE surfaces for pelvic TAE use was analyzed thanks ROC curves. The analysis then focused on the study of risk factors for false positives.

Results: Among 311 severe trauma patients with pelvic ring fracture (mean age 42 ± 19, mean ISS 27 ± 19), 94 (30%) had at least one pelvic ICE on initial CT scan. Patients requiring pelvic TAE had aS2ICE and pS2ICE significantly larger than others (p = 0.001 and p = 0.035, respectively). The global ability of ICE surfaces was either modest (aS2ICE AUC 0.76 [CI 95% 0.64–0.90] + p = 0.011) or non-significant (pS2ICE and exS2ICE) to predict pelvic TAE. The high sensitivity threshold was defined as an aS2ICE greater than 20 mm². Using this threshold, 76% of patients were false-positives. Risk factors of false-positives were—admission low systolic blood pressure (OR 6.7 [CI 95% 1.2–36.7]) and low transfusion needs (OR 15.0 [CI 95% 1.7–133.6]), the extravasation in contact with complex bone fracture (OR 7.8 [CI 95% 1.4–59.7]) or the absence of direct relation between the extravasation and large retroperitoneal hematoma (OR 7.6 [CI 95% 1.4–42.3]).

Conclusion: A significant pelvic ICE during arterial phase does not guarantee the need for pelvic TAE. Three-quarter of patients with an aS2ICE greater than 20 mm² did not experienced pelvic TAE. TAE may be performed in cases of direct relation between ICE and large retroperitoneal hematoma or an ICE without contact with complex bone fracture.

Correspondence: Jahandiez Vincent - vincent.jahandiez01@chu-lyon.fr

Annals of Intensive Care 2018, 8(Suppl 1):F-33

Introduction: Opening of the mitochondrial permeability transition pore (PTP), triggered by Cyclophilin-D (CypD) binding under stress conditions, plays a key role in ischemia–reperfusion injury. We sought to determine, using transgenic mice, whether CypD deletion (CypD −) would improve resuscitability and survival after experimental cardiac arrest (CA). Additionally, we compared the protective effects of CypD deficiency with that of targeted temperature management (TTM).

Patients and methods: Anesthetized mice underwent a 5 min asphyxial CA followed by resuscitation (cardiac massage, resumption of ventilation, epinephrine). Four groups of animals were studied—Sham, Control (Ctrl), CypD-CA using mice lacking CypD (knockout mice), and TTM-CA with fast hypothermia induced by external cooling at reperfusion (33 °C for 1 h). Two hours after CA, the following measurements were carried out (n = 6–9 group)—echocardiography, cellular damage markers (including S100B protein and troponin Ic) and mPTP opening in mitochondria isolated from brain and heart. Additional mice (n = 7–12 group) were included in the same 4 groups for survival follow-up (24 h and 7 days).

Results: Characteristics of CA were similar among groups. Rate of restoration of spontaneous circulation (ROSC) was significantly higher in CypD—and TTM groups compared to controls (p < 0.05). Time to ROSC was shorter in CypD—versus TTM and Ctrl (p < 0.05). Genetic loss of CypD and TTM prevented to a similar extent CA-induced myocardial dysfunction, increase in blood levels of both S100B protein and troponin Ic (p < 0.05 versus Ctrl). CA resulted in a significant increase in PTP opening only in mitochondria isolated from brain (p < 0.05 versus Sham). CypD deletion as well as TTM limited CA-induced PTP opening in brain (p < 0.05 versus Ctrl). Short-term survival (24 h) was significantly improved in the CypD—and TTM groups when compared to controls (p < 0.05). However, only therapeutic hypothermia improved survival at day 7 (p < 0.05 versus Ctrl).

Conclusion: In our murine CA model, genetic loss of CypD increased resuscitability and short-term survival but, unlike therapeutic hypothermia, failed to improve 7-day survival.

Correspondence: Pugin Deborah - Deborah.pugin@hcuge.ch

Annals of Intensive Care 2018, 8(Suppl 1):F-34

Introduction: Early prediction of neurological outcome of post-anoxic comatose patients after cardiac arrest (CA) is challenging. Prognosis of comatose patient relies on multimodal testing—clinical examination, electrophysiological testing and structural neuroimaging (mainly diffusion MRI). This prognostication is accurate for predicting poor outcome (i.e. death) but not sensitive for identifying patients with good outcome (i.e. consciousness recovery). Resting state functional MRI (rs-fMRI) is a powerful tool for mapping functional connectivity, especially in patients with low collaboration. Several studies showed that rs-fMRI can differentiate states of consciousness in chronically brain-damaged patients. A recent study also showed that functional neuroimaging can early detect signs of consciousness in patient with acute traumatic brain injury. However, rs-fMRI has not been assessed for the early prognostication of post-anoxic comatose patient.

Patients and methods: We assessed whole-brain function connectivity (FC) of 17 post-anoxic comatose patients early after CA using rs-fMRI. Nine patients ultimately recovered consciousness (good outcome) while eight died (poor outcome). We estimated FC for each patient following a procedure previously described. We statistically compared whole-brain FC between good and poor outcome group, to assess which brain connections differed between them. Then, we trained a machine-learning classifier (a Support Vector Machine, SVM) to automatically predict coma outcome (good poor) based on whole-brain FC of comatose patients. Finally, we compared this outcome prognostication based on functional MRI to those using standard structural diffusion MRI.

Results: Good and poor coma outcome groups were similar in terms of demographics, except for time to ROSC. Good outcome group showed significant increase in whole-brain FC between most cortical brain regions + with the strongest changes occurring within and between occipital and parietal, temporal and frontal regions (Fig. 1). Using whole-brain FC and a SVM classifier to predict coma outcome yielded to an overall prediction accuracy of 94.4%(AUC 0.94). Interestingly, automatic outcome prognostication using functional neuroimaging achieved better results that structural neuroimaging methods like DWI (accuracy 70.6%).

Conclusion: We used rs-fMRI to predict coma outcome in a cohort of post-anoxic comatose patients early after CA. We deliberately chose to include only patients with indeterminate prognosis after standard multimodal testing, to assess the contribution of rs-fMRI in the early prognostication of coma outcome. We found that automatic prediction based on functional neuroimaging yielded much better results than current DWI methods, notably for identifying patients who recovered consciousness.

Correspondence: Miroz John-Paul - John-Paul.Miroz@chuv.ch

Annals of Intensive Care 2018, 8(Suppl 1):F-35

Introduction: ECMO is increasingly used in critical care. Neurologic complications are frequent however there are no quantitative tools for outcome prediction in this setting. The aim of the study is to examine the value of quantitative pupillometry to predict neurologic outcome in patients on veno-arterial ECMO (VA-ECMO).

Patients and methods: Preliminary cohort analysis (June 2016-August 2017) of consecutive patients with VA-ECMO, in whom Npi^®-200 pupillometer (Neuroptics^®, USA) was repeatedly performed (12–72 h) to measure  % pupillary light response (qPLR) and Neurologic Pupil index (NPi). Mortality was assessed at 3 months. Using simplified 4-channel EEG (Masimo SedLine^®, USA) depth of sedation (using PSI—Patient State Index and SR— % suppression ratio on EEG) was recorded in parallel.

Results: A total of 28 patients (mean age 56 years [23–81] + mean APACHE II 28 [12–47], mean SOFA 14 [9–18]) were analyzed—14 patients (50%) had refractory cardiogenic shock and 14 had refractory cardiac arrest. Mortality at 3 months was 78%. NPi < 3 and qPLR < 11% were associated with 100% mortality. NPi = 0 (n = 7) was associated with highly suppressed EEG (SR 64–100%) and early death within 96 h. Among survivors (n = 5) NPi was > 4 on all measurements.

Conclusion: Automated infrared pupillometry, and particularly the quantitative Neurological Pupil index, may be useful for early neuroprognostication in VA-ECMO patients. Further study is undergoing at our center to confirm these preliminary data in a larger cohort.

Correspondence: Weiss Emmanuel - manuweiss@yahoo.fr

Annals of Intensive Care 2018, 8(Suppl 1):F-36

Introduction: Hospital-acquired (HAP) and ventilator-associated pneumonia (VAP) are often selected for randomized clinical trials (RCTs) aiming at new drug approval. A recent systematic review (Weiss et al. Crit Care 2017) reported a significant heterogeneity in endpoints used in RCTs comparing treatment of severe pneumonia that may influence their ability to demonstrate differences between studied drugs. Clinical cure was the most frequently used endpoint but its definition was highly variable. These results are not surprising as far as even guidance from regulatory agencies on how to evaluate HAP VAP treatments differ. The aim of this work was to reach a consensus on the most appropriate endpoint to consider in future clinical trials evaluating the efficacy of antimicrobial treatment for HAP VAP, using Delphi method.

Patients and methods: Twenty-six international experts from intensive care, infectious disease and from the industry were consulted using Delphi method (four successive questionnaires) from January 2016 to January 2017. More than 70% of similar answers to a question were necessary to reach a consensus.

