Hospital-acquired pneumonia in ICU

The French Society of Anesthesia and Intensive Care Medicine and the French Society of Intensive Care edited guidelines focused on hospital-acquired pneumonia (HAP) in intensive care unit (ICU). The goal of 16 French-speaking experts was to produce a framework enabling an easier decision-making process for intensivists. The guidelines were related to 3 speciﬁc areas related to HAP (prevention, diagnosis and treatment) in 4 identiﬁed patient populations (COPD, neutropenia, postoperative and pediatric). The literature analysis and the formulation of the guidelines were conducted according to the Grade of Recommendation Assessment, Development and Evaluation methodology. An extensive literature research over the last 10 years was conducted based on publications indexed in PubMed TM and Cochrane TM databases. HAP should be prevented by a standardized multimodal approach and the use of local prevalence of MRSA, recent colonisation by MRSA, chronic skin lesions, chronic renal replacement therapy.


Introduction
Hospital-acquired pneumonia (HAP) is the most common infection in the intensive care unit (ICU). This infection encompasses two different entities: pneumonia associated with mechanical ventilation (ventilator-associated pneumonia or VAP) and severe pneumonia developed during the hospital stay. The incidence of VAP ranges from 1.9 to 3.8 per 1000 days of mechanical ventilation in the US and exceeds 18 per 1000 days of mechanical ventilation in Europe [1].
Nosocomial pneumonia is the most common infection in ICU, when considering the timing of these infections. Non-ventilator HAP occurs in patients admitted to the hospital for at least 48 hours and VAP is defined as occurring more than 48 hours after the initiation of mechanical ventilation. Accurate data on their epidemiology are limited by the lack of standardised diagnostic criteria. In the US, the incidence of non-ventilator-HAP was 1.6%, representing a rate of 3.63 per 1000 patient-days [2], while the definition of VAP is not altered by considering time after admission to the hospital and the incidence remain as mentioned above [1].
In the ICU, HAP is associated with an approximate mortality rate of 20% [3]. However, the mortality attributable to HAP is selective digestive decontamination in units where multidrug-resistant bacteria prevalence was below 20%. Diagnosis relies on clinical assessment and microbiological findings. Monotherapy, in the absence of risk factors for multidrug-resistant bacteria, non-fermenting Gram negative bacilli and/or increased mortality (septic shock, organ failure), is strongly recommended. After microbiological documentation, it is recommended to reduce the spectrum and to prefer monotherapy for the antibiotic therapy of HAP, including for non-fermenting Gram-negative bacilli. C 2017 The Authors. Published by Elsevier Masson SAS on behalf of Socié té française d'anesthé sie et de ré animation (Sfar). This is an open access article under the CC BY license (http://creativecommons.org/ licenses/by/4.0/). estimated between 5 and 13% [3]. The attributable mortality is probably higher in some specific populations such as patients with chronic obstructive pulmonary disease (COPD) [1]. In other populations such as trauma patients, HAP seems to have only a minor effect on mortality [1]. Even though the direct impact of this infection on mortality remains debated, it is nonetheless associated with increased morbidity through increased duration of mechanical ventilation (or decrease in ventilator-free days) and increased ICU and hospital length-of-stay [1]. As a result, HAP is also responsible for an overuse of healthcare resources (ventilation, ICU and hospital beds and resources). Finally, it is associated with increased costs related to hospital stay [4].
To date, the French Society of Anaesthesia and Intensive Care Medicine (SFAR) and the French Society of Intensive Care (SRLF) have not proposed guidelines focused on HAP. Sixteen Frenchspeaking experts were selected by an organising committee, itself appointed by the guideline committees of each participating society, following approval by their respective boards. Two independent bibliographic experts analysed the literature of the past 10 years in the field using pre-defined keywords.
Recent American and European guidelines have been published (http://www.idsociety.org/Guidelines/Patient_Care/IDSA_Practice_ Guidelines/Infections_by_Organ_System/Lower/Upper_Respiratory/ Hospital-Acquired___Ventilator_-_Associated_Pneumonia_(HAP/ VAP)/). However, both the SFAR and SRLF wished to share their own interpretation encompassing data from recently available publications. The boards of both societies designated an organising committee, which appointed a panel of sixteen French-speaking experts. Two members of the panel oversaw a literature search using pre-defined keywords limited to the past decade.

Guidelines goals
The goal of these experts' guidelines is to generate a framework enabling an easier decision-making process for intensivists. The group worked to produce a minimal number of guidelines in order to highlight the key points to focus on in each area. When in doubt, published data prevailed over expert opinion. The basic rules of universal medical good practices -hygiene, antibiotic therapy, comprehensive care -were considered as known and excluded from the scope of the guidelines.

Definitions
The criteria to diagnose pneumonia are shown in Table 1. In clinical practice, HAP is suspected when a patient presents with fever, impaired oxygenation and suppurative secretions.
Non-ventilator HAP occurs after 48 h of hospital stay and VAP occurs after 48 h of mechanical ventilation [5]. By definition, neither is present nor incubating at hospital admission nor at the onset of mechanical ventilation. The onset of either HAP or VAP relative to hospital admission discriminates early pneumonia (< 5 days) from late pneumonia (! 5 days). Microbiologically confirmed HAP was defined as the definitive identification of a microorganism isolated from respiratory samples or in blood cultures in a patient with suspected pulmonary infection.

