Clinical review: Probiotics in critical care

Patients in ICUs represent a relatively small subgroup of hospitalised patients, but they account for approximately 25% of all hospital infections. Approximately 30% of ICU patients suffer from infection as a complication of critical illness, which increases the length of ICU stay, morbidity, mortality and cost. Gram-negative bacteria are the predominant cause of ICU-related infections and with the rise in multidrug-resistant strains we should focus our attention on nonantibiotic strategies in the prevention and treatment of nosocomial infections. Probiotics have been proposed as one option in this quest; however, mechanisms of action in the critically ill population require further investigation. Some of the beneficial effects appear to be associated with improvement in gastrointestinal barrier function, restoration of normal intestinal permeability and motility, modification of the balance of intestinal microbiota and immunomodulation. However, the information we have to date on the use of probiotics in the critical care setting is difficult to interpret due to small sample sizes, differences in ICU populations, the variety of probiotic combinations studied and differences in administration techniques. In this review we shall examine the use of probiotics in the critical care setting, look at some of the proposed mechanisms of action and discuss their potential benefits and drawbacks.


Introduction
During an episode of critical illness a number of signifi cant changes occur in the microbiota of the human gut. Th ese changes occur due to alterations in the stress hormone profi le, impairment of blood supply to the gut, immunosuppression, antibiotic use and nutrient defi ciency [1]. In experimental models these changes have been shown to occur within 6 to 8 hours, with endogenous Lactobacillus strains being replaced by pathogenic bacteria [2]. Th is change can lead to a breakdown in the intestinal barrier function that is likely to play a signifi cant role in the pathogenesis of multiple organ dysfunction syndrome [3,4].
Redressing this balance and exploiting the benefi cial eff ects of probiotic bacteria is understandably an area of considerable interest. However, the mechanisms by which these microorganisms exert their eff ects are various and depend upon the dose used, the route(s) of administration and the dosing frequency [5]. Furthermore, a number of these eff ects are strain specifi c.

Probiotics, prebiotics and synbiotics
Probiotics are defi ned as 'live microorganisms that confer a health benefi t on the host when administered in adequate amounts' [6]. Prebiotics are nondigestible food components that stimulate the growth and/or activity of bacteria in the digestive tract in ways that may be benefi cial to health [7]. Synbiotics are a combination of probiotics and prebiotics. Th ere has been an explosion of interest in probiotics and their potential health benefi ts since 2000, with initial attention focusing on the gastrointestinal tract.

Probiotics and the gastrointestinal tract
Th e human intestine is home to hundreds of species of bacteria, archaea and eukarya, many of which are nonculturable but can now be identifi ed using metagenomic approaches. Th e bacterial load tends to be highest in the large intestine (up to 10 11 colony-forming units/g), and while the healthy human gut is dominated by Bacteroides, Firmicutes and Actino bacteria, each individual has their own distinct stool bacterial composition determined by environmental and genetic factors. Th is bacterial profi le remains relatively constant over time unless altered by disease state or antibacterial treatment [8,9].
Culture-based and molecular detection methods have demonstrated that it is possible to signifi cantly alter the composition of gut fl ora in adults and infants by treatment with probiotics. Sepp and colleagues treated 15 neonates with Lactobacillus rhamnosus GG for the fi rst Abstract Patients in ICUs represent a relatively small subgroup of hospitalised patients, but they account for approximately 25% of all hospital infections. Approximately 30% of ICU patients suff er from infection as a complication of critical illness, which increases the length of ICU stay, morbidity, mortality and cost. Gram-negative bacteria are the predominant cause of ICU-related infections and with the rise in multidrug-resistant strains we should focus our attention on nonantibiotic strategies in the prevention and treatment of nosocomial infections. Probiotics have been proposed as one option in this quest; however, mechanisms of action in the critically ill population require further investigation. Some of the benefi cial eff ects appear to be associated with improvement in gastrointestinal barrier function, restoration of normal intestinal permeability and motility, modifi cation of the balance of intestinal microbiota and immunomodulation. However, the information we have to date on the use of probiotics in the critical care setting is diffi cult to interpret due to small sample sizes, diff erences in ICU populations, the variety of probiotic combinations studied and diff erences in administration techniques. In this review we shall examine the use of probiotics in the critical care setting, look at some of the proposed mechanisms of action and discuss their potential benefi ts and drawbacks.
2 weeks of life. Th ey found that L. rhamnosus GG persisted for 1 month in eight of these neonates. Th ere were also signifi cant diff erences in the bacterial composition of the stool compared with the control group, with increased numbers of coliforms, lacto bacilli and Bifi dobacterium spp. [10]. Benno and colleagues demonstrated a statistically signifi cant increase in bifi dobacteria in adults treated with L. rhamnosus GG for a 4-week period. Th ey also found an increase in lactobacilli and a decrease in the proportion of Clostridium spp. [11]. As these techniques are based on faecal profi ling, they tend to refl ect the large bowel bacterial composition with little information being available on the small bowel eff ects of probiotics.

