Year in review 2009: Critical Care - infection

In 2009 Critical Care provided important and clinically relevant research data for management and prevention of infections in critically ill patients. The present review summarises the results of these observational studies and clinical trials and discusses them in the context of the current relevant scientific and clinical background. In particular, we discuss recent epidemiologic data on nosocomial infections in intensive care units, present new approaches to prevention of ventilator-associated pneumonia, describe recent advances in biomarker-guided antibiotic stewardship and attempt to briefly summarise specific challenges related to the management of infections caused by multidrug-resistant microorganisms and influenza A (H1N1).


Introduction
Th e year 2009 was again an interesting one for readers interested in the fi eld of infection in critically ill patients. Several promising new approaches for the prevention of infections in the intensive care unit (ICU) setting were presented. Furthermore, progress was noted in the difficult area of antimicrobial stewardship and risk stratification of infected patients. Finally, several challenges related to infl uenza infections and the management of diffi cult-to-treat infections were tackled or better delineated [1]. Th e present short review will summarise the results of a selection of original studies, with a special focus on articles published in Critical Care in 2009.

Epidemiology of infection in critically ill patients
New insights were reported regarding the epidemiology of infection in ICUs. A global, observational study (EPIC II) on the prevalence and outcomes of infection in 1,265 ICUs was conducted in 75 countries in May 2007. Among the 13,796 patients, 9,084 (66%) patients received an antimicrobial agent and 7,087 (51%) patients were considered infected at the time of data collection [2]. Unfortunately, owing to methodological limitations, no clear-cut distinction could be made between communityassociated and healthcare-associated infec tions. Among those patients who had stayed longer than 7 days in the ICU prior to the study day, however, more than 70% were infected, mostly with multidrug-resistant organisms (MDROs). A clear association was noted between prevalence of infection and hospital mortality, with Greece and Turkey having the highest mortality and Switzerland the lowest [2].
Since this type of prevalence study does not allow one to draw any strong causal inferences between infection rates and excess mortality due to ICU-acquired infections, longitudinal cohort studies with more sophisticated analyses have to be conducted. For instance, a recent French ICU-based case-control study matched 1,725 deceased patients with 1,725 surviving control patients to determine the excess mortality related to ICU-acquired infection [3]. Th e adjusted population-attributable fraction of deaths due to ICU-acquired infection for patients who died before their ICU discharge was 14.6% (95% confi dence interval (CI) = 14.4 to 14.8). Th e attributable mortality of ventilator-associated pneumonia (VAP) was 6.1% (95% CI = 5.7 to 6.5), an estimate close to the 8.1% (95% CI = 3.1 to 13.1%) provided by a multistate model of another cohort study that appropriately handled VAP as a time-dependent event [4].
VAP is a serious complication after major heart surgery in many parts of the world; however, its prevalence and epidemiology varies considerably from hospital to hospital [5,6]. In a recent pan-European cohort study con ducted in 25 hospitals in eight diff erent European coun tries, one or more nosocomial infections were detected in 43 (4.4%) patients. VAP was the most frequent nosocomial infection (2.1%; 13.9 episodes per 1,000 days of mechanical ventilation) [6]. Overall, this rate of VAP is relatively high compared with other surveillance data [7] and warrants further preventive eff orts, as described below.

Prevention of ventilator-associated pneumonia
In many ICUs there is an urgent need to improve adherence to already established infection control measures designed to minimise the risk and rates of VAP. Technology-driven, costly or risky approaches such as

Abstract
In 2009 Critical Care provided important and clinically relevant research data for management and prevention of infections in critically ill patients. The present review summarises the results of these observational studies and clinical trials and discusses them in the context of the current relevant scientifi c and clinical background. In particular, we discuss recent epidemiologic data on nosocomial infections in intensive care units, present new approaches to prevention of ventilatorassociated pneumonia, describe recent advances in biomarker-guided antibiotic stewardship and attempt to briefl y summarise specifi c challenges related to the management of infections caused by multidrugresistant microorganisms and infl uenza A (H1N1). coated endotracheal tubes or selective digestive decontamination should not be implemented as standard of care for all patients [8,9]. Instead, high priority should be given to improving routine hand hygiene, as well as to other routine preventive measures such as backrest elevation >30°, correct cuff -pressure maintenance, avoidance of gastric overdistension and nonessential tracheal suction, and good oral hygiene, which is probably one of the most important and easy-to-perform interventions to successfully prevent VAP [10].
Th e use of chlorhexidine-based oral rinses could be particularly helpful in preventing endogenous and exogenous contamination of patients' upper and lower airways by decreasing the bacterial load present in the oropharyngeal fl ora [11]. Scannapieco and colleagues conducted a randomised, double-blind, placebocontrolled clinical trial of chlorhexidine gluconate on oral bacterial pathogens in mechanically ventilated patients [12]. While 175 subjects were randomised, full follow-up assessment after at least 48 hours of ICU stay was only available for 115 patients. Chlorhexidine reduced the number of Staphylococcus aureus, but not the total number of Enterobacteriacae, Pseudomonas spp. or Acinetobacter spp. in the dental plaque of included subjects. A nonsignifi cant reduction in VAP rates was noted in groups treated with chlorhexidine compared with the placebo group (odds ratio = 0.54, 95% CI = 0.23 to 1.25). A similar study conducted in Spain investigating the eff ectiveness of oral rinses with chlorhexidine in preventing nosocomial respiratory tract infections among ICU patients also failed to demonstrate a signifi cant eff ect [13]. It remains to be elucidated whether the limited power or other methodological issues related to these studies could explain the negative study results [14,15].

