Bench-to-bedside review: Challenges of diagnosis, care and prevention of central catheter-related bloodstream infections in children

Central venous catheters (CVCs) are indispensable in modern pediatric medicine. CVCs provide secure vascular access, but are associated with a risk of severe complications, in particular bloodstream infection. We provide a review of the recent literature about the diagnostic and therapeutic challenges of catheter-related bloodstream infection (CRBSI) in children and its prevention. Variations in blood sampling and limitations in blood culturing interfere with accurate and timely diagnosis of CRBSI. Although novel molecular testing methods appear promising in overcoming some of the present diagnostic limitations of conventional blood sampling in children, they still need to solidly prove their accuracy and reliability in clinical practice. Standardized practices of catheter insertion and care remain the cornerstone of CRBSI prevention although their implementation in daily practice may be difficult. Technology such as CVC impregnation or catheter locking with antimicrobial substances has been shown less effective than anticipated. Despite encouraging results in CRBSI prevention among adults, the goal of zero infection in children is still not in range. More high-quality research is needed in the field of prevention, accurate and reliable diagnostic measures and effective treatment of CRBSI in children.


Multimodal prevention strategies
Avoiding contamination that would lead to subsequent CVC colonization is supposed to be the key element in decreasing the risk of CLABSI [11]. CLABSI occurs through extraluminal contamination (microorganisms migrating from the insertion site along the external of the catheter) or intraluminal contamination (pathogens migrating from the catheter hub through the lumen of the catheter) with subsequent colonization and biofi lm formation [12]. While extraluminal contamination is supposed to be the most common mechanism of CLABSI with short-term catheters [13], the intraluminal route is believed to be the more prevalent route of infection with long-term catheters (duration >10 days) [14].
A number of studies demonstrate the eff ectiveness of implementing standardized procedures and care bundles for CVC insertion and CVC care on CLABSI or CRBSI reduction. Elements for prevention upon CVC insertion include the use of maximum sterile barrier precautions, (alcohol-based) chlorhexidine for skin antisepsis, a checklist to stop non-emergent insertion and establishing fully equipped insertion carts [15][16][17][18][19][20][21]. Elements for prevention in CVC care include standardization of dressing change, skin antisepsis and replacement of tubing, and improving workfl ow at the patient [9, [15][16][17][18][19][20][21]. Outcome reductions range from 70 to 83% using diff erent strategies. No conclusion can be made for single interventions but only for the multimodal use of a defi ned set of procedures. Most studies applied a before-and-after study design and thus the quality of the studies is limited and the eff ectiveness in infection prevention may have been overestimated due to high baseline infection rates (7.8 to 8.6/1,000 days), but multimodal prevention strategies are still probably the most eff ective and important means of reducing CRBSI -especially for short-term catheters.

Biofi lm formation in central venous catheters; prophylaxis and treatment aspects
Biofi lm formation plays a major role in the pathophysiology of CRBSI. Biofi lm acts both as a mechanical barrier and as an environment for genetic exchange and thereby contributes to protection from elimination by the innate host immune defense and to emerging antibiotic resistance. Biofi lm formation is revealed to be a two-stepped process with initial adhesion of planktonic microorganisms and a subsequent maturation phase [22]. Bacterial expression of so-called microbial surface components recognizing adhesive matrix molecules has the capacity to bind to human matrix proteins such as fi brinogen or fi bronectin [23]. Cofactors for the adhesion process are the presence of cations [24][25][26] and bacterial stress. Th e maturation phase of biofi lm formation is characterized by intercellular aggregation and production of extracellular matrix leading to a typical three-dimensional structure. Bacterial lysis, DNA release and quorum-sensing systems play major roles in the development of the structure [27,28].
Most vascular devices develop biofi lm within 24 hours after insertion [22]. Th e occurrence of CRBSI is proportional to the occurrence of microorganisms on the catheter tip, supporting the theory that a critical level of colonization is necessary for the detachment of planktonic bacteria, embolization and systemic infection [29,30]. Th e ability of a pathogen to form biofi lm can be considered a virulence factor, as it is associated with mortality [31]. As metallic cations convey adherence of microorganisms to the surface of catheters, adhesion and thus biofi lm formation is reduced by chelating agents [32][33][34][35]. In vitro studies found that ethylenediamine tetraacetic acid, citrate and N-acetyl-cysteine eff ectively reduce proliferation of Staphylococcus epidermidis and C. albicans and even eradicate existing biofi lm [36].