Microbiological confirmation
The pathogens usually responsible for HAP are Enterobacteriaceae, Staphylococcus aureus, Pseudomonas aeruginosa and Acinetobacter baumannii [1]. The infection is polymicrobial in 30% of cases. In early pneumonia, the most frequently identified pathogens are methicillin susceptible S. aureus, Streptococcus pneumonia and Haemophilus influenza [1].
Microbiological confirmation is a crucial step in the diagnosis of HAP. Routinely, it is based on the qualitative or quantitative cultures of respiratory samples. A pathogen is isolated from these samples and identified in about 70% of suspected cases [1]. Other biological tests can be carried out to support the microbiological diagnosis such as blood cultures and antigenuria. Molecular biology techniques applied to routine microbiology are still being assessed and outside of the scope of these guidelines.

General organisation
These guidelines are the result of the work provided by a panel of experts, brought together by the SFAR and the SRLF. Each expert was required to file a conflict-of-interest disclosure prior to participation in establishing the guidelines. The stated mission was to produce guidelines in three specific areas related to HAP: prevention, diagnosis and treatment as well as the specificities pertaining to different identified patient populations (COPD, neutropenia, postoperative and pediatric).
The schedule of the group was defined upstream (Table 2). First, the organisation committee and the guideline coordinators defined the questions to be addressed by the panelists. It then appointed experts in charge of each question. The questions were formulated according to the Patients, Intervention Comparison Outcome (PICO) format following a first meeting of the expert panel. The literature analysis and the formulation of the guidelines were then conducted according to the Grade of Recommendation Assessment, Development and Evaluation (GRADE) methodology. A level-of-evidence was defined for each reference cited according to the type of study. This level-of-evidence could be re-evaluated taking into account the methodological quality of the study. Overall ''strong'' level-of-evidence resulted in formulating a ''strong'' recommendation (''we recommend. . .'' or ''we recommend not. . .''; GRADE 1+ or 1À). Overall moderate, weak or very weak level-of-evidence resulted in formulating an ''optional'' recommendation (''we suggest. . .'', ''we suggest not. . .''; GRADE 2+ or 2À). In the absence of supporting literature, a question could be addressed by a recommendation under the form of an expert opinion (''the experts suggest that. . .''). Recommendation proposals were presented to the entire panel and discussed one-by-one. The goal was not to obtain consensus on all the proposals, but to Guideline finalisation meeting Table 1 Criteria for defining pneumonia.
Radiological signs Two successive chest radiographs showing new or progressive lung infiltrates In the absence of medical history of underlying heart or lung disease, a single chest radiograph is enough And at least one of the following signs Body temperature > 38,3 C without any other cause Leukocytes < 4000/mm 3 or ! 12000/mm 3 And at least two of the following signs Purulent sputum Cough or dyspnea Declining oxygenation or increased oxygen-requirement or need for respiratory assistance identify points of agreement and disagreement or indecision. Each recommendation was then evaluated by each expert rated using a scale from 1 (complete disagreement) to 9 (complete agreement). Collective rating was established according to a GRADE grid methodology. To validate a recommendation on a criterion, at least 50% of experts had to have concordant opinions, while less than 20% of them had to have discordant opinion. For a recommendation to be strong, at least 70% of participants had to have concordant opinions. Without strong agreement, recommendations were rephrased and resubmitted to reach a consensus.

Areas of guidelines
Three areas related to HAP were: prevention, diagnosis and treatment with population-specific considerations. Four relevant specific populations were considered in the guidelines analyses (COPD, neutropenia, postoperative and pediatric). An extensive literature research over the last 10 years was conducted based on publications indexed in PubMed TM and Cochrane TM databases. To be considered for analysis, publications had to be written in English or in French. It was decided ahead of the analysis to limit the number of expert opinions and to not produce recommendations that are not supported by literature. The analysis was focused on recent data, according to decreasing hierarchical prioritisation of data from meta-analyses of randomised controlled trials (RCTs) or individual RCTs to observational studies. Study sample size and the relevance of the research were considered at the level of each study.

Synthesis of results
The expert panel analyses and application of the GRADE method led to 15 guidelines (and two specific guidelines for the pediatric population) and four care protocols. Among the 15 formalised guidelines for adults, three have a high level-of-evidence (GRADE 1+/À) and 11 a low level-of-evidence (GRADE 2+/À). For one recommendation, the GRADE method could not be applied, resulting an expert opinion. Both paediatric guidelines had a weak level-of-evidence (GRADE 2). The four care protocols, provided only as an indication, are based on expert opinions but were submitted to the same rating and agreement process as the guidelines. After two rounds of rating and one amendment, a strong agreement was reached for all guidelines and protocols.

First area, PREVENTION:
Which HAP prevention approaches decrease morbidity and mortality in ICU patients? R1.1 -We recommend using a standardised multimodal HAP prevention approach in order to decrease ICU patient morbidity.
GRADE 1+, STRONG AGREEMENT Evidence summary and rationale for the recommendation: The implementation of standardised protocols is associated with a decrease in ICU morbidity [6][7][8][9][10]. In before/after studies, the implementation of a multimodal prevention strategy and the adherence to a standardised multimodal approach are associated with a decrease in HAP [11] and in the duration of mechanical ventilation, provided that an early weaning strategy is part of the approach [12]. Experts suggest increasing prevention measures, and applying costly measures to patients with a high risk of HAP and death, such as COPD patients [13] or immunosuppressed patients. R1.1 Paediatrics -We suggest using a standardised multimodal approach aiming at preventing HAP in order to decrease pediatric ICU patient morbidity. GRADE 2+, STRONG AGREEMENT Evidence summary and rationale for the recommendation: Eight monocentric before/after or cohort studies suggest that implementing and applying standardised HAP prevention protocols in pediatric ICU patients are of interest in order to reduce the occurrence of HAP and associated morbidity [14][15][16][17][18][19][20][21].