Mechanisms of action of probiotics
Much of the information available on the mechanisms of action of probiotics is obtained from animal work and in vitro studies; hence we must be careful in extrapolating this to humans. What is clear, however, is that there are multiple mechanisms by which diff erent probiotic bacteria exert their eff ects, and these eff ects may vary with the strain and population studied. Table 1 summarises the main mechanisms by which probiotics exert their eff ects, and Table 2 presents details of commonly used probiotic preparations.
Probiotics may alter the local environment within the lumen of the gut, producing antimicrobial eff ects on pathogenic organisms. Lactic acid-producing and acetic acid-producing probiotics reduce the luminal pH resulting in an unfavourable milieu for pathogens. Th is has been demonstrated in vitro with pathogen growth being reduced in a pH-dependent manner by Lacto bacillus spp. [12]. Venturi and colleagues demonstrated a signifi cant reduction of luminal pH in vivo in ulcerative colitis patients treated with the probiotic mixture VSL#3 [13].
Probiotics also exert a direct antimicrobial eff ect via the production of bacteriocins. Bacteriocins are proteins produced by bacteria that inhibit the growth and virulence of other pathogenic bacteria. Probiotic bacteria defi cient in the bacteriocin gene are less eff ective probiotics, as demonstrated in a murine model where a mutant form of Lactobacillus salivarius UCC118 failed to protect against infection with Listeria monocytogenes [14]. A wide variety of bacteriocins is recognised, and their spectrum of action ranges from antagonism of similar bacterial strains to the inhibition of a wide range of Gram-positives, Gram-negatives, yeasts and moulds [15]. One such example of a broad-spectrum bacteriocin is that produced by a subspecies of L. salivarius. Th e ABP-118 bacteriocin inhibits Bacillus, Staphylococcus, Enterococcus, Listeria and Salmonella spp. [16].
Bacteria communicate with each other using a mechanism known as quorum sensing. Th is involves the produc tion and secretion of signalling molecules known as autoinducers. In their in vitro study, Medellin-Peña and colleagues demonstrated that Lactobacillus acidophilus La-5 secretes molecules that disrupt this interbacterial communication, reducing expression of virulence-related genes by Escherichia coli O157:H7 [17].
Probiotics have also been demonstrated to enhance intestinal barrier function. Intestinal barrier function is complex and its control involves cellular stability at a cytoskeletal and tight junction level, as well as mucus, chloride and water secretion. Probiotics have been shown to exert an eff ect, in vitro and in vivo, via these mechanisms [15]. For example, Lactobacillus plantarum 299v can enhance mucus production and secretion in human intestinal epithelial cells [18]. Th e probiotic strain E. coli Nissle 1917 appears to enhance mucosal barrier function by production of human β-defensin 2 [15]. E. coli Nissle has also has been demonstrated in vitro to reduce adhesion and invasion of intestinal epithelial cells by an entero invasive E. coli.
In addition, by competing with pathogens for nutrients and adhesion in a microbiological niche, probiotics can prevent replication by pathogens, a phenomenon known as colonisation resistance [5]. Probiotics can thus promote the integrity of the gut defence barrier and create an unfavourable environment for pathogen colonisation.
Probiotics can also exert a range of immunological eff ects. Th e interaction between the luminal bacteria and the underlying epithelial and mucosal lymphoid cells is referred to as bacterial-epithelial cross-talk. Th is crosstalk enables probiotics to have an eff ect on both the innate and adaptive host immune system [19] -for example, promotion of B cells into plasma cells, increased production of secretory immunoglobulin A and prevention of activation of the proinfl ammatory nuclear transcrip tion factor NF-κB [5]. Other immunologic mechanisms include altera tion of the cytokine profi le and activation of macrophages to present antigen to B lympho cytes and increase immuno globulin production [20].