Chlorhexidine-based infection control measures
Several recently published high-quality studies have highlighted the potential benefi t of using chlorhexidine for the prevention of catheter-related bloodstream infections. A prospective randomised trial was performed in seven ICUs of fi ve French hospitals to assess the eff ect of two preventive practices on catheter-related bloodstream infection rates: frequency of dressing change (3 days vs. 7 days) and type of dressing (standard vs. chlorhexidine-impregnated sponges) [16]. Th e use of chlorhexidine-impregnated sponges decreased the rate of catheter-related bloodstream infection from an already low level of 1.3 to 0.4 episodes per 1,000 catheter-days without an increase in chlorhexidine-resistant microorganisms. Changing catheter dressings every 7 days was not inferior to changing dressings every 3 days in terms of rate of colonisation [16]. Two studies conducted in the USA suggested that routine chlorhexidine body washes may also help to reduce catheter-related bloodstream infection rates in diff erent settings [17,18].
Chlorhexidine body washes have now become the standard of care in many ICUs to reduce the bacterial load on patients' skin. A British team of investigators examined the impact of several control interventions aimed at reducing cross-transmission of methicillinresistant S. aureus [19]. An educational campaign and cohorting had little impact on methicillin-resistant S. aureus transmission. Th e introduction of chlorhexidine as a skin antiseptic reduced methicillin-resistant S. aureus transmission of all but one of the strains prevalent in this ICU: the TW strain that carries the qacA/B genes that code for chlorhexidine resistance [19]. Owing to its chlorhexidine resistance, the acquisition of this methicillin-resistant S. aureus strain increased dramati cally during the period of this interrupted timeseries study. Th e emergence of resistance has also been ob served with other topical decontamination regimens; it is therefore important to actively look for emerging chlorhexidine resistance in settings with widespread chlorhexidine usage [20].