R1.2 -In units where multidrug-resistant bacteria prevalence is low (< 20%), we suggest applying routine selective digestive decontamination using a topical antiseptic administered enterally and a maximal 5-day course of systemic prophylactic antibiotic to decrease mortality. GRADE 2+, STRONG AGREEMENT Evidence summary and rationale for the recommendation: Meta-analyses of RCTs comparing selective digestive decontamination (SDD) to standard care show a significant decrease in hospitality mortality, lengths of mechanical ventilation and HAP incidence in ICU patients [22]. In a sub-group analysis from a multicentre trial, the effect of SDD on mortality decrease was similar in medical and surgical patients [23]. In the sub-group analysis of a meta-analysis of RCTs on pneumonia prevention strategies, the effect of SDD on the decrease in mortality was only observed for strategies including a topical antiseptic administered enterally and systemic prophylactic antibiotic use [24]. The effect of SDD on the decrease in mortality was greater in patients with high overall mortality, demonstrating greater efficiency in the more critically ill [24]. In the sub-group analysis of a large multicenter trial, SDD was associated with a decrease in the acquisition of multidrug resistant bacteria [25]. No link between SDD and the development of bacterial resistance in ICU patients was shown in a meta-analysis or in a randomised-controlled trial comparing SDD to standard care [26]. It is probably best to limit the duration of systemic antibiotic therapy in SDD protocols to a maximum of 5 days, because prolonged antibiotic therapy may lead to the emergence of multidrug-resistant bacteria [27].
The major studies that have demonstrated the efficiency of SDD were conducted in environments where the prevalence of multiresistant bacteria was low [25]. It is therefore recommended to follow the effects of this prevention strategy on the local bacterial ecology on a regular basis. Consequently, we cannot recommend such a strategy in units where the prevalence of multiresistant bacteria is high.
R1.3 -Within a standardised multimodal HAP prevention approach, we suggest combining some of the following methods to decrease ICU patient morbidity: promote the use of non-invasive ventilation to avoid tracheal intubation (mainly in post-operative digestive surgery patients and in patients with COPD); favor orotracheal over nasotracheal intubation when required; limit dose and duration of sedatives and analgesics (promote their use guided by sedation/pain/agitation scales, and/or daily interruptions); initiate early enteral feeding (within the first 48 hours of ICU admission); regularly verify endotracheal tube cuff pressure; perform sub-glottic suction (every 6 to 8 hours) using an appropriate endotracheal tube.

GRADE 2+, STRONG AGREEMENT Evidence summary and rationale for the recommendation:
Within multimodal HAP prevention approaches, methods with an effect on patient morbidity should be favored, in particular those with an effect on length of mechanical ventilation (ventilator-free days), ICU and hospital lengths-of-stay and use of antibiotics.
For each of the methods proposed above, a significant decrease in the risk of HAP and/or in the length of mechanical ventilation and/or in ICU/hospital length-of-stay were reported in: a meta-analysis of RCTs comparing non-invasive and invasive ventilation [28], a RCT including patients in postoperative digestive surgery patients of [29] and two meta-analyses focused on weaning from mechanical ventilation in patients with COPD [30,31]; two before/after studies on the implementation of a sedation protocol titrated by nurses [32,33]; three meta-analyses of RCTs comparing early enteral feeding (initiated before 48 h) to late feeding [34][35][36]; RCTs comparing continuous to discontinuous monitoring of tracheal intubation tube cuff pressures [37][38][39]; two meta-analyses comparing sub-glottic suction to standard care [40,41]. Tracheal intubation tubes with sub-glottic suction may also have a positive cost-benefit ratio; a randomised study comparing nasotracheal versus orotracheal intubation [42].
R1.4 -Within a standardised multimodal HAP prevention approach, we suggest not using the following methods to decrease ICU patient morbidity: systematic early (< day 7) tracheotomy (except for specific indications); anti-ulcer prophylaxis (except for specific indications); post-pyloric enteral feeding (except for specific indications); administration of probiotics and/or synbiotics; early systematic change of the humidifier filter (except for specific manufacturer recommendations); use of closed suctioning systems for endotracheal secretions; use of antiseptic-coated intubation tubes or with tubes an ''optimised'' cuff shape; selective oro-pharyngeal decontamination (SOD) with povidone-iodine; use of prophylactic nebulised antibiotics; daily skin decontamination using antiseptics. GRADE 2À, STRONG AGREEMENT Evidence summary and rationale for the recommendation: Although some of the abovementioned strategies have been associated with significant decrease in the risk of HAP, no reduction in morbidity (length of mechanical ventilation or ICU/ hospital length-of-stay) or in mortality was demonstrated with the use of these strategies individually. However, it is possible that their combined use in prevention bundles may have some beneficial effect on patient morbidity. However, there is lack of available data in the literature to support this hypothesis. Experts do not take position against these HAP prevention techniques, but their first line use is not recommended, either because of cost or potential adverse effects.
R1.5 -In weaning of COPD patients from ventilation, we suggest using non-invasive ventilation to reduce length of invasive mechanical ventilation, incidence of HAP, morbidity and mortality. GRADE 2+, STRONG AGREEMENT Evidence summary and rationale for the recommendation: Invasive mechanical ventilation is the most important risk factor for VAP. Several studies have analysed the role of noninvasive ventilation (NIV) within the process of weaning from mechanical ventilation. Their goals were the reduction of the length of invasive mechanical ventilation and in VAP incidence. Two meta-analyses [30,31] assessed the impact of NIV, as a means of rapid weaning from invasive mechanical ventilation, on the incidence of VAP.