Results: According to 60% the experts, clinical cure was the most desirable primary outcome among those found in the literature but two other endpoints were highly rated—all-cause mortality and mechanical ventilation (MV)-free days. Consequently, 88% of the panelists agreed to use a composite endpoints and even a hierarchical composite endpoint to combine these items together in which clinical cure and MV-free days would be assessed at day 28 and clinical cure at day 7 after end of therapy. For VAP, mortality was considered as the most clinically significant item by 75% of the experts, followed by MV-free days and finally clinical cure (Fig. 1). For HAP, a dual composite endpoint that only included all-cause mortality and clinical cure was chosen (Fig. 1). Among the various elements of clinical cure definition found in the literature, only three were retained by the experts—resolution at end of therapy of signs and symptoms present at enrolment, no further antimicrobial treatment needed and resolution or lack of progression of radiological signs of pneumonia. Finally, we found a consensus on the signs and symptoms that should trigger the suspicion of pneumonia—worsening of gaz exchange, purulent tracheal secretions, hypotension and or vasopressor requirements and fever or hypothermia.

Conclusion: We provide here two consensual endpoints (for VAP and HAP) that would help addressing the efficacy of antimicrobial molecules for HAP VAP treatment in future clinical trials.

Correspondence: Merceron Sybille - sybille07@gmail.com

Annals of Intensive Care 2018, 8(Suppl 1):F-37

Introduction: Stenotrophomonas maltophilia (SM) is a Gram negative bacteria rarely involved in ICU acquired infections. Empirical treatment is challenging, as SM is resistant for almost all clinically-relevant antibiotics. SM ventilator-associated pneumonia (SM-VAP) have been associated with frequent inadequate antimicrobial therapy and increased ICU stay and mortality. Identifying risk factors for SM-VAP could be useful.

Patients and methods: We conducted a retrospective analysis of prospectively collected data in a multicenter database in France. VAP was determined as microbiologically documented pneumonia occurring after at least two days of mechanical ventilation. All patients admitted from 2000 to 2015 presenting at least one episode of VAP were included. SM-VAP were compared with VAP due to other micro-organisms (VAP-other). To assess risk factors and day-30 (D30) mortality, SM-VAP were matched (1 − n) with VAP-other based on center, time from ICU admission to VAP occurrence, and SAPS II score.

Results: Of the 20 968 patients included, 1 659 presented at least one episode of VAP. Among them, 102 (6.2%) suffered from SM-VAP. Patients with SM-VAP were men (61.3%), aged 62 [53–76] with SAPS II score 50 [40–65] and 20% were immunocompromised. Median length of stay in ICU was longer in SM-VAP (Table). 93 SM-VAP were matched with 380 control patients. In univariate analysis, risk factors for SM-VAP were—male gender, chronic heart failure, respiratory, cardiovascular and coagulation SOFA scores two days before VAP, median number of antibiotics used, percentage of time with antibiotics before VAP, parenteral nutrition, dialysis, catecholamine use and exposure to Ureido-carboxypenicillin, Ciprofloxacin, Tazobactam or Imipenem-Meropenem during the week before VAP (Table). Patients with SM-VAP were less likely to receive initial adequate therapy (56 vs 70%, OR 0.53, p = 0.01). There was no statistical difference for ICU or D30 mortality. D60 mortality was higher for SM-VAP (Table). In multivariate analysis, exposure to Imipenem-Meropenem during the week before VAP, respiratory and coagulation SOFA scores two days before VAP were independent risk factors for SM-VAP.

• SAPSII: Simplified Acute Physiology Score; SOFA: SOFA (Sequential Organ Failure Assessment); SOFA Resp: SOFA respiratory score;

• SOFA coag: SOFA coagulation score; SOFA cardio: SOFA cardiovascular score

Conclusion: SM-VAP represented 6.2% of VAP. We observed no differences in patients characteristics between the groups. Imipenem-Meropenem use during the week before VAP was the most important risk factor for SM-VAP. The higher risk of inadequate initial therapy with SM-VAP had no impact on D30 mortality but D60 mortality was significantly higher.

Correspondence: Lehingue Samuel - samlehingue@gmail.com

Annals of Intensive Care 2018, 8(Suppl 1):F-38

Introduction: Education of undergraduate students is key to improve Hand hygiene (HH) behavioral changes amongst doctors [1.2]. Our aim was to evaluate personal feedback using Ultraviolet (UV) light inspection cabinets in a 2 years program. Our hypothesis was that its use for Alcohol hand rub (AHR) application on first year would increase complete AHR application on 2nd year.

Patients and methods: This was a simple blind randomized trial comparing HH training with personal feedback using UV cabinet to a control group. On first year, students had access to a theoretical formation then were convened by groups for a demonstration of the correct execution of world health organization’s (WHO) procedure [3]. Before HH training, each group underwent a cluster randomization. In the control group, the student hand rubbed under visual supervision and advises of a trainer. In the intervention group after the same visual assessment, completeness of AHR hand application was recorded under UV light and shown to the student. He was given free access to the UV cabinet to repeat the technique, until perfect application complete under UV light. An enhancement with a scenario-based learning was proposed to both groups. On second year, every student were asked to hand rub with the fluorescent AHR. A supervisor blinded to the group of randomization assessed the quality of the HH procedure visually, the completeness of hand application under UV light and compliance with the WHO’s opportunities for HH during the simulation.

Results: After randomization 140 students were included in the intervention group and 102 in the control group. On second year, the rate of complete application of the AHR under UV was increased in the intervention group as compared with the control group (60% versus 30.4% p < 0.001) (Fig. 1) despite that visual assessment of HH procedures was similar between the two groups. In a logistic regression model including gender, intercurrent HH formation, intercurrent UV cabinet use, surgical unit traineeship and report of regular use of AHR, the hazard ratio for the intervention was 3.837 (IC 2.086–7.058). The rate of perfect compliance with the HH opportunities in the intervention group was increased (58.1% versus 42.4% p < 0.018) and the effect persisted in the logistic regression.

Conclusion: UV cabinets for undergraduate students’ HH education improve the technique and the compliance with HH opportunities. Included in a multifaceted education program, it must be considered a key tool for training.

Correspondence: Missri Louai - louai.missri@gmail.com

Annals of Intensive Care 2018, 8(Suppl 1):F-39

Introduction: Bacterial colonization of environmental surfaces or medical devices can lead to nosocomial infections, especially in ICU. Mobile phones are commonly used by healthcare workers. However, very few studies assessed bacterial colonization of healthcare workers’ mobile phones in ICU and the impact of disinfection.

Patients and methods: We conducted a prospective monocentric study in a 15-bed medical ICU with 54 healthcare workers. Swabs were used to assess bacterial colonization of healthcare workers’ mobile phones before and after 5 min of disinfection with bactericidal Anios© wipes. Results were compared to those observed in administrative staff. Swabs were incubated at 37 °C for 24 h on different types of agar. Bacterial identification was conducted by mass spectrometry (MALDI-TOF) combined with the VITEK2 automated antibiotic device for the search for antibiotic-resistant strains.

Conclusion: Colonization of mobile phones is similar in healthcare workers in ICU and in administrative staff. In our ICU, all mobile phones are colonized by skin flora and 43% with pathogens. Our assessed disinfection protocol appeared insufficient. Intensivists should be aware of these results to help prevent nosocomial infections.

Correspondence: Prevel Renaud - renaud.prevel@hotmail.fr

Annals of Intensive Care 2018, 8(Suppl 1):F-40

Introduction: Extended-spectrum beta-lactamases-producing Enterobacteriaceae (ESBL-E) are disseminating worldwide. ESBL-E fecal carriage is known to increase the risk of ESBL-E infections and so untailored empirical probabilistic antibiotherapy and mortality. The aim of this work was to study the mechanisms of dissemination of ESBL-E fecal carriage in an intensive care unit (ICU).

Patients and methods: We screened every in-patient at admission and every 7 days in a two units of medical ICU (25 beds) for 6 months according to current practice. Every ESBL-E found was characterized by ESBL genes PCR amplification and sequencing and the clonal dissemination was assessed by Pulsed-Field Gel Electrophoresis (PFGE).

Results: Over that 6-month period, among the 608 screened patients, 57 (9.4%) were found to be colonized by ESBL-E. Among these 57 patients, 7 (1.2%) acquired the ESBL-E fecal carriage during their stay in the ICU. No cross-transmission of ESBL-E occurred (as assessed by PFGE analysis) even for ICU-acquired fecal carriage. ESBL genes were found to be mostly from the CTX-M group mainly CTX-M group 1 (36 57) enhancing the community origin of the ESBL-E fecal carriage even in ICU. ESBL-E caused infections occurred in 5 57 patients always consisting in ventilator-associated pneumonia caused by the colonizing strain.