Management of severe and diffi cult-to-treat infections
Treatment of VAP caused by MDROs has been limited by the poor diff usion of certain intravenous antibiotics (for example, aminoglycosides) into the alveolar compartment of the lungs. An elegant solution to this challenge could consist of the aerosolisation of antibiotic agents with special methods and devices [21]. In a recent pilot study, French investigators showed that a new mode of delivery of aerosolised amikacin achieved very high drug concentrations in the lung, while maintaining safe serum levels in 28 mechanically ventilated patients with Gramnegative VAP treated for 7 to 14 days, adjunctive to intravenous therapy [22]. Despite these recent promising fi ndings, the widespread use of aerosolised antibiotics to treat VAP cannot be recommended at present and should be restricted to the treatment of multidrug-resistant Gram-negative VAP, as pointed out by the same group of investigators in a recent review [21].
Th e management of postoperative peritonitis caused by MDROs may also represent a clinical challenge [23,24]. Augustin and colleagues determined risk factors for the presence of MDROs in postoperative peritonitis in 100 patients, as well as optimal empirical antibiotic therapy choices among diff erent, commonly suggested treatment options [25]. Adequate empirical therapy was achieved in only 64% of cases. Adequacy decreased signifi cantly in patients with MDROs, as compared with patients presenting other bacteria (39% vs. 81%, P <0.0001). However, as also observed in another recent article on staphylococcal bacteremia [26], mortality in the study by Augustin and colleagues did not diff er between patients who received adequate empiric therapy and those who did not (30% vs. 31%), or between patients with peritonitis caused by MDROs and other bacteria (29% for MDRO group vs. 35% for others). Importantly, the defi nition of adequacy in this study was based purely on microbiological criteria and did not take yeasts into account. Th e single antibiotics providing the best activity rate were imipenem/cilastatin and piperacillin/tazobactam. Th e best adequacy for empiric therapy was obtained by combinations of imipenem/cilastatin or piperacillin/tazobactam, amikacin and a glycopeptide [25]. Th is fi nding is in line with two recent studies from 2010 on the use of antibiotic combinations. Both studies recommend antibiotic combination therapy over monotherapy for the initial empiric treatment phase of the most severely ill patients with septic shock [27,28].
Antifungal therapy has been revolutionised within the past 10 years. New treatment options and indications have continuously entered critical care and have increased the competition and marketing pressure. In this overheated area of medicine with continuous infl ux of new products and industry-sponsored clinical studies [29], it remains rather diffi cult for the nonexpert critical care physician to evaluate true progress and the eff ectiveness of diff erent antifungal agents in daily clinical practice, including the toxicity profi le of older agents [30].
Marriott and colleagues [31] undertook a nationwide prospective clinical and microbiological cohort study of all episodes of ICU-acquired candidaemia occurring in non-neutropenic adults in Australian ICUs between 2001 and 2004 [32]. Overall, 183 patients had ICU-acquired candidaemia with a 30-day case-fatality rate of 56%. Host factors (older age, mechanical venti lation and ICU admission diagnosis) and failure to receive systemic antifungal therapy were signifi cantly associated with mortality on multivariate analysis. Process of care measures advocated in recent guidelines were implemented inconsistently: follow-up blood cultures were obtained in 68% of patients, central venous catheters were removed within 5 days in 80% of patients and ophthalmological examination was performed in 36% of patients. Th is study showed that crude mortality remains high in Australian ICU patients with candi daemia. Among those who were treated, mortality was overwhelmingly related to host factors but not treatment variables (the time to initiation of anti fungals or fl uconazole pharmacokinetic and pharmaco dynamic factors) [31].
Zilberberg and colleagues investigated the costeff ective ness of a new echinocandin antifungal agent (micafungin) as an alternative to fl uconazole in the empirical treatment of suspected ICU-acquired candidaemia among septic patients in a simulation model [33].
In the base case analysis, the authors assumed a high attributable mortality of ICU-acquired candidaemia (40%) and an overly optimistic risk reduction (52%) in mortality with appropriate timely therapy. Of note, in the Australian cohort study cited above, antifungal therapy was commonly started among treated patients >48 hours after drawing the fi rst positive blood culture; this delay was not associated with increased mortality [31]. Moreover, the model assumptions were mainly based on the North-American epidemiology of azole-resistant Th is cost-eff ectiveness analysis has severe limitations, since the methodology used is defi cient both in terms of the modelling strategy as well as the reliability of the probability estimates. Th e authors used an oversimplifi ed approach and, sometimes, questionable probability estimates, result ing in biasing their analysis in favour of the intervention (providing empiric anti-Candida therapy) and in favour of micafungin versus fl uconazole. Although empiric micafungin may well be an attractive treatment strategy, the defi ciencies in this analysis preclude its widespread use. Th is study therefore should only represent the starting point for further investigations of the cost-eff ectiveness of diff erent treatment strategies of suspected and confi rmed fungal infections in the critical care setting.

Antibiotic stewardship and risk prediction
At the current time, procalcitonin (PCT) represents the best studied biomarker for guiding antibiotic treatment duration in the hospital setting [34,35]. Several highquality clinical trials investigating the diagnostic performance and clinical eff ectiveness of PCT have been published within the past 3 years [36][37][38][39]. Two large-scale studies confi rmed the potential usefulness of PCT to guide antibiotic use in critically ill patients [37,39]. Nevertheless, in the study by Bouadma and colleagues more than one-half (53%) of patients enrolled in the PCT-guided arm did not follow the protocol for initial antibiotic treatment decisions -and thus antimicrobial use was not completely determined by PCT levels, as recommended [39]. PCT in critically ill patients therefore probably remains a suboptimal marker to strongly infl uence initial treatment decisions or even to withhold empiric therapy for potentially life-threatening infections. PCT measure ments may, however, increase the confi dence of clinicians to withdraw antimicrobial therapy at an earlier timepoint in the majority of patients.
To further clarify the kinetics of PCT within the fi rst days of sepsis in relation to adequacy of antibiotic therapy, Charles and colleagues conducted an obser vational cohort study in 180 septic patients [40]. Appro priate initial antibiotic therapy was associated with a signifi cantly greater decrease in PCT until day 3. Th e baseline PCT level failed to predict outcome, but on day 3 higher PCT levels were measured in the nonsurvivors when compared with the survivors. Th is is the fi rst study to demonstrate that the PCT dynamics within 72 hours after onset of sepsis may be correlated both with appropriateness of the empirical antibiotic therapy and with overall survival. Whether this interesting obser vation can be incorporated into clinical management guidelines needs to be further evaluated.
Another marker of infl ammation, C-reactive protein remains widely used throughout the world for diagnosis of infectious conditions -despite its rather limited diagnostic accuracy when used as a single measurement in time [41]. Paran and colleagues therefore investigated the dynamic nature of C-reactive protein in a cohort of patients admitted to an emergency department in Israel [42]. Th ey constructed a new index, C-reactive protein velocity, which was defi ned as the ratio of C-reactive protein on admission to the number of hours since the onset of fever. Th e C-reactive protein velocity improved diff eren tiation between febrile bacterial infections and non bacterial febrile illnesses compared with C-reactive protein alone. If confi rmed by other groups, this approach could provide clinicians with a valuable tool for estab lishing the correct diagnosis and better identifying individuals who need prompt therapeutic interventions [42].