In the first meta-analysis (9 studies, 632 patients with COPD), NIV reduces the risks of mortality (risk ratio [ [31]. However, this metaanalysis had several issues: several studies included in this meta-analysis were not randomised; number of patients included in most of the studies was under 50; definition of VAP was not the same in every study.

Second area, DIAGNOSIS:
What methods to diagnose HAP should be used to decrease ICU patient morbidity and mortality?
GRADE 2À, STRONG AGREEMENT Evidence summary and rationale for the recommendation: The performance of the clinical pulmonary infection score (CPIS) in the diagnosis of pneumonia depends on the comparator. Its sensibility and specificity vary from 60 to 80% compared to microbiological samples obtained by bronchoalveolar lavage [58][59][60][61]. The diagnostic performances are low when the comparator is based on histology and pathology of postmortem lung biopsies in autopsy series [62]. The performance of the CPIS depends on the pre-test probability of pneumonia [63]. Nevertheless, CPIS variation over time may be useful in pneumonia resolution [64] and antibiotic de-escalation [65,66]. Initial CPIS at HAP onset has little value in predicting patient outcomes, as compared to severity scores [67].
R2.2 -We suggest collecting microbiological airway samples, regardless of type, before initiation of or any change in antibiotic therapy. GRADE 2+, STRONG AGREEMENT Evidence summary and rationale for the recommendation: A meta-analysis comparing different airway samples (endotracheal aspirates, protected specimen brush and bronchoscopic or blind bronchoalveolar lavage) and microbiological culture methods (quantitative or not) found no significant effect on patient outcomes (28-day mortality, ventilator-free days, or length-of-stay) nor antibiotic treatment [68]. Therefore, both sampling and culture methods are left to the discretion of clinicians according to strategic choices at the unit, department or institutional levels. There are little data specifically addressing the effect of sampling and culture methods on antibiotic consumption or antibiotic-free days. One study found an approximate increase of 2 antibiotic-free days with invasive compared to non-invasive airway sampling [69]. In the absence of any true gold standard, using post-mortem histology and pathology findings upon autopsy as gold-standard surrogates suggests a slightly greater diagnostic performance of quantitative cultures, regardless of airway sample type. Indeed, in the various autopsy case series summarised in the 2016 IDSA/ATS guidelines [70] and in a dedicated meta-analysis [71]: (a) quantitative culture, regardless of airway sample type had a diagnostic odds-ratio (DOR) > 1, i.e. could establish a diagnosis in the presence of pneumonia, albeit with a high variation in DOR ranging from close to 1 to much higher, whereas (b) semior non-quantitative cultures had a low DOR, sometimes < 1, i.e. a high probability of establishing a diagnosis in the absence of pneumonia, and c) all positive likelihood ratios (LR+) were low, close to 0.5, i.e. even in the case of high prevalence there would be a low post-test probability (40-50%) of establishing a diagnosis compared to pre-test suspected diagnosis regardless of culture method and thresholds. Therefore, there may be an advantage to quantitative cultures in establishing the diagnosis and there may be an advantage to invasive sampling in increasing antibiotic-free days. While obtaining airway samples should not delay initiation of antibiotic treatment in severe cases and/or acute respiratory distress syndrome (c.f. R3.1) and it remains essential to guiding and/or de-escalating antibiotic therapy when microbiological identification is achieved (c.f. R3.6).
In onco-hematology patients admitted to the ICU for respiratory distress, lack of an established diagnosis is significantly associated with increased mortality [72]. However, in these patients the diagnosis is most often HAP, in two-thirds of the cases [73]. In a randomised trial, 119 onco-hematology patients admitted to the ICU for respiratory distress, among which 31% were neutropenic, invasive lower respiratory tract sampling [BAL] compared to non-invasive sampling was not associated with increased risk of intubation nor increased mortality, neither was there any difference in diagnostic yield between sampling methods, around 80% [74].
R2.2 Paediatrics -We suggest collecting microbiological airway samples, regardless of type, before initiation of or any change in antibiotic therapy. GRADE 2+, STRONG AGREEMENT Evidence summary and rationale for the recommendation: Labenne et al. compared protected distal brush and blind bronchoalveolar lavage samples for VAP diagnosis [75]. Of the 103 patients, 29 were diagnosed with VAP. The combination of brushing with culture of the blind bronchoalveolar lavage sample and a bacteriological index (BI) resulted in a sensitivity (Se) of 90% and a specificity (Sp) of 88% for VAP diagnosis. Gauvin et al. compared several types of microbiological samples and qualitative or quantitative culture methods for VAP diagnosis, using the 1988 CDC VAP diagnosis criteria [76]. The best performing method for the diagnosis of VAP was the bacteriological index, BI > 5 (Se 78%, Sp 86%, positive predictive value, PPV of 70%, and negative predictive value, NPV of 90%) but the number of VAP was low. Sachdev et al. prospectively found in 30 patients a superiority of protected distal samples compared to endotracheal aspirates to diagnose HAP [77]. The same authors confirmed the excellent reproducibility of protected distal samples in 34 children with VAP, both for the cellular analysis and bacteriological diagnosis [78]. From 335 samples obtained prospectively in 61 children, Willson et al. reported the lack of usefulness of tracheal samples for VAP diagnosis, suggesting that the distinction between colonization and infection is challenging [79].