Conclusion: Hygiene isolation procedures seem to be well applied in our units as we do not have any ESBL-E fecal carriage cross transmission case. But still cases of ICU-acquired ESBL-E fecal carriage occurred suggesting an endogenous origin in a non-outbreak situation. Therefore, the mechanisms of ESBL-E expansion in the community and of endogenous ESBL-E acquisition in ICU are still to investigate.

Correspondence: Perrin Caroline - perrincaroline@hotmail.fr

Annals of Intensive Care 2018, 8(Suppl 1):F-41

Introduction: Detection of nosocomial infections (NIs) occurring during extracorporeal membrane oxygenation (ECMO) is difficult and their related side effects appear to contribute to the high mortality rates (40–50%) of these patients. The most common ones are ventilator-associated pneumonia and bloodstream infections (BSI). Thus we empirically developed a standardized protocol with systematic blood cultures on days 2, 6, 10 and 15. The objectives of this study were to analyze BSI incidence, distribution of microorganisms, risk factors in adult undergoing ECMO support and to evaluate the efficiency of our protocol.

Patients and methods: We prospectively included all adult patients who underwent veinoarterial (VA-ECMO) or veinovenous extracorporeal membrane oxygenation (VV-ECMO) between May 2014 and March 2016. Anamnestic data and blood culture results using the protocol and additional blood samples were analyzed.

Results: Among the 81 patients who underwent ECMO support for more than 48 h, 10 patients in the VA-ECMO group (n = 10 56 + 17.9%) and 3 patients in the VV-ECMO group (n = 3 25 + 12%) developed 18 BSI, corresponding to a global rate of 22.4 BSI per 1000 ECMO days. Microorganisms associated with these infections were most frequently gram-negative bacilli. Only the duration of ECMO was significantly correlated with BSI occurrence, with a median duration of 6.5 days for “patients without BSI” and of 14 days for “patients with BSI” (p = 0.0170). The efficiency of the protocol in diagnosing BSI was roughly 5% (5 96) for the VA-ECMO group and around 6% (3 49) for the VV-ECMO group.

Conclusion: Probability of developing BSI in adults undergoing ECMO increases with ECMO duration. Systematic blood cultures do not appear to be an effective way of improving BSI detection.

Correspondence: Perrin Caroline - perrin.ca@chu-toulouse.fr

Annals of Intensive Care 2018, 8(Suppl 1):F-42

Introduction: Nosocomial infections (NIs) occurrence during veno-arterial extracorporeal membrane oxygenation (VA-ECMO) support and their complications appear to contribute to the high mortality rates (40–60%) of these patients. The most common ones are ventilator-associated pneumonia and bloodstream infections (BSI). In clinical practice, detection of NIs is difficult due to confounding factors caused by the device. The objectives of our study were to analyze BSI and positive tracheal aspirate (TA) culture incidences, distribution of microorganisms and risk factors in adult undergoing veinoarterial ECMO (VA-ECMO).

Patients and methods: We prospectively and consecutively included all adult patients who underwent VA-ECMO between May 2014 and March 2016 in our intensive care unit. Anamnestic data, blood culture and TA culture were analyzed. Primary endpoint was the occurrence of a BSI or a positive TA culture more than 24 h after ECMO initiation and within 48 h after ECMO discontinuation.

Conclusion: Use of an antibioprophylaxis at ECMO implantation is associated with a decrease of global positive cultures development.

Correspondence: Alexandre Destere - alexandre.destere@aphp.fr

Annals of Intensive Care 2018, 8(Suppl 1):F-43

Introduction: The majority of recommended drugs to treat status epilepticus (SE) patients have a limited safety margin with potential risk of drug–drug interactions. Cytochrome P450 (CYP) induction may result in insufficient dosage and subsequent therapeutic inefficiency of the patients’ co-treatments including the anticonvulsive drugs administered to treat the SE. Our objective was to describe the anticonvulsive drugs administered to SE patients and determine the resulting theoretical risk of drug–drug interactions with the patients’ treatments.

Patients and methods: Patients admitted during 1 year (Jan–Dec 2016) in our University Hospital intensive care unit (ICU) were selected using the Program of Medicalization of the Information system (PMSI). The usual demographic, clinical, and outcome data were collected with a special focus at the patients’ treatments taken before ICU admission and treatments administered during the ICU stay including the anticonvulsive drugs. We used the table of drug–drug interactions of the Geneva University Hospitals (1) and the DrugBank website (2) to determine the risks of drug–drug interactions.

Results: Sixty-one patients (74% males 26% females + age—56 years [16 + 92] (median [percentiles 25, 75]) were included in this study. Patients were chronic alcoholics (41%) and drug users (16%). The following anticonvulsive drugs were administered—levetiracetam (N = 46, 75%), fosphenytoine (N = 40, 66%), clonazepam (N = 37, 61%), clobazam (N = 28, 46%), midazolam (N = 22, 36%), valproic acid (N = 10, 16%), carbamazepine (N = 7, 12%) and phenobarbital (N = 7, 12%). Among the 46 61 patients (77%) treated with a CYP inducer, 26 patients (63%) received at least one drug with liver metabolism of which plasma concentrations might have been altered by such potential drug–drug interactions, thus representing 43% of the SE patients admitted to the ICU. Drugs at risk of pharmacokinetic alterations were cardiovascular, metabolic and psychotropic drugs.

Conclusion: Drugs with potential risk of drug–drug interactions used to treat SE patients in the ICU are relatively frequent. Such drug–drug interactions are at risk of altering the plasma concentrations of the anticonvulsive drugs as well as several other patients’ co-treatments, highlighting the usefulness of pharmaceutical advice at the bedside when choosing the anticonvulsive drugs to administer.

Correspondence: Fahmi Dachraoui - dachraoui.fahmi@gmail.com

Annals of Intensive Care 2018, 8(Suppl 1):F-44

Introduction: Delirium in the ICU is often under-diagnosed despite its related burden and impact on patients’ morbidity, mortality and prolongation of hospital length of stay. The aim of this study was to assess the medical and paramedical community beliefs and practices regarding delirium in Tunisian ICUs.

Patients and methods: Between August 1st and 31/2017, healthcare professionals working at the ICUs of university hospitals of Monastir and Mahdia (Tunisia) were asked to participate in the survey by completing a questionnaire anonymously (that specified participants’ characteristics (age, gender, function, years of experience in ICU) and their knowledge and perception of delirium in ICU. The questionnaire consisted in 10 questions of different types: Likert style (: widespread scale in psychometric questionnaires in which the respondent expresses his or her degree of agreement or disagreement with an assertion), multiple choice, ranking and yes/no).

Results: During the study period, 96 respondents out of 163 (60% female, nurses: 52%), aged between 20–30 years in 70%, responded to the questionnaire. Healthcare professionals experience in the ICU was < 1 year in 39.6%; 1–5 years in 32.3%, and > 5 years in 28.1%. Participants asserted that the “most characteristic signs of delirium” were: insomnia (25%); confusion (22%); agitation (21%) and aggressiveness (20%). Three-quarters of participants said they did not systematically search for signs of delirium in their patients. 33% thought that delirium was “an insignificant problem” or that “it was not a problem”. Only one and three participants respectively, said they attended a conference and read an article about delirium in ICU the last year. Half of the respondents felt that the most appropriate treatment for a patient with delirium was restraint. Nearly one-third of participants thought that delirium was an under-diagnosed entity and only 6% felt that it was associated with long-term neuropsychological deficits. Factors considered to be determinant in the occurrence of delirium were ARDS, shock, age, mechanical ventilation, postoperative status in 70, 66, 63, 55 and 17%, respectively.

Conclusion: Most Tunisian healthcare professionals consider delirium as a common, underdiagnosed, and serious problem in the ICU. Yet, few participants actually monitor this condition.

Correspondence: Khzouri Takoua - takoua_kh2@yahoo.fr

Annals of Intensive Care 2018, 8(Suppl 1):F-45

Introduction: Delirium is a common manifestation of acute brain dysfunction in critically ill patients. It is associated with a healthcare cost increase, and extension of the hospital stay length. The present study aimed to explore influence of patient characteristics and analgesic-sedation on delirium incidence and to analyze its risk factors.

Patients and methods: It is a prospective single blind randomized controlled trial, started on the first July 2017 in a 12-bed toxicological Intensive Care Unit, including all mechanically ventilated patients requiring sedation who were admitted for acute poisoning. They were randomly divided into two groups G1 et G2 receiving respectevily Propofol-Remifentanil and Midazolam—Remifentanil. Delirium assessment scores were judged not adapted to our population and we retained the diagnosis of delirium on arguments inspired from diagnostic and statistical manual of mental disorders fourth edition (DSM-IV).