Preventive lock solutions
Lock solutions in CRBSI prevention were tested almost exclusively in long-term CVCs. A recent study among adult hemodialysis patients found catheter lifespan prolongation using a lock solution with a highly concentrated chelating agent (sodium citrate 46.7%) in the absence of antimicrobials [37]. CRBSI was not reduced. Similarly, a lock solution combining citrate 7%, methylene blue and paraben eff ectively reduced infection rates in a recent study among adult hemodialysis patients [38].
Antibiotic CVC locks versus heparin locks were found eff ective in adults by a recent systematic review (relative risk = 0.37/catheter-day (95% confi dence interval = 0.30 to 0.47)) [39]. Only one randomized controlled trial (RCT) has compared a vancomycine lock with heparin in a predominantly pediatric population, reporting a reduction of bacteremia among non-neutropenic patients [40]. Two RCTs evaluated the eff ect of adding vancomycine to PEN infusions in neonates and found bacteremia and CVC colonization signifi cantly reduced [41,42]. In vitro studies suggested a synergistic eff ect by combining antibiotics, chelators and disinfectants [43][44][45]. A small proof-of-concept study confi rmed the eff ectiveness of a lock solution combining minocycline and ethylenediamine tetraacetic acid on infection rates and prolonged catheter survival as compared with heparin [46]. An additional small study reported a similar eff ect for minocycline/ethylenediamine tetraacetic acid among a small cohort of children with an implantable device [47]. Th ere are no comparative studies between antibiotics alone and in combination with chelating agents.
Various non-antibiotic locking techniques have been proposed, such as taurolidine-citrate for hemodialysis catheters [48]. Taurolidine acts as a disinfectant, irreversibly damaging bacterial and fungal cell walls and disrupting biofi lm, while citrate is a chelating agent [48]. Taurolidine-citrate locks reduced the incidence of bloodstream infection (BSI) due to Gram-negative organisms in two adult studies, but showed limited eff ect when BSI was caused by Gram-positive organisms [49,50]. In contrast, a recent study among pediatric hema tology patients reported reduced overall BSI incidence in the taurolidine group as compared with the heparin group [51]. Th e study was very small, however, which puts the fi nding into perspective.
Ethanol locks, preferably at a concentration of 70%, have been studied mostly in children requiring PEN [52][53][54]. Ethanol appears to aff ect biofi lm formation through protein denaturation [36]. A recent systematic review evaluated four studies among children treated by PEN and calculated a relative risk of 0.19 (95% confi dence interval = 0.12 to 0.32) for CRBSI [52]. However, the included studies were heterogeneous, retrospective and small.

Impregnated catheters and dressings
In adults, technologies for infection prevention include catheters impregnated with chlorhexidine-silver sulfadiazine or antibiotics [55][56][57][58] and the use of chlorhexidine-impregnated dressings [59]. Availability of impreg nated catheters for small children is limited and chlorhexidine-impregnated dressings were found to be causing contact dermatitis in neonates [60,61]. Furthermore, no studies have been done for children other than neonates.
Two systematic reviews about the use of impregnated catheters in adults found only signifi cant and substantial reductions in CRBSI for heparin-coated and antibioticimpregnated catheters, while no benefi ts were identifi ed for antiseptic CVCs, coated with chlorhexidine and silver sulphadiazine, or silver-impreg nated CVCs. [57,62]. Th e only RCT in children compared heparin-bonded against standard catheters [63]. Routine blood cultures were performed every 3 days and the catheters were cultured after removal. Th e hazard ratio for the endpoint of any positive culture result was 0.11 (95% confi dence interval = 0.04 to 0.31) for children with heparin-bonded catheters.
Preventive strategies are summarized in Table 1.