To summarise, since endotracheal tube and upper airway colonisation occur rapidly following intubation, the use of protected distal samples may provide superior specificity to endotracheal aspirates. In the absence of any demonstrated impact on outcomes, the choice of airway sampling and culture methods depends on the environment, institutional and departmental strategic choices, taking into account the appropriate threshold of the chosen method when paired with quantitative cultures.
R2.3 -We suggest not measuring plasma or alveolar levels of procalcitonin or soluble TREM-1 to diagnose HAP. GRADE 2À, STRONG AGREEMENT Evidence summary and rationale for the recommendation: Eight studies have assessed the contribution of measuring procalcitonin (PCT) plasma concentrations to the diagnosis of HAP, on their own or in combination with clinical criteria [80][81][82][83][84][85][86][87]. A total of 589 patients were included for a pneumonia prevalence of 55% and the following test characteristics: Se = 54% (48-59%), Sp = 67% (51-73%), PPV = 67% and NPV = 54%. The positive likelihood ratio (LR+) was 1.65 and negative likelihood ratio (LRÀ) 0.68. These performances are insufficient to contribute to the diagnosis decision process. Diagnostic test thresholds varied, ranging from 0.15 to 3.9 ng/mL. Seven studies assessed the contribution of alveolar concentrations of the soluble form of the Triggering Receptor Expressed on Myeloid cells-1 (sTREM-1) to the diagnosis of HAP [87][88][89][90][91][92][93]. A total of 317 patients were included for a prevalence of 48%. The sTREM-1 test characteristics were: Se = 83% (76-89%) and the Sp = 77% (69-83%), resulting in a PPV of 77% and a NPV of 83%n, a LR+ of 3.56 and LRÀ of 0.22. Even if these performances seem interesting, mainly for the exclusion of the diagnosis, the diagnostic thresholds varied widely, ranging from 5 to 900 pg./Mr. This could be explained by different and non-standardised dosage techniques, excluding the possibility of using this test in clinical practice at the moment.
It is surprising to observe the low number of patients included in these biomarker studies (PCT or sTREM-1), contributing to their poor robustness.

Third area, TREATMENT:
What therapeutic options for HAP should be used to decrease ICU patient morbidity and mortality?
R3.1 -We suggest immediately collecting samples and initiating antibiotic treatment taking into consideration risk factors for multidrug resistant bacteria in patients with suspected HAP and haemodynamic or respiratory compromise (shock or acute respiratory distress syndrome) or frailty such as immunosuppression.
GRADE 2+, STRONG AGREEMENT Evidence summary and rationale for the recommendation: The urgent need for antibiotics in septic shock does not fall within the scope of these guidelines. Clinicians should consult current guidelines on the management of septic shock. The immunosuppressed patients receive specific treatment.
In HAP, no RCT has compared a strategy based on early empirical antibiotic therapy to definitive antibiotic therapy after pathogen isolation, identification and susceptibility testing on patient outcomes, with stratification on pre-test diagnosis probability and clinical severity. However, several observational studies and a meta-analysis have shown that inappropriate initial antibiotic therapy, i.e. ineffective according to susceptibility testing of the causal pathogen, was associated with worse outcomes (length of mechanical ventilation, lengthof-stay and mortality), suggesting that the appropriate empirical antibiotic therapy is associated with improved outcomes [94][95][96][97][98][99].
R3.2 -We recommend treating HAP in mechanically-ventilated immunocompetent patients empirically by a monotherapy, in the absence of risk factors for multidrug-resistant bacteria, non-fermenting Gram negative bacilli and/or increased mortality (septic shock, organ failure). GRADE 1+, STRONG AGREEMENT Evidence summary and rationale for the recommendation: We identified three RCTs published over the last ten years [100-102] and a meta-analysis [103] of these three studies (including a prior study [104]) including 1163 patients. They compared monotherapy vs. combined therapy. All studies reported on the impact of combined therapy on mortality. Only two reported on clinical cure and the incidence of adverse effects [100,104]. No difference between monotherapy and combined therapy was found in terms of mortality ( . The moderate methodological quality mainly comes from the inaccuracy of the results and/or the heterogeneity of studies included in the meta-analysis. We decided to grade this recommendation as strong in spite of a moderate overall methodological quality because of: a convergence of published studies towards the absence of benefit of combined therapy, the high relevance of the endpoints studied in the studies, and data demonstrating adverse effects of combined therapy in terms of increased toxicity, emergence of resistance and cost [105,106]. However, several elements issues remain: the methodological quality of analysed trials in the meta-analysis ranges from moderate (mortality, ICU length-of-stay, adverse effects) to very low (clinical cure); the number and size of studies are limited. In fact, if the mortality analysis is based on four studies (n = 1163), the other judgment criteria are analysed using two studies (n = 350 for clinical recovery; n = 921 for adverse effects; n = 813 for ICU length-of-stay); observational studies and a few controlled studies have suggested that in the presence of multidrug-resistant bacteria [102] or non-fermenting Gram-negative bacilli [107][108][109][110], the risk of an inappropriate empirical antibiotic therapy is higher, resulting in higher mortality, and ICU or hospital length-of-stay.
Empirical combined therapy increases the rate of appropriate empirical antibiotic therapy in VAP caused by multidrugresistant bacteria [102,109]. It is therefore important to systematically screen for risk factors to develop an infection to or the carrying of multidrug-resistant (MDR) bacteria or nonfermenting Gram-negative bacilli, mainly Pseudomonas aeruginosa.