Results: Until the 15th September 2017, 35 patients were included, with 22 patients in G1 and 13 in G2. The two groups were comparable in terms of epidemiological characteristics. Delirium was developed in 9 patients (26%) (n = 5 in G1 and n = 4 in G2) with an average duration of 18 ± 13 h [6–48] with no difference between the 2 groups (13 ± 10 h for G1–24 ± 16 h for G2, p = 0.268). Compared to those without delirium, no differences were found in the patient characteristics among these two groups with regard to sex, age, psychiatric history and severity of illness (APACHE II, IGS II score) and even with regard to hypnotic choice (5 vs 4 p = 0.599). Delirium was associated to prolonged duration of mechanical ventilation (69 ± 73 h vs 39 ± 69 h, p = 0.280) and length of ICU stay (91.51 h vs 54.05 h, p = 0.132) without significant differences. Delirious patients had more hypotension (p = 0.008), and received more atropine (0.026). Multiple logistic regression analysis identified atropine (OR 2.333, 5%Cl 0.992–5.489, p = 0.026) as an independent risk factor for delirium.

Conclusion: The diagnosis and prevention of ICU delirium are subjects of multiple ongoing investigations. We carried out this study to detect the risk factors of delirium in order to prevent it. It is important to note that our results are influenced by the studied population and are only preliminary. We rely on the study pursuit and the sample enlargement to better inform us as well on risk factors as protective.

Correspondence: Mateos François - francois.mateos@gmail.com

Annals of Intensive Care 2018, 8(Suppl 1):F-46

Introduction: Background: Severe alcohol withdrawal syndrome is a common cause of hospital admission. Delirium tremens is a potentially fatal complication of alcohol withdrawal. In severe delirium, very large dosages of benzodiazepines can be required despite well described side effects, such as coma and hypoxic cardiac arrest, although there is no recommendations for standardized treatments. Objective -The aim of this study was to describe outcomes and risk factors for complications in patients with severe alcohol withdrawal syndrome treated in intensive care unit with continous infusion of benzodiazepine (BZD).

Patients and methods: We retrospectively reviewed the medical records of all patients hospitalized for alcohol withdrawal syndrome between 2006 and 2016. Only those who received continous-infusion of BZD, associated with close clinical monitoring and the evaluation of RASS and Cushman scores, without systematic recourse to mechanical ventilation, were included.

Results: We studied 104 patients hospitalized in ICU for severe alcohol withdrawal syndrome. The mean age (SD) was 48.7 ± 8.7 years, mean ICU admission SAPS (simplified acute physiology score) II score was 20 ± 6.1. All of them have received continous infusion of midazolam, with a median maximum perfusion velocity of 8 mg h (interquartile range, (5, 12)). The median duration of treatement was 2 days (interquartile range, (1, 3)). Thirteen patients (12%) developed pneumonia, and or required intubation, and 8 (8%) have had seizures. No cardiac arrest and death was observed. ICU length of stay (LOS) was 3 days (2, 5) (median, interquartile range). Patients who requiried intubation and or developed pneumonia, received substantially more BZD (median total dose, 428 mg of midazolam vs. 147 mg in the non-complicated group + p < 0.001), and their ICU LOS was higher (median, 5 days vs. 3 days + p < 0.001). Endotracheal intubation and or development of pneumonia were associated with a higher maximum perfusion velocity of midazolam (> 8 mg h) (OR 18.83, IC 95% (2.11–168.15), p = 0.009). Previous episodes of delirium tremens before ICU admission were associated with higher complications such as mechanical ventilation and or pneumonia (OR 11.29, IC 95% (1.45–87.60), p = 0.02).

Conclusion: In severe delirium, very large dosages of benzodiazepines can be used without systematic mechanical ventilation with a low incidence of complications.

Correspondence: Smonig Roland - r.smonig@gmail.com

Annals of Intensive Care 2018, 8(Suppl 1):F-47

Introduction: Delirium is frequent in intensive care unit (ICU) patients and is associated with increased mortality, increased hospital stay, increased cost and long term cognitive impairment in survivors. Numerous pharmacological and non-pharmacological strategies have been investigated for delirium treatment without success. Therefore delirium prevention strategies are recommended by current critical care practice guidelines. Among the potentially modifiable risk factors for delirium, the impact of daylight exposure on delirium incidence and or duration has not been studied. The objective of this study was to investigate whether daylight exposition would reduce delirium burden in critically ill patients.

Patients and methods: We conducted a prospective study in a 27-bed medical intensive care unit (ICU) over a 1-year period (January 2016–January 2017). All consecutive adult patients receiving invasive mechanical ventilation (MV) for 2 days or more were eligible for the study. Patients were assigned to a room with windows allowing daylight exposure (“Light” group) or without window (“Dark” group), depending on bed availability. Delirium was evaluated with the Intensive Care Delirium Screening Checklist (ICDSC) for a maximum period of 28 days. Delirium was defined by a ICDSC score ≥ 4 for two consecutive days. Agitation was defined by a RASS > or = +2. The primary endpoint was cumulative incidence of delirium. Data are presented as median (interquartile range) or number (percentage).

Results: A total of 195 patients were included (age—50 [50 + 69] years, SAPS2—51 [36 + 64], SOFA score—9 [7 + 11], medical admission—69%). Of them, 110 patients were admitted to a “Light” group and 85 to a “Dark” group. Incidence of known risk factors for delirium was similar in the two groups. Delirium occurred in 65 (64%) patients in the “Light” group and in 55 (71%) patients in the “Dark” group (p = 0.28). The duration of delirium was 3 [1 + 7] days. Patients in the “Light” group received significantly less neuroleptics to treat agitation than patients in the “Dark” group (13 vs. 25%, p = 0.04). This protective association persisted after adjustment for confounders in multivariate analysis (Odds ratio = 0.40 + [0.17 + 0.90] + p = 0.03).

Conclusion: Daylight exposure does not impact on delirium burden in ICU mechanically ventilated patients. However, daylight exposure is independently associated with a reduced prescription of neuroleptics to treat agitation.

Correspondence: Zeidan Sinéad - sineadzeidan@gmail.com

Annals of Intensive Care 2018, 8(Suppl 1):F-48

Introduction: Patients with convulsive status epilepticus (CSE) frequently require mechanical ventilation (MV), either for general anesthesia in case of refractory generalized CSE, or for airway protection. Guidelines for the management of refractory generalized CSE currently recommend general anesthesia for 24–48 h, followed by gradual withdrawal. Our objective is to evaluate the incidence of refractory generalized CSE among patients who required MV during pre-hospital management of status epilepticus, and to describe the management of general anesthesia in intensive care unit (ICU).

Patients and methods: This ongoing multicenter retrospective observational study is conducted in 4 French ICUs. All patients admitted in ICU under mechanical ventilation between 01-01-2014 and 12-31-2016 with disease-code “Status Epilepticus” are included. Exclusion criteria are—age < 18 years, post anoxic SE, acute traumatic brain injury, initiation of MV in ICU, transfer from another ICU, inclusion in a therapeutic trial on SE, non-convulsive SE. Collected data include reason for MV, antiepileptic treatment, dosage and duration of general anesthesia, mode of EEG monitoring. Outcomes are—relapse of SE, MV duration, in-ICU length of stay and mortality.

Results: Among the 211 medical files reviewed, 77 met the inclusion criteria and were analyzed, and 134 were excluded. A minority of patients (18.2%) had a refractory generalized CSE, most patients (64.9%) had a non-refractory generalized CSE + the others had mostly partial CSE. The main reason for intubation was coma (n = 45, 58.4%). The duration of general anesthesia was not significantly different in refractory CSE patients compared to non-refractory CSE patients (p = 0.18). Data regarding main outcomes are summarized below-.

Conclusion: These preliminary data suggest that the majority of the patients admitted in ICU under MV for CSE do not have a refractory status. Indication of MV is mainly coma without persistent convulsions. The mean duration of general anesthesia before withdrawal is < 24 h, and thus in discrepancy with guidelines, but does not seem associated with a frequent relapse of SE. If this low rate of RSE for patients admitted in ICU and the safety of rapid withdrawal of GA are confirmed, the recommended 24–48 h duration of general anesthesia in ICU could be challenged.

Correspondence: Nèji Henda - hendaneji@gmail.com

Annals of Intensive Care 2018, 8(Suppl 1):F-49

Introduction: Induced coma in intensive care patients protect them against pain and neurologic disorders. However, a few of them may present a delayed wake-up when the sedation is interrupted. The aim of this work is to assess brain imaging findings in patients with this condition.

Patients and methods: Retrospective review of imaging data of 27 patients (21 males and 6 females), aged between 32 and 84 years, admitted in intensive care unit (ICU) between June 2015 and September 2017, who had sedation or general anesthesia and presented a delayed wake-up. They were explored either by MRI (n = 11) or computed tomography (CT) (n = 16). Patients with traumatic lesions were excluded.

Results: Patients were admitted in the ICU because of chronic obstructive pulmonary disease exacerbation (n = 7), infectious pneumonia or pleural effusion (n = 6), acute respiratory failure (n = 5), heart disease (n = 2). Two patients underwent general anesthesia. Septic shock and circulatory collapse occurred in 3 and 5 patients respectively. MRI and CT showed lesions that may explain the wake-up delay in 6 of 11 and 4 of 16 patients, respectively. Brain anomalies included anoxic lesions (n = 4) with basal ganglia involvement (n = 3), ischemic or hemorrhagic strokes (n = 4), hepatic encephalopathy (n = 1) and herpetic encephalitis (n = 1).