Thrombosis and infection
Th e association between catheter infection and thrombosis has been suggested repeatedly although the results of the older studies were never really confi rmed and the more recent publication is limited by size [64][65][66].
Although some authors argue that infection precedes thrombosis, a recent pediatric study suggested that thrombotic occlusion rather precedes infection [67]. Although fi brin deposition may provoke infection [68], the vast majority of hypercoagulability disorders do not seem to predispose for CRBSI [69]. A recent Cochrane systematic review of 12 RCTs comparing the preventive administration of anti-thrombotic agents (heparin or vitamin K antagonists) against placebo in 3,611 pediatric cancer patients only found a nonsignifi cant trend towards decreased incidence of symptomatic deep vein thrombosis and a nonsignifi cant reduction of CRBSI [70]. Another Cochrane systematic review identifi ed two studies about preventive urokinase fl ushing versus heparin fl ushing, and no signifi cant CRBSI reduction was identifi ed [71]. One of the included studies reported fewer occlusive events in the urokinase group but no BSI reduction [72].
In summary, the interplay between occlusion, infection and deep vein thrombosis remains unclear. Although pathophysiologically such association appears likely, no formal association has been demonstrated in children, most probably due to the fact that both events are relatively rare.

Catheter-related bloodstream infection diagnosis
Confi rming CRBSI in children is challenging. According to guidelines, blood cultures should be taken both from the central catheter and from a peripheral vein upon clinical symptoms, and CRBSI is most likely when the colony count from the catheter is fi vefold to 10-fold higher than the colony count from the peripheral vein, or when the diff erential time to positivity between the two blood culture samples exceeds 2 hours [73]. Peripheral sampling is painful and unpleasant, and thus is not routine practice in children. Blood culture sampling in children uses smaller volumes (1 to 3 ml compared with 10 to 20 ml in adults), and although specially enriched culture media are used, the chance of isolating live organisms drops below 70% as compared with adults [74,75]. Correctly performed blood cultures are more likely to be positive [76]. Only two-thirds of the blood samples from infants <1 month old were adequate for culture, adequate being defi ned as containing an appropriate (age-related) volume of blood and being submitted in the correct blood culture bottle type. Th is information should prompt healthcare workers to be more careful when taking blood from neonates for blood culture [77]. If antibiotic treatment is initiated before sampling, the sensitivity of the test is further reduced [78]. Incubation times often exceeding 72 hours and low sensitivity due to blood sampling make traditional blood cultures a lengthy process to rule out BSI [79,80]. As a consequence, new identifi cation systems such as nucleic acid testing are promoted. Nucleic acid testing, which is based on pathogen lysis, nucleic acid extraction, purifi cation, amplifi cation and identifi cation, can be used either to address a single pathogen of interest or to identify a large range of bacteria and fungi, even in a simul taneous process [81]. Techniques such as nucleic acid testing may serve to identify colonies isolated by culture or they may be applied directly to blood samples [81]. Protein-based identifi cation via mass spectrometry by matrix-assisted laser desorption/ionization-time-offl ight (MALDI-TOF) is a rapid tool for pathogen identifi cation, and is already widely established [82][83][84]. Th e technique uses mass spectral signals from clinical samples, which are compared with a reference database, thus allowing swift and accurate identifi cation. MALDI-TOF has been shown to have limitations in the identifi cation of some Gram-positive organisms or when the infection is polymicrobial [83]. However, prior culturing of organisms from clinical samples overcomes such limitation.
PCR-based diagnostics addressing housekeeping genes of microorganisms can theoretically be used directly in clinical material.
Quantitative detection of pathogen DNA without species identifi cation may add information in confi rming suspicion of CRBSI in children without prior culturing and can also detect bacterial DNA debris after initiation of antibiotic therapy [78,85]. In one recent study, there was a dose-dependent association between the amount of bacterial DNA per microliter of blood and suspicion of confi rmed BSI [86]. Concentrations >0.5 pg/μl had a high positive predictive value for CRBSI.
Th e most intensively studied clinical method of simultaneous, pathogen-specifi c diagnosis is SeptiFast (Roche, Mannheim, Germany), a multipathogen probebased real-time PCR system target ing DNA sequences of the bacterial and fungal genome simultaneously, which allows the detection of 25 common pathogens from one single blood sample [87]. Th e sensitivity/ specifi city values (60 to 95%/74 to 99%, depending on the pathogen) and positive predictive values of SeptiFast were superior to blood cultures in a number of studies, but the diagnostic value in clinical practice still awaits the results of large, well-designed clinical trials [81,88,89]. Th e test is rapid, but it identifi es only a limited range of pathogens and does not provide information on antibiotic resistance, which is an important limitation in times of emerging resistance [87,89].