The established risk factors for Pseudomonas aeruginosa pneumonia are COPD, bronchiectasis and cystic fibrosis. The risk factors associated with MDR bacteria airway colonisation and HAP are [70]: antibiotic therapy in the previous 90 days; a hospital stay of more than 5 days prior to suspected HAP; renal replacement therapy requirement during HAP; septic shock; ARDS.
In the presence of at least one of these risk factors, combined empirical therapy is indicated. However, after pathogen identification and susceptibility testing, no study has shown that pursuing combined therapy remains beneficial in VAs, including VAP due to Pseudomonas aeruginosa [70,103,111].
Finally, for a given patient, a predictable mortality rate above 25% probably remains an indication for combined empirical antibiotic therapy. In a meta-analysis including observational and randomised studies, a benefit to combined therapy is found in terms of mortality (31% vs. 41%; hazard ratio: 0.71; IC 95% [0.57-0.89]) in patients with sepsis due to lung infection [112]. These results have to be taken cautiously since they depend on the spectrum of the primary antibiotic; combination therapy yielded no advantage when the primary antibiotic was itself a broad-spectrum antibiotic. This metaanalysis also showed that combined therapy in patients with low expected mortality rate (< 10%) tends to have deleterious consequences. The recommendation is therefore quite limited since it only applies to immunocompetent patients, without any risk factor for multidrug-resistant bacteria, and at low risk of mortality. In the presence of one or several of such risk factors, empirical combined therapy is probably indicated before pathogen identification and susceptibility testing results are available.
R3.3 -The experts suggest not systematically directing empiric antibiotic therapy against methicillin-resistant Staphylococcus aureus in the treatment of HAP.
EXPERTS OPINION, STRONG AGREEMENT Evidence summary and rationale for the recommendation: A study randomising patients treated empirically with an antibiotic regimen including or not an antibiotic active against methicillin-resistant Staphylococcus aureus (MRSA) did not show any difference between the two strategies in terms of length-of-stay and mortality [113]. An observational and prospective study found a decrease in mortality in patients who had received an empirical antibiotic therapy including vancomycin [114]. The prevalence of MRSA VAP (15%) supports this result. The literature suggests that inappropriate empirical antibiotic therapy plays a major part in the mortality of patients with MRSA HAP [115]. However, a study found an increase in ICU length-of-stay in patients with MRSA HAP, independently from the adequacy of the empirical treatment [116]. The adequacy of the empirical antibiotic therapy is a critical factor for survival and the length-of-stay [113,117]. A randomized controlled study highlighted an increase in MRSA emergence in patients treated empirically with vancomycin [113]. This encourages highly selecting patients to be treated empirically with an antibiotic directed against MRSA.
In conclusion, only one study conducted in an environment in which the prevalence of MRSA was high (15% of VAP episodes) showed an association between the empirical use of an antibiotic against MRSA and an improvement in HAP outcomes. There is therefore no rationale for the systematic empirical use of an antibiotic against MRSA in France, the prevalence being lower than 3% [118]. However, the consideration of the local ecology is important. Some risk factors encourage including an antibiotic active against MRSA in the empirical antibiotic therapy of HAP, without the possibility to establish an exhaustive list (experts' opinion): high local prevalence of MRSA; recent colonisation by MRSA; chronic skin lesions; chronic renal replacement therapy.
R3.4 -We suggest reducing the spectrum and preferring monotherapy for the antibiotic therapy of HAP after microbiological documentation, including for non-fermenting Gramnegative bacilli.
GRADE 2+, STRONG AGREEMENT Evidence summary and rationale for the recommendation: Both reducing the spectrum and preferring a monotherapy for the antibiotic therapy of HAP after documentation are encompassed by the term ''de-escalation''. De-escalation is possible as soon as the microbiological results are obtained. Nine studies and a meta-analysis address this question. A meta-analysis by Silva et al. did not draw conclusions regard-ing safety and efficacy of de-escalation [119]. Two randomized open controlled studies compared mortality in a ''de-escalation'' compared to a ''control'' group [113,120]. These two studies did not find any significant difference in overall mortality between the two strategies.
Kim et al. found an increase in mortality in the de-escalation group for MRSA documented VAP. Seven observational studies have also compared the de-escalation to the maintenance of the initial antibiotic regimen in terms of mortality. In three prospective studies [114,121,122] and a retrospective study [123], mortality decreased in the de-escalation group. Finally, three retrospective studies [124][125][126][127] did not find any mortality difference between the two strategies. The ICU length-of-stay was prolonged (non-inferiority not obtained) in the de-escalation group of a randomised study [120]. On the contrary, a prospective observational study found an association between a shorter ICU length-of-stay and de-escalation [122].
The risk of relapse associated with de-escalation remained unchanged in two observational studies [126,127] and increased in one randomised [120] and one retrospective study [124]. In a randomised study, de-escalation was associated with an increase in the duration of antibiotic treatment, even though less active antibiotics against Pseudomonas aeruginosa were used in the de-escalation group [120]. However, it is important to note that in this study, VAP represented 48% of included patients [120]. A sub-group analysis on the 56 VAPs did not find differences in these outcomes.
In summary, no data allows to confirm that a de-escalation strategy is associated with increased mortality. It is impossible to conclude concerning the risk of increased length-of-stay, antibiotic consumption and infectious relapse. However, in a strategy of antibiotic stewardship, preservation of large spectrum antibiotics, and preventing the emergence of multidrugresistant bacteria, and in the absence of proven deleterious effects on survival, de-escalation strategy seems reasonable.