Probiotics in the critically ill
Th e effi ciency of intestinal barrier function is demonstrated by the fact that the faecal bacterial concentration approaches 10 12 organisms/ml in the caecum, while tissues one cell deep to the intact intestinal mucosa are usually sterile [21]. Any signifi cant insult to the gut or alteration to its microbiota is likely to play a role in promoting systemic infl ammation and infection in the critically ill population [22]. In contrast to the large bowel, the stomach, duodenum and jejunum have a relative paucity of bacteria (10 3 to 10 4 organisms/ml). Th e presence of enteric organisms in gastric aspirates is therefore abnormal and represents gastric colonisation. In the context of critical illness, this colonisation is the result of bacterial overgrowth in the proximal gastro intestinal tract [21].
Colonisation of the stomach by pathogens or potential pathogens is believed to occur due to a combination of poor gut motility, increased gastric pH (due to acid suppression) and the use of broad-spectrum antibiotics. Th is combination of factors leads to an overgrowth of bacteria in the duodenum, which refl ux into the stomach and are ultimately regurgitated and aspirated into the lungs [23].
Th e normal intestinal microbiota of critically ill patients is altered and replaced by pathogens for a number of reasons. Th erefore, it would seem logical to consider that probiotics may have a role in reducing intestinal

Mechanism of action Specifi c probiotic examples
Luminal pH modifi cation Production of lactic acid and acetic acid reduces luminal pH resulting in unfavourable milieu for pathogens Lactobacillus spp.: pH-dependent reduction in pathogen growth [12] VSL#3: in vivo luminal pH reduction in ulcerative colitis patients [13] Bacteriocin production Bacteriocins are proteins produced by bacteria that inhibit the growth and virulence of other microorganisms. The may be narrow spectrum (inhibit related bacterial strains) or broad spectrum (inhibit a wide range of bacteria, yeasts and moulds) [15] Mutant Lactobacillus salivarius defi cient in bacteriocin gene are unable to protect mice against Listeria monocytogenes infection [14] L. salivarius subspecies produce broad-spectrum bacteriocins [16] Disruption of interbacterial communication Autoinducers are the signalling molecules produced and secreted by bacteria that form the basis of quorum sensing (bacterial communication) Lactobacillus acidophilus La-5 disrupts quorum sensing and expression of virulence-related genes by Escherichia coli O157:H7 [17] Enhanced mucosal barrier function Increased intestinal epithelial cell mucus production and secretion

Reduced adhesion and invasion of intestinal epithelial cells by enteroinvasive bacteria resulting in reduced translocation
Increased production of human β-defensin 2 by epithelial cells