Community-acquired pneumonia risk stratifi cation
Th e severity of community-acquired pneumonia may be diffi cult to judge clinically. As a consequence, multiple scores have been proposed with the aim of predicting the risk of adverse outcomes in critically ill patients [43][44][45]. None of the existing rules is ideal; weaknesses include low sensitivity or specifi city, excessive complexity, underestimation of severity in younger patients, and poor prediction of ICU admission.
In view of both the high cost and potential benefi t of critical care, there is a need for tools that help ensure timely ICU admission for all patients with pneumonia for whom this is likely to improve outcome. Th e REA-ICU index developed by Renaud and colleagues aims to preemptively identify patients at risk of requiring secondary transfer to ICU within the fi rst 3 days of their hospital admission [46]. Th e prediction rule was derived from a cohort of 4,593 patients initially presenting without overt circulatory or respiratory failure and was based on 11 criteria. Nursing home residents were excluded. Th e highest risk class was assigned to 3.6% of evaluated patients; among this group, the rate of ICU transfer within 3 days of admission was around 30%.
Do we need yet another community-acquired pneumonia severity score? Th e merit of the study by Renaud and colleagues is its focus on patients who are at high risk despite not being obvious ICU candidates on admission.
Th e REA-ICU index may not, however, constitute a major advance in the overall endeavour of identifying those patients who will or should benefi t from critical care [47]. Compared with existing prediction rules, the REA-ICU index is neither less complex nor does it appear to be clearly superior in guiding patient management (Table 1). A head-to-head validation of the existing scores in a prospective study with separation of evaluators and clinical decision-makers would be desirable to better judge their utility in clinical practice.

H1N1 infl uenza A
Th e infl uenza A (H1N1) pandemic was certainly the most featured infectious disease in 2009. Several highly accessed contributions were published in Critical Care during this year. Rello and Pop-Vicas highlighted the clinical challenges associated with primary infl uenza pneumonia [48]. Infl uenza A (H1N1) illness severity and the case-fatality rate were described in an interesting case series of 32 relatively young patients (median, 36 years) hospitalised in Spain between 23 June and 31 July 2009 [49]. Twenty-four patients (75%) developed multiorgan dysfunction, and eight patients died. As confi rmed by later cohort studies from Australia and the UK [50,51], pulmonary compli cations of infl uenza A (H1N1) infection in pregnant and young obese but previously healthy persons were associated with adverse health outcomes. Th e same Spanish group investigated the host immune response following infection with infl uenza A (H1N1) [52]. Interestingly, severe H1N1 disease with respiratory involvement was characterised by early secretion of specifi c cytokines usually associated with cell-mediated immunity but also commonly linked to the pathogenesis of infl ammatory diseases.

Conclusions
Infection remains one of the key challenges of critical care and signifi cantly contributes to morbidity and mortality. Papers published in recent months remind us that further reductions of nosocomial infection rates are possible -often with the help of simple interventions. Antimicrobial resistance is a permanent threat for ICU patients and there is growing awareness that available antimicrobial agents should be used wisely. Biomarkers of infection can help to make more appropriate treatment decisions. Th e rapid proliferation of published research data entails a need for consolidation of existing knowledge as exemplifi ed by the growing number of community-acquired pneumonia severity scores. Clearly, infections in the ICU continue to be an exciting and important topic for ongoing research.