R3.5 -We recommend not prolonging for more than 7 days the antibiotic treatment for HAP, including for non-fermenting Gram-negative bacilli, apart from specific situations (immunosuppression, empyema, necrotising or abscessed pneumonia).
GRADE 1À, STRONG AGREEMENT Evidence summary and rationale for the recommendation: There are two meta-analyses comparing treatment durations of 8 days or less to 9 days or more [128,129]. Data are only available in the area of VAP. The first of these meta-analyses relies on six RCTs, including a meeting summary, which was not reported here [130][131][132][133][134]. This meta-analysis included 508 patients. Mortality (28-day, ICU and hospital), length-ofstay and of mechanical ventilation did not change according to treatment duration. Patients in the short-course group had a mean increase in antibiotic-free days at day 28 of 4.02 days; IC 95% = 2.26 to 5.78 days), and a decrease in secondary infectious episodes due to multidrug-resistant bacteria. However, this meta-analysis highlighted a trend towards an increase of the number of relapses and a significant increase in the number of pneumonia relapses in patients with a non-fermenting Gram-negative bacillus infection. The weight of a single multicenter study was essential to explain this result [130]. The increase in relapses was not associated with increased mortality. The second meta-analysis found similar results based on three randomised studies including 883 patients [130,132,133]. However, the increased risk of relapses for VAP linked to non-fermenting Gram-negative bacilli was not confirmed. No medico-economic study compared the costs and efficacy of shorter and longer antibiotic courses.
In summary, there is no benefit to prolonged courses of antibiotic therapy (longer than 7 days) for VAP. A shorter course reduces the exposure to antibiotics and infectious relapses to multidrug-resistant bacteria, albeit with a potential risk of relapse of VAP associated with non-fermenting Gram-negative bacilli. The studies excluded immunosuppressed patients (Human Immunodeficiency Virus, neutropenia, immunosuppressants, corticosteroids > 0.5 mg/kg/d for more than a month, cystic fibrosis) and situations known to require prolonged antibiotic therapy such as empyema, necrotising or abscessed pneumonia [131; 133, 135]. A shorter antibiotic course has yet to be evaluated in these specific patient populations or situations and this recommendation may therefore not be applied to them.
R3.6 -We suggest administering nebulised colimycin (sodium colistimethate) and/or aminoglycosides in documented HAP due multidrug-resistant Gram-negative bacilli documented pneumonia established as sensitive to colimycin and/or aminoglycoside, when no other antibiotics can be used (based on the results of susceptibility testing).
GRADE 2+, STRONG AGREEMENT Evidence summary and rationale for the recommendation: Among the studies published over the last 10 years, four RCTs [135][136][137][138] (with a low to moderate methodological quality) out of seven [135][136][137][138][139][140][141] concluded to the favorable impact of nebulised antibiotics in terms of clinical [135,138] and microbiological [136][137][138] cure in VAP. Following these results, three meta-analyses [70,142,143] (144)). Only one meta-analysis concluded to the superiority of nebulised antibiotics in terms of microbiological response [142]. Two positive studies (in terms of clinical and microbiological cure for the first [138] and microbiological cure for the second [136]) were not included in the meta-analyses.
The benefits in terms of decrease in the emergence of multidrug-resistant bacteria were shown in five studies [136][137][138][139]144], of which four were randomised and controlled [136][137][138][139]. The decrease in the emergence of multidrug-resistant bacteria during treatment was significant in three studies [136][137][138]. The effect on the decrease in the emergence of multirdrug-resistant bacteria is to be taken into account according to the epidemiological context. In spite of thee favorable meta-analyses, it is important to note that: the investigation area is limited to VAP due to Gramnegative bacilli, mainly Enterobacteriaceae or nonfermenting Gram-negative bacilli, implying a treatment with amikacin and/or colimycin treatment. In fact, six of the seven randomised and controlled studies tested the administration of nebulised aminoglycosides. The pathogens responsible for VAP were mainly [137,138] or exclusively [135,136,[139][140][141] Gram-negative bacilli. Three of these seven studies specifically tested the efficacy of nebulised antibiotics in VAP caused by Pseudomonas spp. and Acinetobacter spp [135,139,141]; there is a heterogeneity among the controlled studies included in the three meta-analyses concerning the modalities of nebulisation, dosing of nebulised antibiotics, concomitant systemic antibiotics, and nonconsensual composite criteria for clinical and/or microbiological cure; the meta-analysis of Valachis et al. included seven observational studies and one randomised controlled study, leading to an assessment of its level of methodological quality, varying from weak to very weak, according to the studied judgment criteria [142]; the literature does not support the superiority of an empirical combined therapy containing either of the two antibiotics administered by nebulisation; deleterious effects of nebulised antibiotics have been described, mainly in the most hypoxemic patients [145].
In terms of microbiological eradication of multidrug-resistant Gram-negative bacilli, nebulised antibiotics have proven superior in two controlled studies [137,138] and non-inferior in one observational [144] and one controlled study [141]. In this context, a meta-analysis including six randomized controlled studies and five observational studies concluded to a therapeutic benefit of nebulised compared to systemic antibiotics on mortality (RR: 0.64; CI 95% [0.44-0.94]). A decrease in the risk of nephrotoxicity was also observed using nebulisation (RR: À0.33; CI 95% [À0.54, À0.12]). Consequently, despite the weak methodological quality of this meta-analysis, due to heterogeneity and given: studies assessing the eradication kinetics of bacteria causing VAPs [136][137][138][139]144] and PK/PD parameters [136,140] have shown rapid bactericidal effect, through the high colimycin and/or aminoglycoside concentrations obtained at the alveolar level, with nebulised antibiotics [146][147][148]; experimental data showing that pulmonary penetration of aminoglycosides and colimycin is null or very weak following systemic administration [149][150][151], the benefit-risk balance of nebulised antibiotics is probably favorable for documented VAP due to multidrugresistant Gram-negative bacilli that are susceptible to colimycin and/or aminoglycosides. The specifics of the equipment used (vibrating plate or ultrasonic nebulization chambers, position of the nebulisation chamber in the ventilation circuit, specific formulations of nebulised antibiotics) and the specific modalities of mechanical ventilation during nebulisation require prior training of teams, in order to reproduce the conditions described in the positive studies.