Management of catheter-related bloodstream infection
Suspected CRBSI creates the dilemma of whether or not to remove the CVC. Guidelines recommend this decision is made based on individual, clinical judgment [59,90]. Th e patient often still requires hemodynamic monitoring, fl uid intake, PEN, and medication. CVC removal would therefore result in the insertion of a new catheter with the risk of mechanical and infectious complications and would expose the child to anesthesia. Th ere is no RCT evaluating the evidence for early versus late CVC removal in neonates [91]. Only retrospective studies with confl icting conclu sions are published [92][93][94][95].
CRBSI must always be treated with systemic antibiotics. However, while antibiotics are able to clear planktonic, free-fl oating microorganisms, their ability to eradicate biofi lm-embedded organisms is limited due to low penetration into biofi lm, changes in bacterial metabolism, antimicrobial resistance, and local alterations in the microenvironment of the biofi lm that impair the activity of the antimicrobial agent [96]. To eradicate microorganisms embedded in biofi lm, antibiotic concentrations would have to be 100-fold to 1,000-fold that required to kill free-fl oating organisms [97,98]. High antibiotic concentrations instilled in the catheter lumen for hours or days alongside systemic antibiotic therapy have been used with success rates between 31 and 100% depending on the pathogen [99][100][101][102][103][104][105][106]. Most studies reported treatment durations of 14 days. Four small pediatric studies about antibiotic locks reported immediate success rates ranging from 83 to 100%, but no information about long-term eff ects was provided [107][108][109][110]. While inclusion criteria and choice of antibiotic vary greatly in these studies, the vast majority of infections in all studies are caused by CoNS.
In summary, attempts at CVC salvage are only recommended in uncomplicated CRBSI or CLABSI caused by bacteria that are neither too virulent nor too diffi cult to eradicate -predominantly CoNS or enterococci [23].
Alternative strategies for CVC treatment include nonspecifi c disinfectants such as hydrochloric acid (HCl). HCl is thought to disrupt biofi lm through denaturation of protein components in the extracellular matrix, exposing the microorganisms to subsequent antibiotic treatment [111]. Th e instillation of 2 M HCl was des cribed in 2004 for the fi rst time [112]. Th ree cycles of HCl exposure for 10 minutes within 1 day in combination with systemic antibiotic therapy that took into account the susceptibility of the identifi ed pathogen resolved the infection in 28 out of 42 (67%) patients with persistent CVC-related bacteremia, defi ned as positive blood cultures >48 hours after the initiation of antibiotics. Another study of HCl instillation using a similar protocol was able to prolong CVC survival as compared with a historical control [111].
Although of concern, there are no data about catheter damage due to HCl treatment [113].
Ethanol also acts as a nonspecifi c substance interfering with the biofi lm matrix and has a direct disruptive eff ect on microorganisms [114]. Attempted catheter salvage by ethanol locks with dwell times between 12 and 24 hours have been reported to be successful in 67 to 88% of published observational studies and case reports [114][115][116]. A small RCT identifi ed a reduced CRBSI incidence using a 70% ethanol lock in tunneled central lines among hematology patients [117]. A recent large RCT failed to show any signifi cant reduction in the incidence of endoluminal CRBSI, however, and thus its true benefi t remains uncertain [118]. Th is uncertainty is further supported by preliminary results of another large RCT including 359 long-term tunneled or implanted CVCs, which failed to show any benefi t of a 50% ethanol lock for CRBSI prevention [119].
Th erapeutic strategies are summarized in Table 2.