Conclusion: Brain imaging techniques help diagnosing causes of delayed wake-up after induced coma. Anoxic lesions and strokes are mostly behind this condition. MRI is more accurate than CT.

Correspondence: Abily Julien - julien.abily@chu-rouen.fr

Annals of Intensive Care 2018, 8(Suppl 1):F-50

Introduction: Gastric tubes are common in intensive care units used for enteral feeding, administration of drugs or aspiration of the digestive tract. These tubes offer an excellent tolerance but malposition may have serious consequences that can lead to patient’s death. The actualy gold method to confirm their correct placement is chest X-ray. We report a study which evaluate the performance of gastric ultrasonography for the validation of the good positioning of the gastric tube.

Patients and methods: We carried out a prospective, monocentric study in a Medical Intensive Care Units. For each included patient, we compared the results of a gastric ultrasonography to the interpretation of a chest X-ray.

Results: One hundred and thirteen gastric ultrasonographies were performed from July 2016 to May 2017. In 87 cases, ultrasonography concluded that the gastric tube was correctly positioned, confirmed by chest X-ray. In 24 cases, ultrasonography did not visualize the tube in gastric area. Among these 24 cases, only 4 malpositions were detected by the chest X-ray. The sensitivity and specificity of gastric ultrasonography were 0.81 [0.72 + 0.87] and 1 [0.51 + 1]. Positive and negative predictive values were 1 and 0.17, respectively. The ultrasonography was performed 54 min [50.3 + 57.6] after the gastric tube placement while the chest X-ray was interpreted 211 min [170.9 + 251.7] after this same placement (p < 0.0001).

Conclusion: Our results suggest a good performance of gastric ultrasonography to check the positioning of the gastric tube. This result must be interpreted with caution because of a low power of the study. We planned a multi-center study to confirm our results.

Correspondence: Giabicani Mikhael - mikhael.giabicani@aphp.Fr

Annals of Intensive Care 2018, 8(Suppl 1):F-51

Introduction: Prognosis of cirrhotic patients hospitalized in intensive care unit (ICU) remains poor. In many ICUs, cirrhotic patients are widely admitted and revalued after receiving optimal treatments. Little is known about risk factors involved in the evaluation of the prognosis at day 3, except the persistence of organ failure. This susceptibility to organ failure would be related to an alteration of the regulation mechanisms of the systemic inflammatory response. The blood neutrophil-to-lymphocyte ratio (NLR) is an inflammation biomarker reported to predict clinical outcome in unselected critically ill patients and in patients with stable liver cirrhosis, but has never been studied in critically ill cirrhotic patients. The aim of this study was to evaluate the blood NLR as parameter to predict mortality of cirrhotic patients hospitalized > 3 days in ICU.

Patients and methods: Retrospective monocentric study including consecutively cirrhotic patients hospitalized in a medical ICU from 2010 to 2016. For each patient, clinical and biological data at admission and day 3 were collected. NLR at admission (“NLRD0”), at day 3 (“NLRD3”) and its variation between admission and D3 (“delta NLR”) were calculated. Statistical analysis used appropriate non parametric tests and Cox regression for survival analysis. The ability of the variables to discriminate survivors from non-survivors was determined using ROC curves and a Net Reclassification Index (NRI).

Results: 140 patients (median Child–Pugh score = 9 [7–11], median MELD score = 25 [19–30]) were hospitalized more than 3 days in ICU. The major causes for ICU admission were sepsis (56.4%), gastrointestinal bleeding (15%) or respiratory failure (6.4%). Patients were followed up for 38.5d [13–339]. 74 (53%) patients died—38 (27%) in ICU, 21 (15%) after ICU discharge and 15 (11%) after hospital discharge. In univariate analysis, factors significantly associated with mortality were—at D3, NLR, MELD and SOFA scores + and between D0 and D3—delta NLR, delta SOFA and delta MELD. Predictors of death in multivariate analysis are shown in Table 1. Area under delta NLR ROC curve was 0.74 (CI = 0.69–0.79). NRI revealed that delta NLR was more efficient than delta SOFA (NRI = 8.7%) to identify patients with a 85% mortality risk at least.

Conclusion: NLR is a novel inflammation index known to predict poor clinical outcomes. Delta NLR is an independent predictor of mortality in critically ill cirrhotic patients and could be more effective than delta SOFA in predicting hospital mortality in these patients.

Correspondence: Ichai Philippe - philippe.ichai@aphp.fr

Annals of Intensive Care 2018, 8(Suppl 1):F-52

Introduction: Severe acute liver injury and failure (sALI ALF) is a grave complication of exertional heatstroke (EH). Liver transplantation (LT) may be a therapeutic option, but the criteria for, and timing of, transplantation have not been clearly established. The aim of this study was to define the profile of patients who require transplantation in this context.

Patients and methods: This was a multicentre, retrospective study of patients admitted with a diagnosis of exertional heatstroke-related sALI ALF with a prothrombin time (PT) lower than 50%, with or without hepatic encephalopathy.

Results: 24 male patients (median age—27.5 years) with ALI ALF related to exertional heatstroke were studied + nine of them (37.5%) were listed for emergency LT. The latter differed from those who were not listed with respect to their more severe liver failure after D1, a clear deterioration in their PT and ALT values between D0 and D3, and more marked organ dysfunction. Four of these nine patients were subsequently transplanted. At the time of LT, all had PT levels lower than 10%, a marked rise in bilirubin levels and required support for at least one organ (or x organs were involved). Histological findings on the explanted livers demonstrated massive or sub-massive necrosis and little potential for effective mitosis with a mitonecrotic appearance. The 15 unlisted patients (62.5%) were still alive 6 months later and had not experienced any after-effects.

Conclusion: Survival without liver transplantation in patients with heatstroke-related ALI ALF reaches 83.5%. The indication for liver transplantation is based on an evolving dynamic. The lack of any signs of an improvement in liver function at or after D3, in patients presenting with other organ dysfunctions or failure, means that liver transplantation should be envisaged. The peculiar histological features observed on all the explanted livers, and the aspect of abortive mitoses in hepatocytes could be attributed to the effects of heatstroke.

Correspondence: Roubin Johanna - roubinjohanna@gmail.com

Annals of Intensive Care 2018, 8(Suppl 1):F-53

Introduction: Without any treatment, patients with fulminant hepatitis die in 70 to 90% of cases. The Prometheus system allows to purify toxin bounded to albumin through adorption. Up to now, there a little data concerning its use during fulminant hepatitis. The aim of this study was to evaluate the Prometheus system on mortality and the need for liver transplant among patients with fulminant hepatitis.

Patients and methods: We conducted a retrospective observational study in a single polyvalent ICU of a teaching hospital. From December 2011 to December 2015 every patient admitted for fulminant hepatitis was included. We defined two groups—Prometheus group versus standard medical care group. Exclusion criteria were the contraindications for liver transplant.

Results: 45 patients were admitted for fulminant hepatitis during the study period, 21 patients were excluded. 12 patients had Prometheus and 12 had standard medical treatment. The average age was 40 (18–61), the mean SAPSII was 42 (15–99). On admission, the mean PT was 18.5% (8–42), the mean total bilirubin was 157 umol l. Paracetamol poisoning was the principal etiology with 45% of the patients—66% in the Prometheus group versus 25% in the standard group (p = 0.106). The hepatic encephalopathy grade was significantly higher in the Prometheus group—3 versus 0.5 in the standard group (p = 0.029). There was no difference between the two groups concerning mortality on day 28 (p = 1) or day 90 (p = 0.59). There was no difference concerning the length of stay in intensive care unit or in hospital between the two groups. 13 patients (37.1%) were transplanted. There was a statistical difference between the two groups concerning liver transplantation (p = 0.041)—4 transplant (40%) in the Prometheus group versus 9 transplant (75%) in the standard medical care group. There was a significant improve of encephalopathy after the Prometheus session (p = 0.002).

Discussion: Our study does not show any difference on mortality in fulminant hepatitis by using Prometheus. We found a statistical difference concerning the need for transplant which is significantly diminished after Prometheus. Prometheus may be a bridge to recovery. It is difficult to conclude because our study is retrospective and with only 24 patients. Also, there is a trend to have a majority on paracetamol poisoning in the Prometheus group.

Conclusion: Prometheus does not change the mortality in fulminant hepatitis. Prometheus may be a bridge to recovery with less need for liver transplantation.

Correspondence: Monet Clément - clemen.t@hotmail.fr

Annals of Intensive Care 2018, 8(Suppl 1):F-54

Introduction: The Molecular Adsorbent Recirculating System (MARS^®) is a device based on an albumin-enriched dialysis, allowing removal of albumin-bound toxins that accumulate during acute liver failure. Data on MARS^® are scarce and little is known on efficiency and tolerance. MARS^® therapy has been used in our ICU since 2005. We report here our experience with a focus on efficacy as well as tolerance.