Stabilisation of intracellular tight junctions and reduced chloride/water secretion Epithelial cell regeneration and reduced apoptosis
Lactobacillus plantarum 299v: increased mucin gene expression in vitro [18] and adherence to colonic cells via a mannose-specifi c adherence mechanism [74] Lactobacillus casei rhamnosus adheres to colonic cells in vitro [75] E. coli Nissle 1917: increase in mucin gene expression [76] and production of human β-defensin 2 by colonic cells [77] Streptococcus thermophiles and L. acidophilus reduce water and chloride secretion in response to pathogenic bacteria [78,79] Lactobacillus pretreatment of intestinal epithelium reduces disruption of epithelial tight junctions by pathogenic E. coli [80]. Probiotic preparation VSL#3 (see Table 2) prevents redistribution of epithelial tight junction proteins on exposure to pathogenic bacteria [76]. Lactobacillus rhamnosus GG prevents cytokine-mediated apoptosis of intestinal epithelial cells [81]. Lactobacillus casei and Clostridium butyricum both stimulate gut epithelial proliferation in rats [82] Colonisation resistance The probiotic competes with pathogen for nutrients and adhesion in a microbiological niche [5] L. casei rhamnosus adheres to colonic cells, reduces pathogenic bacterial growth and can persist within the gastrointestinal tract [75,83] E. coli Nissle 1917 inhibits growth of Shiga-toxin producing E. coli [84] Immunological eff ects Bacterial-epithelial cross-talk enables luminal probiotic organisms to infl uence gut-associated lymphoid tissue and innate and adaptive host responses [19,85]. Toll-like receptors play a central role in mediating this process [86] Increased promotion of B cells to plasma cells and increased production of immunoglobulins [5] Activation and modulation of macrophages, T cells and natural killer cells VSL#3 has been associated with increased anti-infl ammatory and reduced proinfl ammatory cytokine activity, reduced inducible nitric oxide synthase and matrix metalloproteinase activity in patients with pouchitis [87]. L. plantarum 299v increases IL-10 secretion from macrophages and T cells in patients with ulcerative colitis [88]. L. casei and Lactobacillus bulgaricus signifi cantly reduce TNFα release from infl amed mucosa in Crohn's disease [89]. E. coli Nissle 1917 shows local and systemic anti-infl ammatory eff ects in a murine model of lipopolysaccharide-induced sepsis [90] L. rhamnosus GG: increased circulating IgA, IgG and IgM concentrations in children with gastroenteritis [91,92]. Pretreatment with probiotic prior to typhoid vaccination leads to increased anti-typhoid antibody titres [93] L. casei Shirota: cell wall structure potently induces IL-12 production and the probiotic diff erentially controls the infl ammatory cytokine responses of macrophages, T cells and natural killer cells [30,94,95]. L. casei Shirota and Bifi dobacterium breve administered preoperatively to biliary cancer patients signifi cantly reduce postoperative IL-6, C-reactive protein and white cell count concentrations [30]. L. acidophilus and Bifi dobacterium longum increased macrophage phagocytic activity in a murine model [96] colonisation by pathogens and thus in the prevention of infection and sepsis syndromes in this population.

Probiotics in the prevention of nonrespiratory infection
Probiotics have been studied in the prevention of postoperative infection. Th ree studies in patients undergoing major colorectal surgery have shown no signifi cant reduc tion in postoperative infection rates [24][25][26]. In each study, however, the eff ectiveness may have been limited by a relatively short postoperative period of probiotic administration (4 to 5 days). In contrast, several studies in patients undergoing pancreatic resection [27,28] and hepatic resection [29,30] have shown signifi cant reduc tions in postoperative infection rates of up to 30%. Th ese patients received probiotic for 8 to 14 days post operatively. Liver transplant patients have multiple risk factors for infection, including profound immunosuppression. Two randomised trials have shown probiotics to be safe and eff ective in this group of patients. In the fi rst study 95 patients were randomised to receive standard enteral feed plus selective bowel decontamination, fi bre-containing enteral feed plus live L. plantarum 299 (Lp299) or fi bre-containing enteral feed plus heat-killed Lp299 [31]. Th e live Lp299 group developed signifi cantly fewer infections than the other two groups (48% vs. 13% vs. 34%, respectively). In addition, the mean duration of antibiotic therapy, the mean total hospital stay and the length of ICU stay were also shorter than in the groups with inactivated Lp299 and selective bowel decontamination. However, these diff erences did not reach statistical signifi cance. Th e second study compared only Synbiotic 2000 and prebiotic fi bre, reporting postoperative infection rates of 3% and 48%, respectively [32]. No serious side eff ects or infec tions caused by the probiotics were noted in either study. Oláh and colleagues randomised 45 patients with severe acute pancreatitis to receive enteral oat fi bre and live Lp299 or enteral oat fi bre and heat-killed Lp299 [33]. In the group treated with the live probiotic, only one patient required surgery for a septic complication involving the pancreas, compared with seven such compli cations in the control group (P = 0.02). Th ere was also a nonsignifi cant trend toward a shorter length of hospital stay (13.7 days vs. 21.4 days, respectively). Th e same group carried out a single-centre, double-blind, randomised placebo-controlled trial using Synbiotic 2000 in a further 62 patients with severe acute pancreatitis [34]. Th is trial showed no statistically signifi cant diff erences in the incidence of mortality, septic complications or develop ment of multiorgan failure between the two groups. However, the total incidence of systemic infl amma tory response syndrome, multiple organ failure and rate of complications was signifi cantly less in the treatment group versus the control group (8 vs. 14, P <0.05 and P <0.05, respectively).
Th e trial that has raised most concern with regard to adverse outcomes and the use of probiotics is the PROPATRIA trial [35]. In this multicentre, placebocontrolled trial, 296 patients with predicted severe acute pancreatitis were randomised to receive the synbiotic preparation Ecologic 641 or placebo. Th is was administered together with fi bre-enriched enteral feed via the nasojejunal route for 28 days. Th e rate of infectious complications was similar in both groups (30% vs. 28%) but the mortality rate was higher in the synbiotic group. Nine patients in the synbiotic group developed bowel ischaemia, eight of these being small bowel ischaemia. Th ere were no cases of bowel ischaemia in the placebo group. One possible explanation for this outcome is a diff erence in the two groups, with more patients in the synbiotic group having established organ failure at the time treatment began. Another theory is that such a signifi cant intestinal burden of bacteria and high-fi bre feed could result in increased oxygen consumption and local bowel ischaemia. Nevertheless, this is the fi rst time such a complication has been reported.