Discussion
Avoiding contamination upon catheter insertion and catheter care is the cornerstone of CRBSI prevention, and a number of studies support the positive eff ect of improved routine procedures in the topic [9, [15][16][17][18][19][20][21]. In pediatric settings, study designs are mostly quasiexperimental and neither high-quality studies nor robust analyses clearly document the eff ect of such inter ventions. Th is does not imply that best-practice procedures would not work, however, and it is most likely that they contribute notably to a reduced risk of infection. Documented declines in infection rates [21,[120][121][122] allow the assumption that good clinical practice is pivotal in CRBSI prevention. Th ese results are largely based on short-term CVCs in adult and pediatric or neonatal ICUs. Although the contribution of the individual steps of what is called best practice is uncertain, multimodal intervention strategies were eff ective even if the combination of the specifi c interventions varied across the studies [123]. Bringing the risk of infection to the attention of healthcare providers already positively impacts patient outcomes. Th is phenomenon, called the Hawthorne eff ect, contributes to the positive outcome of many CRBSI prevention programs and thus should always be taken into account when interpreting data from quality improvement studies -especially in open-label protocols and when using a historical control group.
Although best-practice procedures of catheter insertion and catheter care to prevent bacterial coloni zation have been proven eff ective in CRBSI prevention, the principal pathogenesis of CRBSI is related to the catheter material itself promoting colonization with microorganisms and the formation of biofi lm. As eradicating biofi lms is diffi cult, if not impossible, preventing attachment and limiting biofi lm formation are crucial tasks alongside correct catheter care in achieving the goal of zero infection -especially with long-term indwelling devices.
Maintaining best practice may be challenging with long-term CVCs and thus prophylactic locks have been intensively studied in pediatric patients using both antibiotics [40][41][42] and other antimicrobial substances [52][53][54][55] attempting to prevent the formation of biofi lm. Despite evidence supporting antibiotics in lock solutions [39], there has been some reluctance to recommend such locks in routine use due to growing emergence of multidrug-resistant microorganisms [41,42,46,47]. A recent systematic review of 16 studies with a total of 176,332 CVC-days reported only one single case of bacterial resistance, but the authors argued that only few studies thoroughly addressed this issue and none performed surveillance cultures or examined long-term eff ects (>12 months) [39]. As for catheter salvage upon suspected BSI, pediatric studies are limited in quantity and quality. Th ey provide limited evidence in favor of highly concentrated antibiotic lock solutions with long indwelling times, although the studies vary in design and choice of antibiotic [107][108][109][110]. On the other hand, the vast majority of detected and successfully treated CRBSIs were caused by CoNS and thus the fi ndings are in line with the advisory guidelines, which propose that an antibiotic lock may be envisaged in uncomplicated CRBSI due to CoNS and enterococci [56]. It is of interest that primary outcome was either a subsequent negative blood culture or the absence of subsequent positive blood cultures. Bearing in mind the limited reliability of blood culture sampling in children, such an outcome may overestimate the treatment eff ect. A recent adult RCT failed to demonstrate a treatment eff ect because the calculated sample size was not achieved due to strict exclusion criteria [124]. Most frequently, patients were excluded because of permanent CVC use, which made an 8-hour to 12-hour lock impossible. Th is is frequently the case in children with CRBSI because the catheter may represent the only route of administration for fl uids and nutrients, and the insertion of a temporary peripheral or central line may be technically diffi cult or undesirable due to anesthesia.
Prophylactic locks by non-antibiotic agents such as taurolidine-citrate [51], ethanol [52][53][54][117][118][119]125] and HCl [111,112] have been shown eff ective in pediatric settings. However, such locks imply a risk of inadvertent fl ushing or spillover to the systemic circulation. Due to low body weight this has more serious consequences in children, particularly in neonates. Taurolidine-citrate locks containing 4% sodium citrate provoked adverse events such as nausea, vomiting, and abnormal taste sensations in 20% of the children, possibly due to spill over of citrate causing a drop in plasma concentrations of calcium and magnesium [51]. Possible adverse events at high concentrations such as 46.7%, as used in a successful trial [37], may include severe or even fatal arrhythmias due to electrolyte disturbances [126]. Th e side eff ects of ethanol are dose dependent and include dizziness, lightheadedness, tiredness, headache and liver toxicity [115]. In addition, disseminated intravascular coagulation and thrombosis have been reported as possible adverse events in children [53,54]. Ethanol locks, like antibiotic locks, require long indwelling times and thus their use is limited given the short time potentially available for locking in children often having only a single catheter. Spillover of HCl may cause hemolysis [112]. Such complications have not yet been reported, but generally the possible risks of HCl remain insuffi ciently investigated. Antimicrobial coating of CVCs is another means to prevent infection, and impregnations with antibiotics or heparin have been shown of benefi t [55,62,63]. However, most studies are unblinded and only one study has been performed in children. Th e study compared heparin coating with nonimpregnated catheters, and although CRBSI data had been obtained the reported outcome was all-cause BSI, which weakens the fi ndings [63]. Most studies looked at catheter colonization as a primary outcome, which is questionable since a catheter may contaminate on removal. In addition, impregnated catheter tips can leak antimicrobials into the agar plate and thus provoke false negative culture results [62].
While there is evidence to support the use of antibiotic locks for CRBSI prevention, data for non-antibiotic locks in children are absent and well-designed studies are necessary to address this issue.
Quick and accurate microbiological diagnosis upon suspected CRBSI is of pivotal importance in pediatric critical care, but is challenging. New molecular techniques are promising, but not without shortcomings [81]. Methods such as MALDI-TOF are primarily add-ons requiring prior cultivation and although they shorten the time of pathogen identifi cation, they do not overcome the challenge of low sensitivity of blood cultures in children due to inconsistent blood sampling [83].
Rapid multiplex real-time PCR tests to be applied directly to clinical samples may reduce the time to diagnosis [81]. However, existing PCR-based methods have not overcome shortcomings such as limited spectra of identifi able pathogens, limited antibiotic resistance testing, central laboratory facility requirements, lack of 24-hour availability and cost issues [87,88]. For the time being, PCR-based methods still require conventional culture being performed because susceptibility testing is paramount in today's emerging resistance. Ideally, PCRbased methods would be performed for rapid diagnosis of a pathogen, and blood cultures would be taken to confi rm pathogen identifi cation and to obtain antibiotic susceptibility. However, in pediatric and neonatal care settings, an additional sampling volume (for example, +1.5 ml for SeptiFast) may not be feasible due to practical limitations.