Patients and methods: From March 2005 to August 2017 all patients who have undergone MARS^® therapy in our ICU were included consecutively and prospectively in the cohort. MARS^® therapy performed using a double lumen dialysis catheter in the femoral or jugular vein. We used the monitor MARS^® 1 TC (Teraklin) coupled with the dialysis machine Prismaflex^® (Gambro). The albumin dialysate circuit consisted of 500 ml of 20% human albumin and was regenerated by an anion-exchange column and an uncoated charcoal column (diaMARS^® IE250, diaMARS^® AC250).

Results: Ninety patients were included for 300 sessions. The mean duration was 7 h 38 min (± 1 h 43 min). The population treated consisted of 5 groups—acute-on-chronic liver failure (AoCLF), acute liver failure (ALF), post-surgery liver failure (post transplantation, post hepatectomy), refractory pruritus and drug intoxication (Fig. 1). Regarding biological efficacy—total bilirubin was lowered in AoCLF and post-surgery groups (p < .001), also in the ALF group although not significatively. MELD score was lowered in the AoCLF and ALF group (p < .001). However clinical variables (Glasgow score and encephalopathy) didn’t improve significatively. In the refractory pruritus group, pruritus decreased in 20 out of 25 patients (p < .001). Bile acid levels decreased to 34.8% of its mean baseline level (p < .001). In the drug intoxication group improvement of the Richmond Agitation-Sedation Scale (RASS) from deeply sedated (RASS < = − 3) to minimal sedation (RASS > = − 2) was obtained in 5 out of 6 patients. Out of 300 sessions, catheter-related adverse effects were low (1.3%), thrombocytopenia was the main adverse effect (26.7%).

Conclusion: We report our MARS^® experience with the largest cohort of patients referred from a single hospital. We showed biological efficacy in all 5 indications, although clinical efficacy was uneven. MARS^® therapy in patients with refractory pruritus yielded promising results. Tolerance was good and the main adverse effect was thrombocytopenia. Global transplantation-free survival was low in patients with liver failure, reinforcing the need for a liver transplantation center when using MARS^®.

Correspondence: Margetis Dimitri - dimargetis@free.fr

Annals of Intensive Care 2018, 8(Suppl 1):F-55

Introduction: Postoperative low cardiac output syndrome occurs in 4–26% after cardiac surgery. In case of refractory cardiogenic shock, veno-arterial extra corporeal membrane oxygenation (VA-ECMO) is used when pharmacological support fails to restore adequate perfusion. There is little evidence about severe digestive complications in this specific population. The aim this study was to investigate the incidence and the risk factors of gastrointestinal bleeding and mesenteric ischemia in patients assisted by VA-ECMO after cardiac surgery.

Patients and methods: Retrospective, observational study in a university hospital cardiac surgery department. All VA ECMO-assisted patients admitted in intensive care unit after cardiac surgery in a two-year period were included. Our study has been approved by ethic committee CPP Ile de France 5.

Results: At all, 160 patients have been included. Digestive complications occurred in 53 patients (33.1%). We observed 30 (18.7%) gastrointestinal bleeding. No risk factor has been identified. Bleeding was associated with prolonged assistance by VA-ECMO, mechanical ventilation, hospital length of stay, use of vasoactive drugs and transfusion. Mesenteric ischemia occurred in 32 patients (20%). In multivariate analysis, dialysis was the only significant risk factor, OR 6.35 [IC95 1.76–22.9] (p = 0.005). Heart transplant seems to be protective, OR 0.11 + [IC95 0.01–0.90 (p = 0.04). Attributed mortality increases consistently from 43 to 84.4% (p < 0.0001).

Conclusion: Severe digestive complications occur in 1 3 of VA-ECMO assisted-patients after cardiac surgery. The mortality increases by 95% in this population. Renal replacement therapy is identified as risk factor of ischemic complications.

Correspondence: Messika Jonathan - jonathan.messika@aphp.fr

Annals of Intensive Care 2018, 8(Suppl 1):F-56

Introduction: Colonoscopy is crucial for the management of lower gastro-intestinal disorders, but its profitability is discussed in critically ill patients, mainly because of the complexity of colonic preparation. As the profitability of colonoscopy in Intermediate or Intensive Care Units (CICU) has been scarcely reported (1), we investigated its indications and usefulness.

Patients and methods: Retrospective bicenter observational study (2004–2015). Main endpoint: diagnostic profitability of CICU in unselected critically ill patients. Profitability was a priori defined as “high” if CICU led to adapt ongoing therapies; allowed an endoscopic intervention; or participated in the decision to limit therapeutic effort. Secondary endpoints: describe the quality of CICU and its preparation; determine its position in diagnosis strategy; describe its morbidity.

Results: One hundred and eleven CICU were performed in 84 patients (male sex n = 50 + 60%, age 72y [63–84] + SAPSII 73 [62–84]) within 3 days [1–10] after admission. Gastro-intestinal bleeding or haemorrhagic shock was the cause of admission in 35 patients (42%). On the day of CICU, 37 patients (44%) had shock. CICU indications were—haemorrhage (n = 77) with 35% of acute anaemia (n = 39) + suspected colitis (ischemic [n = 19, 17%], infectious [n = 17, 15%], toxic [n = 5, 5%], inflammatory [n = 1, 1%]) + investigation of a Gram negative bacilli sepsis (n = 7, 6%) + sigmoid volvulus (n = 4, 4%) + and cancer diagnosis (n = 4 + 4%). CICU profitability was deemed high in 50% (n = 56), with an endoscopic intervention performed in 32% (n = 35). The CICU lead to antimicrobial adaptation (n = 9), emergent surgery (n = 10), or to limit therapeutics effort (n = 2). In 23 cases (21%) the CICU was considered normal. Patients’ preparation was rated as good in 27% (n = 30) + and the colonoscopy was complete in 33 (30%). The CICU was mainly performed as a 2nd (58%) or 3rd (40%) investigation after an abdominal CT-scan or an upper digestive endoscopy (respectively performed in first instance in 69 and 55%). Three CICU were complicated by 3 hemodynamic and 2 respiratory failures, none were fatal.

Discussion: In our series of unselected critically ill patients, CICU were mainly performed to investigate lower gastro-intestinal bleeding. Despite a low rate of good preparation, CICU is safe and its profitability is high in the majority of cases.

Conclusion: Although performed in poor conditions, CICU seems useful in the diagnostic and the therapeutic management of critically ill patients, and not only in gastro-intestinal bleeding. (1) Church, Surgical endoscopy 2014.

Correspondence: Masse Juliette - juliette.masse2911@gmail.com

Annals of Intensive Care 2018, 8(Suppl 1):F-57

Introduction: Accurate pain assessment is associated with better outcomes in intensive care unit (ICU) patients. Specific scales for non-communicative patients have been developed and validated but their routine use still remains inaccurate and subjective. Analgesia Nociception Index (ANI) is based on high-frequency heart rate variability. This study objective was to assess the correlation between the Behavioral Pain Scale (BPS) and ANI during care procedures in deeply sedated patients.

Patients and methods: We conduced a French multicentric prospective observational study with blinded continuous recording of ANI during 8 h with spotting of care procedures in patients with RASS less or equal to − 2. We compared pain assessment using BPS and ANI before (t1) and during (t2) each care procedure. The cares analyzed included prick glycaemia, turning, catheter insertion, dressing change and others. A behavioral pain reactivity (BPR) was defined by a BPS elevation of at least 1 point. We analyzed minimal ANI values and its variations with calculation of DeltaANI (ANIt1– ANIt2). Because of the analysis of several cares per patient we used a Bonferroni’s correction in comparison of BPR and no BPR groups with a significant p value < 0.025 for this comparison. For others analyses the p value considered as significant was p < 0.05. Correlation between ANI and BPS was analyzed using a Spearman correlation rank test.

Results: Nighty two cares were analyzed in 46 patients. A BPR was noted for 29 cares (32%). ANI before care procedure was comparable in BPR and no BPR groups (p = 0.19). The most frequent care was turning in the BPR group (55%) and prick glycaemia (40%) in the no BPR group. Compared to the period before care, minimal ANI decreased significantly during care procedure in BPR group (median (IQR) 53 [35–82] vs 43 [28–50] + p < 0.001), whereas it did not reach statistical significance in the no BPR group (49 [35–58] vs 44 [31–60] + p = 0.049). DeltaANI was higher in the BPR group than in the no BPR group (8 [2–33] vs 3 [− 4 to 13] + p = 0.009). A low correlation was noted between DeltaANI and BPS during care (r = 0.211, p = 0.044).

Conclusion: ANI decreases significantly during procedures where BPS raises and not during procedures with constant BPS. A correlation was founded between BPS and DeltaANI. Thus, ANI monitoring could be a helpful tool for daily assessment of pain in sedated critically ill patients.