R3.7 -We recommend not administering statins as adjuvant treatment for HAP. GRADE 1À, STRONG AGREEMENT Evidence summary and rationale for the recommendation: Five randomised controlled studies [152][153][154][155][156] included in a meta-analysis (n = 867) [157]) and one observational study (n = 349) [158] were analysed. Among these studies, three controlled studies included patients with HAP outside of the ICU [153,154,156], two controlled studies [152,155] and an observational study [158] included patients admitted to the ICU and requiring mechanical ventilation [155,158] or not [152]. Papazian et al. [155] and Bruyè re et al. [158] included patients with sepsis from all causes, in whom pneumonia represented approximately 50% of patients. No benefits in terms of mortality, ICU or hospital length-of-stay, and in length of mechanical ventilation were demonstrated with statins. Given the overall high methodological quality (high in the controlled studies [152][153][154][155][156] and in the meta-analysis [157]), the importance of the tested judgment criteria, the relative diversity of population (outside ICU, ICU, mechanically-venti-lated or not) and the varied therapeutic regimens (atorvastatin or simvastatin, low or high dose), we obtain a strong recommendation covering a large application range. The sub-group of 27 observational studies of the meta-analysis is the only one to suggest a beneficial effect of statins on mortality [157]. High heterogeneity, publication bias and the limits to the formats of observational studies published over several decades limit the relevance of this result. This effect was therefore not retained as open to interpretation by the experts. Finally, it is important to note the high tolerance of the treatment by statins, regardless of molecule or dose, without any increase in the risk of rhabdomyolysis or hepatic cytolysis.
We did not identify any studies published on the eventual role of corticosteroids administered for HAP in ICU patients. The only study published that has tested anti-lipopolysaccharide monoclonal antibodies as an adjuvant treatment is a posteriori analysis [159] of a previous study of the Phase IIA [160] comparing 17 patients in treatment and 17 patients using placebos. The post-hoc characteristic, the small study sample size, added to the heterogeneous demographic characteristics between groups, and the non-crucial judgment criteria (clinical resolution of pneumonia), do not allow concluding to an effect of this type of treatment. No recommendation was formulated on the adjuvant administration of these adjuvant treatments for HAP in adult ICU patients. The remaining authors declare that they have no competing interest.

Disclosure of interest
Appendix A Protocol 1 suggested by experts: Multimodal healthcare associated pneumonia prevention protocol (EXPERTS' OPINION) Multimodal healthcare associated pneumonias prevention protocol 1 -Favour non-invasive ventilation (NIV) (mainly following digestive surgery and for COPD patients) When invasive ventilation is required: 2 -Apply* a selective digestive decontamination protocol with prophylactic systemic antibiotic treatment < 5 days *If the prevalence of multiresistant bacteria is low (< 20%).
3 -Associate some of the following methods (1st line): favour the use of NIV to prevent intubation; limit dose and duration of sedatives and analgesics associated with mechanical ventilation; initiate early enteral feeding; regularly verify endotracheal tube cuff pressures; perform sub-glottic suction (/6-8 hours) using an appropriate endotracheal tube; favour the orotracheal route for intubation.
NB: The association of head of bed elevation < 30˚and/or oropharyngeal decontamination with 0.12 or 0.2% chlorhexidine could be proposed in association to these measures, despite low efficiency, because they do not cost much and are well-tolerated.
4 -Avoid using the following methods: systematic early tracheotomy (apart from specific indications); anti-ulcer prophylaxis (apart from specific indications); post-pyloric enteral feeding (apart from specific indications); probiotics; systematic early changing of humidifier filters (apart from a recommendation from the manufacturer); closed endotracheal suction systems; the use of intubation tubes lined/coated or incorporating silver or antiseptics, or with an ''optimised'' cuff shape; oropharyngeal decontamination using povidone-iodine; prophylactic nebulized antibiotics; daily skin decontamination using antiseptics.

Protocol 2 suggested by experts: Selective digestive decontamination (EXPERTS' OPINION)
Oro-pharyngeal application (Â 4/day, until tracheal extubation) of a paste or gel containing polymyxin E (2%); tobramycin (2%); amphotericin B (2%).  a Doses are given for information purposes only in patients with normal renal function and standard weight. b Favour the use of aminoglycosides over fluoroquinolones to limit emergence of MDR bacteria. c According to the guidelines' criteria ''Reduce de use of antibiotics in intensive care unit''. d Favour the use of amikacin over gentamicin due to enhanced efficacy against non-fermenting Gram-negative bacilli. e Risk factors for non-fermenting Gram-negative bacilli: antibiotic therapy in the previous 90 days, prior hospital stay of more than 5 days, renal replacement therapy requirement during pneumonia, septic shock, acute respiratory distress syndrome.
f Methicillin-resistant Staphylococcus aureus (MRSA) risk factors: high local prevalence of MRSA, recent colonisation by MRSA, chronic skin lesions, chronic renal replacement therapy.