Probiotics in the prevention of respiratory infection
Th e respiratory tract is consistently the most common site of nosocomial infection, accounting for 65% of ICU-acquired infections [36]. Ventilator-associated pneu monia (VAP) complicates the care of up to 30% of patients receiving mechanical ventilation, accounting for 50 to 60% of total antibiotic days [37][38][39][40]. Patients with VAP present increased morbidity and mortality, prolonged ICU and hospital lengths of stay, and increased costs [41].
Current VAP prevention strategies aim to reduce colonisation of the oropharynx and upper gastrointestinal tract with pathogenic bacteria and prevent their subsequent aspiration. Th ese measures include elevation of the head of the bed, silver-coated tracheal tubes, oral care, subglottic secretion drainage and use of sedation breaks and weaning protocols. Selective digestive tract decontami nation using antibiotics in the oral cavity or whole gastrointestinal tract decontamination have been shown to reduce rates of VAP and mortality [42,43]. However, these strategies have not gained widespread favour in critical care owing to concerns about promoting antibiotic resistance. Oostdijk's group demonstrated a statistically signifi cant increase in intestinal colonisation with Gram-negative bacteria resistant to ceftazidime, tobra mycin or cipro fl oxacin (P <0.05) [44]. Th ese concerns were also borne out by a large-cluster, randomised cross-over study of selective decontami nation of the digestive tract that showed a marked increase resistance to ceftazidime in faecal Entero bacteriaceae, together with a small but signifi cant increase in bacterial resistance from the respiratory tract [45]. In a previous study, the use of cefotaxime as part of selective decontamination of the digestive tract regime was found to select for an outbreak of extended-spectrum β-lactamase-producing E. coli and Klebsiella pneumonia [46].
To date there have been eight randomised controlled trials of probiotic therapy as a strategy to prevent VAP [38,[47][48][49][50][51][52][53]. Th e inclusion criteria, sample size (range 50 to 348), populations studied and diagnostic criteria for VAP varied between studies. Th e probiotic formula, dosing and route of administration also varied but all trials contained Lactobacillus spp. (see Table 3). Six of the eight trials showed a lower incidence of VAP in the probiotic group [38,47,48,[50][51][52], but this diff erence was statistically signifi cant in only three of the studies [38,47,48]. Interestingly, one study used chlorhexidine oral disinfection as a control and found that probiotic Lp299 was at least as eff ective in preventing oropharyngeal colonisation (61.9% vs. 34.8% new colonisation, respect ively; P = 0.13) [50]. Th e study by Forestier and colleagues found no diff erence in incidence of VAP between groups but did demonstrate a median delay in respiratory colonisation with Pseudomonas aeruginosa of 50 days versus 11 days in controls (P = 0.01) [49]. Th is is the most commonly isolated antibiotic-resistant Gramnegative species in VAP [39].  Confl icting results also arise from meta-analyses of probiotics in critical care. Th e work by Watkinson and colleagues in 2007 analysed the use of prebiotics, probiotics and synbiotics in 999 adult critical care patients from eight randomised controlled trials and concluded that there was no benefi t in the probiotic prophylaxis of VAP [54]. In 2010, however, Siempos and colleagues examined fi ve randomised controlled trials (689 patients) and showed that probiotic administration was associated with a lower incidence of VAP when compared with standard care (odds ratio = 0.61; 95% confi dence interval = 0.31 to 0.91) [55]. Importantly, both of these were published before the studies by Morrow and colleagues [38], Oudhuis and colleagues [52] and Barraud and colleagues [52].
Th e trial by Morrow and colleagues is unique in that it included oropharyngeal slurry as one of the routes of administration for the probiotic [38]. Th e research group randomised 146 ventilated patients who were considered at high risk for VAP to receive probiotic L. rhamnosus GG or placebo (inulin) within 24 hours of intubation until extubation, tracheostomy or death. Th e primary outcome was microbiologically confi rmed VAP based on quantitative culture of distal airway samples obtained by bronchoscopy. Th e incidence of VAP was signifi cantly reduced in the probiotic group (19.1% with probiotic vs. 40.0% with placebo, P = 0.007).
Morrow and colleagues also examined the incidence of Clostridium diffi cile and ICU-associated diarrhoea in their patients. Th e probiotic group had signifi cantly less C. diffi cile cytotoxin-positive diarrhoea compared with the placebo group (5.6% vs. 18.6%, P = 0.02), although the duration of diarrhoea was not signifi cantly lower. However, patients treated with probiotic received fewer days of antibiotics for C. diffi cile-associated diarrhoea (0.5 ± 2.3 days vs. 2.1 ± 4.8 days in placebo group, P = 0.02). Th e duration of ICU-associated diarrhoea was also signifi cantly reduced in the probiotic group (4.1 ± 3.7 days vs. 5.9 ± 3.8 days in placebo group, P = 0.03).
To date, studies of probiotics in the critically ill have trialled a number of species, Lactobacillus featuring frequently. Currently unknown, however, is whether one species is superior in the prevention of infection associated with critical illness. Similarly, the optimal administration route, dosage and duration of treatment are not clear. Further research is undoubtedly warranted, perhaps considering Gram-negative probiotic species.