Correspondence: Mokline Amel - dr.amelmokline@gmail.com

Annals of Intensive Care 2018, 8(Suppl 1):F-58

Introduction: The pain associated with burn was one of the most painful injuries to treat. Pain was induced by therapeutic acts such as wound debridement, dressing and other painful procedures. Burn pain caused changes in neurophysiology and pharmacokinetics that may make standard pharmacologic analgesic therapy less effective than usual.virtual reality has been explored as an adjunct therapy for the management of acute pain for a number of conditions. In our study, we attempt to assess the impact of virtual reality on management of burn pain during dressing changes.

Patients and methods: Before the therapeutic procedure (dressing changes), the concept of virtual reality therapy was explained to the patient (technology and equipment used). The video used was snow mountain. During the act, pain was assessed until the end of the procedure. The assessment of pain was based on visual analog scale (vas). For pain intensity, the scale was most commonly anchored by “no pain” (score of 0) and “very intense pain” (score of 10).

Results: During the study period, 20 patients were included. The mean age was 32 ± 17 years. 75% of our patients were adults aged over 20 years. They were 17 men and 3 women. the average burned surface area was 30 ± 13%. Pain was evaluated before the start of the therapeutic procedure. The mean initial pain severity score was 8.85 ± 0.74 (range 8 to 10). The pain assessment after virtual reality condition showed a significant decrease in the intensity of pain (p < 0.01). The mean pain decreased from 8.85 to 4.6 ± 0.84 with extremes ranging from 1 to 4.

Conclusion: Our study supports the use of virtual reality, simple non-invasive, as an adjunct therapy in the management of pain associated with dressing changes in burn patients.

Correspondence: Amar Yaël - yamar80@gmail.com

Annals of Intensive Care 2018, 8(Suppl 1):F-59

Introduction: Hypno-analgesia (HA) is used in the operating room and for complex pain. Before implementation of HA in our intensive care unit (ICU), most protocols for algogenic procedures included intravenous or epidural morphine and Nitrous Oxid. Since 2016, many caregivers have been trained, HA has been implemented and patient comfort is evaluated using 1) a specific analogic scale of comfort (0 to 10) before and after the procedure + 2) at the end of the procedure, a score of patient and caregiver comfort using a five item questionnaire (5 to 20 points).

Patients and methods: This pilot prospective study compares HA versus the standard protocol in the removal of abdominal drains after digestive surgery. The main objective was to evaluate the patient comfort before after the procedure using a scale of comfort + the secondary objectives were to test the patient and caregiver comfort scores and evaluate in the impact on consumption of analgesic. Between May 2016 and September 2017, two groups were obtained, according whether the procedure was performed by HA-trained or non-HA-trained professionals (depending on caregivers availability in the unit). The number of subjects required to compare scales of comfort before vs. after drain removal was 66, using a nonparametric Wilcoxon–Mann–Whitney test.

Results: Eighty-eight patients were analyzed. The mean note in the comfort scale remained unchanged after vs. before drain removal in patients without HA (n = 43, +0.04 points, ± 2.8), while it increased in patients with HA (n = 45, +2.07, ± 2.7 + p = 0.001). Using our specific five item comfort score, patients and caregivers had a comparable level of satisfaction in HA and non-HA groups (Patients 17.4 20 and 16.3 20 + caregivers—19 20 in both). A trend was observed in reduction of the consumption of morphine and Nitrous Oxid with HA, without altering their comfort.

Discussion: Despite its limitations (mainly, its open non-randomized design), this study suggests that—HA may be used for algogenic procedures and is willingly adopted in ICU by patients and professionals + specific scales scores, adapted for HA, may be useful to assess the effectiveness + finally, HA seems to be at least as efficient as classical procedures and could reduce the use of analgesic drugs.

Conclusion: HA adds value to patients and to all caregivers. Prospective randomized studies are needed to valid the comfort scores we proposed, and to prove that HA reduces the consumption of analgesic drugs.

Correspondence: Guilbaut Victoria - victoria.guilbaut@hotmail.fr

Annals of Intensive Care 2018, 8(Suppl 1):F-60

Introduction: Pain has long been a focus of concern for doctors and caregivers. In intensive care unit, the inability to verbalize discomfort and pain are major stressors for patients. Music therapy has demonstrated in many international studies its effect on the blood pressure and on the respiratory frequency. In this context, we conducted a study to evaluate the effects of standardized musical intervention on pain during painful cares in vigils patients hospitalized in critical care.

Patients and methods: Design—We conduct a prospective, observational, randomised, single blind, mono center study. 140 painful cares were studied and then distributed in two groups (n = 70 with music, n = 70 without music). The patients were equiped with a Bose© helmet, and had or not music therapy during the care. Our main criteria was the pain, it has been evaluated by a numeric scale before and after the painful care. We also estimated anxiety with the COVI’s heteroevaluation scale before and after the car. We also noticed if the care were stopped because of the pain, then we used a semi quantative numeric scale in order to estimate the feeling of the caregiver and the patient on the session.

Results: Concerning pain, there is no significant difference between the two groups (p > 0.005). However, in the music group, pain decreased by 35% after the care (p < 0.001). Anxiety was way lower in the music group than in the group without music (p < 0.001). We also noticed a decrease of 50% of the anxiety in the music group. The patients and the caregivers’ feeling were the same in the two groups, with no significant difference (p > 0.05). On the other hand, caregivers tended to underestimate the difficulty of the session in comparison with the patients’ (p < 0.01) in both groups.

Conclusion: Music therapy did not improve the pain in a significant way, in the music group versus the group without but allowed a decrease of 35% of the pain after the care. Nevertheless, music reduced by two patients’anxiety.

Correspondence: Pottier Véronique - veroniquepottier@icloud.com

Annals of Intensive Care 2018, 8(Suppl 1):F-61

Introduction: Sedation and analgesia is one of the basic themes in ICU as complications associated with excessive sedation negatively impact the morbidity and mortality of patients. The objective of this study is to show that the nurse implementation of a sedation and analgesia algorithm is beneficial to the patient in terms of sedative drugs reduction and thus overall decrease in duration of mechanical ventilation (MV) and the morbidity and mortality which is associated with it, without altering patient comfort and tolerance of the environment.

Patients and methods: A before and after prospective, observational, non-interventional study was conducted in surgical ICU in Caen University Hospital, between November 2014 and April 2017. Mechanically ventilated patients under sedation predicted to last 48 h or more were included. During the “Before” period, sedation and analgesia was managed by the physician, while during the “After” period, it was managed by the nurses according to the protocol.

Results: 1156 intubated and mechanically ventilated patients were admitted during the study period. Among the 145 eligible patients, 100 were included during “Before” period and 45 during “After” period. The duration of MV after inclusion was significantly shorter in group “After” (10.5 [7 + 18] vs 8 [5 + 11.5] days, p = 0.042), as the duration of target RASS (-2 à 0) was significantly longer (0 [0 + 2] vs 1 [0 + 3] day, p = 0.038), the duration of RASS < − 2 significantly shorter (6.5 [4 + 12] vs 3 [2 + 5] days, p < 0.001), dose of sedative drugs was significantly decreased (1330 [517.5 + 2543.8] vs 315 [15 + 720] mg, p < 0.001 for hypnotics and 1803 [1097.5 + 4290] vs 900 [450 + 1680] µg, p < 0.001 for opioids, respectively), and sedation cost (25.1 [12.7 + 67.4] vs 12.2 [7.3 + 34] euros, p = 0.004). The patients experienced less of ventilator-acquired pneumonia (VAP) and delirium during the “After” period (55 vs 24.4%, p = 0.004, and 41 vs 26.7%, p = 0.015, respectively).

Conclusion: The nurse implementation of a sedation and analgesia algorithm was associated with a trend towards reduction in duration of MV, ICU and hospital length of stay. Moreover, prevalence of VAP and delirium was reduced, in correlation to the significant decrease in sedative drugs. This type of algorithm is necessary to reduce morbidity and mortality associated with MV.

Correspondence: Pichon Xavier - pichon.x@chu-toulouse.fr

Annals of Intensive Care 2018, 8(Suppl 1):F-62

Introduction: Central venous catheter insertion is a common practice for anesthetists and intensivsts. This invasive procedure generates pain and anxiety for patients. We aim to demonstrate that remifentanil improves the analgesia during scheduled central venous catheter insertion in mindful patients.

Patients and methods: A prospective, randomized, double-blind, controlled study in patients requiring central venous access. Patients were randomly assigned to receive 3 ng ml⁻¹ remifentanil target controlled infusion (TCI) and local anesthesia (LA) with lidocaine or placebo and LA. All patients were monitored in intensive care or post-intervention care unit and systematically received oxygen. Patients were asked to assess verbal numeric rating pain scale (VNRPS) during the procedure. The primary outcome was the maximal VNRPS. Secondary outcomes were pain at each step, anxiety, patient satisfaction, operator ease and side effects.