Strengths and limitations
While attempting to provide a broad review of the microbiological and clinical challenges with CVCs in children, long-term catheters appear to be over-represented compared with short-term CVCs, peripherally inserted central catheters and umbilical catheters. Th is imbalance is partially due to the fact that short-term catheters or CVCs in the neonatal ICUs are addressed often in best-practice multimodal improvement strategies while technical prevention methods were more probably applied to long-term catheters.
Some discussed results are based on various adult populations [37,39,55,57,59,62]. Catheters used for adults are of diff erent length, diameter and material than those used for children, and there is a considerable diff erence in the disease spectrum indicating CVC placement and use in children and adults. Th ere are also age-related physiological diff erences, which may aff ect the risk of infections related to the devices, especially the immaturity of the immune system and therefore relative immune incompetence in children.
Finally, children may have their CVCs longer than adults due to diff erent clinical treatment regimes (that is, conversion from central to peripheral administration of medicine with increasing age) or merely due to the fact that critically ill children have better survival rates compared with adults.
Keeping in mind all of the above allows few overall evidence-based conclusions, and as such this review may leave the impression that little has been achieved. Th is is not the case, however, as multiple promising strategies have been proposed and indicate possible directions for future research.

Conclusion
Despite increasing research activity in CRBSI prevention in the past years, the goal of zero infection is still not in range -not for adults, and even less so for children. Insertion and care bundles addressing behavior change are diffi cult to implement in many hospital settings, and technology such as locks and antimicrobial impregnation of catheters are less eff ective than anticipated and many of them were tested in small studies of limited quality and rarely in children. For the time being, physicians must live with the limitations of blood sampling in children upon suspicion of CRBSI. New molecular tests, although promising, must show their eff ectiveness and reliability in clinical practice. In general, more highquality research is needed in the fi eld of infection prevention in children. Th is does not only apply for CRBSI prevention, but is indeed true for the entire fi eld of infection control.