Administration of probiotics and monitoring of their eff ects
Probiotics are commercially available in various preparations including yoghurt-based products, capsules, powders and suspensions. Th e studies in critically ill patients discussed above involve enteral administration of a variety of probiotic strains using diff erent dosing regimes.
In eight of the nine studies involving mechanically ventilated patients (Table 3), probiotic powder or capsule contents were dissolved in water and delivered via a feeding tube into the stomach. Morrow and colleagues used an oropharyngeal slurry of L. rhamnosus GG (suspended in a sterile water-based surgical lubricant) in addition to nasogastric administration [38]. After 72 hours, the patients receiving this regime were found to have lower rates of oral (38.2% vs. 70%, P = 0.001) and gastric (32.3% vs. 45.7%, P = 0.03) colonisation with pathogenic species than those receiving placebo. Klarin and colleagues used topical application of Lp299 to the oral cavity alone and found it to be at least as eff ective as chlorhexidine 0.1% in reducing oropharyngeal pathogenic load [50].
Testing for colonisation of the gastrointestinal tract with the probiotic species is reported in only a minority of studies. McNaught and colleagues collected gastric aspirates at induction of anaesthesia in elective surgical patients who had received at least 1 week of oral Lp299 [24]. Th e probiotic species was not isolated in any subject. In the study by Forestier and colleagues, however, gastric aspirates were taken at inclusion, at day 7 and at discharge. Lactobacillus casei rhamnosus was detected in 52 out of 102 patients on probiotic treatment after a median of 13 days [49]. In the study by Klarin and colleagues described above, the probiotic species Lp299 was detected in all oropharyngeal cultures and in the tracheal cultures from 56% of patients in the probiotic arm [50]. Knight and colleagues demonstrated detection of probiotic species in stool culture after 3 days treatment with Synbiotic 2000 Forte, indicating its survival from the stomach to the distal gut [56]. However, they did not routinely analyse stool samples in their more recent study [51]. None of the other studies cited in Table 3 reported detection of probiotic species in any microbiological specimens.