Conclusion: TCI remifentanil is a safe procedure to reduce pain during central venous catheter insertion in awake patients. Trial Registration—clinicaltrials.gov Identifier—02206022, REMIDOLCATH.

Correspondence: Paul Marine - marine.1604@hotmail.fr

Annals of Intensive Care 2018, 8(Suppl 1):F-63

Introduction: Although guidelines on post-resuscitation care recommend the use of short-acting agents for sedation during targeted temperature management (TTM) after cardiac arrest (CA), the potential interests of this strategy have not been clinically demonstrated.

Patients and methods: Before-after study. We compared two sedation regimens (propofol-remifentanil, period P2 vs midazolam-fentanyl, period P1) among comatose TTM-treated CA survivors. Management protocol, apart from sedation and neuromuscular blockers use, did not change between the two periods. Baseline severity was assessed with Cardiac-Arrest-Hospital-Prognosis (CAHP) score. Time to awakening was measured starting from discontinuation of sedation at the end of rewarming. Awakening was defined as delayed when it occurred after more than 48 h.

Results: 460 patients (326 in P1, 134 in P2) were included. CAHP score in P2 and P1 did not significantly differ (p = 0.93). Sixty percent of patients awoke in both periods (194 326 vs 81 134, p = 0.85). Median time to awakening was 2.5 (IQR 1–9) hours in P2 vs. 17 (IQR 7–60) hours in P1. Awakening was delayed in 6% in P2 vs. 29% of patients in P1 (p < 0.001). After adjustment, P2 was associated with significantly lower odds of delayed awakening (OR 0.08, 95% CI 0.03–0.2 + p < 0.001). Patients in P2 had significantly more ventilator-free days, and lower catecholamine-free days between admission and day 28. Survival and favorable neurologic outcome at discharge did not differ across periods.

Time course for awakening according to sedation period.

These figures report the time course of patients included after discontinuation of sedation. At each time point, we report in the upper part, proportion of patients awoken after discontinuation of sedation, in the lower part, patients who died without awakening, and in the middle part (in white), patients still comatose. Red dots indicate, for each period, the last patient awakening (after 12 days in P1 and 5 days in P2).

Black dots indicate the median time to awakening (too early to appear for period 2).

Conclusion: During TTM following resuscitation from CA, sedation with propofol-remifentanil compared with midazolam-fentanyl was associated with an earlier awakening, and an increase in ventilator-free days.

Correspondence: Perez Justine - justine.perez03@orange.fr

Annals of Intensive Care 2018, 8(Suppl 1):F-64

Introduction: The new recommendations of the French society of anesthesia-intensive care (SFAR) on perfusion and medication errors were revised in 2016 to promote proper use relating to drug administration with medical devices. To advance that of inquiry, practices of our intensive care unit (ICU) were assessed in order to improve drug administration by central venous catheter (CVC).

Patients and methods: Prospective evaluation by pharmacist resident and technical nurse during seven weeks, using a standard evaluation tool, in a 16 bed ICU. Drug recommendations and SFAR documents from 2016 were used as referential of conformity. The following parameters were evaluated—1 Central venous lines mounting, 2 Drug administration and identification with a focus on narrow therapeutic index (NTI) drugs.

Results: 50 patients with CVC were analyzed between June and July 2017. 19 entered directly in the ICU. 23 were hospitalized for surgical reasons. 42 had triple-lumen CVC. Regarding the first parameter, no conformity was found due to lack of line identification (100%) or anti-return valve well positioned (96%). Perfusion ramp position was above heart level in 10%, infusion tubing had contact with floor in 6%, and absence of plug on non-used lines is found in 8% of cases. Regarding second parameter, non-conformities were due mostly to syringe label—absence of drug’s concentration (72%), preparator identification (99%), patient identification (99.7%), drug identification (3.5%, all concerning propofol), date and time of medication preparation (67%), lack of color code of labels (100%). Regarding NTI, 70% were not administrated according to the recommendations—absence of dedicated line (63%), absence of administration on the nearest insertion site of the catheter (67%).

Conclusion: The evaluation highlight some improvement axis such as complete identification on syringes, sensitizing of ICU healthcare team, or homogenization of CVC perfusion system. It calls for a second evaluation round after implementation of improvements.

Correspondence: Piriou Pierre-Guillaume - pgpiriou@wanadoo.fr

Annals of Intensive Care 2018, 8(Suppl 1):F-65

Introduction: Sedation is a corner stone of the care of patients receiving mechanical ventilation in the ICU. Sedation was associated with increased comfort and adherence to care, but also with increased morbidity, including delirium, increased duration of mechanical ventilation and length of ICU stay. Previous studies reported beneficial impact of reduced doses of sedative drugs and careful monitoring of patients comfort and consciousness. Our goal was to assess the impact of the introduction of a nurses-dedicated sedation protocol in our ICU.

Patients and methods: This monocentre retrospective before-after study included all the patients admitted in our ICU, over two three-month periods, from July 2015 and January 2016, treated with invasive mechanical ventilation for more than 24 h and older than 18yrs. After the first period, all physicians and nurses were trained to a new sedation management protocol. Analysis was performed to assess the prescription and application of the protocol, its impact on the use of sedative drugs, ICU length of stay, and duration of mechanical ventilation. Major complications were also recorded.

Results: 134 patients were included—78 before and 56 after the protocol implementation. Patients in both groups had similar baseline characteristics (Men 67 vs. 50%, p = 0.08 + mean age 56 ± 18 vs. 60 ± 16 years, p = 0.1 + weight 73.2 ± 20.6 vs. 70.2 ± 15.9 kg p = 0.61 + IGS2 44 ± 17 vs. 44 ± 19, p = 0.99 + medical admission 91 vs. 88%, p = 0.51). Recordings of RASS and BPS did not differ between groups (9 ± 4 vs 10 ± 9, p = 048 + 19 ± 25 vs, 16 ± 11, p = 0.25). The duration of sedation was significantly shorter after introduction of protocol (4.1 ± 4.7 vs 1.9 ± 1.0, p < 0.0001), as was the duration of mechanical ventilation (9.3 ± 13.8 vs 4.1 ± 3.2, p = 0.001) and ICU length of stay (12.3 ± 15.8 vs 7.3 ± 5.3, p = 0.028). There was no difference in major ICU complications, nor in mortality between groups (25 and 30%).

Conclusion: Although the implantation of a sedation protocol did not translate in increased recording of RASS and BPS scores, it was associated with improved outcomes. Our data suggest that, more than the protocol by itself, beneficial effects reported after the implementation of a sedation protocol may be ascribed to increased awareness of the care givers and thus better management of sedation.

Correspondence: Goyer Isabelle - goyer-i@chu-caen.fr

Annals of Intensive Care 2018, 8(Suppl 1):F-66

Introduction: Critically ill patients often necessitate multiple high-risk supportive pharmacologic treatments and exhibit complex and rapidly changing pharmacokinetic parameters. Their constantly evolving medical condition warrants continuous reevaluation and adjustments in pharmacotherapy. Multiple studies have shown the impact of pharmacist’s interventions on the reduction of medication prescription errors and extrapolated hypothetic cost savings. Critical care societies around the world recommend full time specialized clinical pharmacy services in the intensive care unit (ICU). Nonetheless, the full impact of specialized clinical pharmacy in the critical care unit should be appreciated in terms of meaningful direct patient outcomes such as length of ICU stay, length of mechanical ventilation (MV), risk of MV associated pneumonia or other nosocomial infections, risk of delirium, thromboembolic events or mortality and direct actual expenditure savings.

Patients and methods: Literature review (PubMed, google scholar and cross-referencing of selected articles) with selected Mesh terms (Intensive Care Units [Mesh] OR Critical Illness [Mesh] OR Resuscitation [Mesh]) AND (Pharmacists [Mesh] OR Pharmacy [Mesh]) done on August 28th, 2017. Original research studies written in English or French evaluating the impact of an ICU clinical pharmacist’s care on clinical endpoints were selected. All available articles were included without time limit, oldest article dating back to December 1991.

Results: The literature search yielded 157 references to which additional relevant references were added by cross-referencing selected articles. 26 studies evaluating the impact of clinical pharmacists on patient outcomes in the critical care setting were selected for this analysis. One hundred percent of the studies reported positive measurable clinical outcomes of the presence of a clinical pharmacist in critical care setting. Eighty-eight percent of the studies were done in the USA and 31% were prospective trials (n = 1348 patients). Six studies evaluated actual monetary expenditure and 100% of them concluded in cost-effectiveness of critical care clinical pharmacy services.

Conclusion: ICU clinical pharmacists do not only intercept medication prescription errors once they happen and decrease related hypothetic institutional costs. Their full time presence and interdisciplinary team inclusion in the ICU globally improves patient care and translates into tangible benefits in terms of patient outcomes and direct monetary savings. We can no longer ignore international recommendations warranting full time specialized clinical pharmacy services in critical care units to ensure patient care efficiency, safety and cost-effectiveness. Critical care clinical pharmacy services should be considered as an essential element of a standard ICU.