Quality and safety
Probiotics are now widely available and are being consumed daily in large quantities. Overall they have an excellent safety record, but there are some concerns that are likely to lead to caution in their widespread use in clinical practice.
Th e availability of diff erent probiotics varies from country to country and there can be lack of consistency between manufacturers, and even batches, in terms of density of bacteria, adhesion characteristics, stability and viability [57]. Strain-specifi c adhesion properties and viability have been shown to vary between batches from the same manufacturer, which could lead to confl icting clinical trial results [58].
Th ere have been a number of publications reporting serious infections caused by Lactobacillus spp. related to those used as probiotics [59]. Th e Finnish group of Salminen and colleagues examined 89 cases of Lactobacillus bacteraemia. In 11 cases, the strain was identical with the probiotic L. rhamnosus GG [60]. However, they could not directly relate these cases to probiotic consumption. Salminen and colleagues also examined trends in Lactobacillus bacteraemia in Finland over the period 1990 to 2000. Th is period coincided with a rapid increase in the consumption of probiotic L. rhamnosus GG. Th e group concluded that increased probiotic use had not led to an increase in Lactobacillus bacteraemia [61].
Th ere are case reports in the literature of Lactobacillus infection and bacteraemia that appear to be directly related to probiotic consumption [62][63][64][65]. All of the patients involved were immunocompromised to some degree and the causative organism was linked to the probiotic by molecular techniques. Infections caused by Lactobacillus spp. from probiotics have also been reported in immunosuppressed patients -including those with acquired immunodefi ciency syndrome and those following lung and liver transplantation [66][67][68]. Lactobacillus bacteraemia has been associated with structural heart abnormalities, valve prosthesis or prior endocarditis [69]. However, the majority of clinical trials using Lactobacillus spp. probiotics report few adverse eff ects.
Th e only reported infection associated with probiotic E. coli Nissle 1917 is in a premature neonate (gestational age 28 weeks) [70]. Th e child had an extremely low birth weight of 935 g and developed gastroenteritis due to rotavirus and adenovirus 14 days into the postnatal period. E. coli Nissle treatment initially led to improvement but the child developed severe sepsis 10 days later and subsequently E. coli Nissle 1917 was isolated in blood cultures. Th e child was treated with antibiotics and made a full recovery.
A wide range of probiotic species is being investigated for an increasing number of indications. Th ere has been little work carried out on the rationale behind which probiotics are used and in what combination. Timmerman and colleagues attempted to address this issue by examining specifi c strains in an attempt to produce an eff ective multispecies mixture [71]. Th e symbiotic preparation Ecologic 641 was used in the PROPATRIA trial. Th is group selected six strains of Lactobacillus based on survival in a simulated gastrointestinal environment, antimicrobial activity and ability to induce IL-10, highlighting the point that there should be a diseasespecifi c rationale for selection of probiotics.

Conclusions
Concerns are mounting about multidrug-resistant Gramnegative bacteria with the extensive spread of extendedspectrum β-lactamases [72], and in particular the emergence of Enterobacteriaceae with resistance to carba penems conferred by metallo-β-lactamase NDM-1 [73]. New antimicrobial agents with which to tackle resistant bacteria are in limited supply, and a recently announced EU-US taskforce has called for a commitment to the development of 10 new antibacterial agents by 2020. Th is will require a substantial public fi nancial investment and will need to be sustained long term because continued antibiotic use will maintain the pressure on organisms to evolve new resistant strains. In the absence of universally eff ective treatments, strategies that could prevent the development of ICU-acquired infection are needed.
Th e human, animal and in vitro studies of probiotics carried out to date exhibit a high level of heterogeneity in the conditions targeted, models used and probiotics tested. Th ese studies are likely to refl ect an oversimplistic view of the mechanisms of action of probiotic species. As alluded to above, probiotics are likely to bring about their eff ects through multiple processes with diff erent strains having very specifi c eff ects.
We are still far from understanding fully the probiotichost interaction but, given the potential benefi ts that probiotic bacteria have to off er, further study is warranted. Careful consideration should be given to further well-powered studies addressing the questions of which probiotic by what route, in what dose and at what time.