Is systematic fecal carriage screening of extended-spectrum beta-lactamase-producing Enterobacteriaceae still useful in intensive care unit: a systematic review

We searched the MEDLINE database for English language articles published from the inception of the database to February 15, 2019. A combination of MeSH/Emtree and title/abstract keywords was used. The search terms were “ESBL,” “ESBL-E,” “ESBLE,” “extended spectrum beta-lactamase,” “ICU” with non-relevant terms “neonatal,” “pediatric,” “children,” and “infants”.

All the available data were extracted from each study by two investigators (DG and RP) independently according to the aforementioned inclusion criteria, and any differences were resolved by discussion with a third investigator (JRZ). The following data were collected from each study: the name of the first author, publication year, study design, number of patients, primary outcome, and risk factors for the primary outcome.

Using the previously described request in MEDLINE database, 330 articles were referenced. Every abstract was read and 39 articles that appeared to address issues relevant for this review were selected for full-text assessment. After full-text assessment, 25 studies were included in the systematic review (Fig. 1). Seven were relevant regarding the risk of ESBL-E cross-transmission (Table 1), 4 regarding the evaluation of hygiene procedures’ efficacy (Table 2), 10 assessed the link between colonization and infection or the prognostic value of ESBL-E carriage to guide the empirical antimicrobial therapy (Table 3), and 4 the efficacy of a selective decontamination strategy (SD) (Table 4).

The first study demonstrating ICU ESBL-E cross-transmission in a non-outbreak situation describes two clones responsible for clusters of 5 and 12 patients. Nevertheless, this study included patient in 1990 and 1991 and did not focused on carriage but infection [22]. Another study involved 2883 patients with 28 (0.97%) ESBL-E carriers. Only 9/28 cases of ESBL-E were explained by cross-transmission despite the fact that the screening method was suboptimal limiting their impact on the prevention of ESBL-E dissemination [23]. These results were consistent with another one detecting 97/1806 (5%) of ESBL-E fecal carriers including 23/97 (24%) of ICU-acquired ESBL-producing E. coli fecal carriage with only 3/23 (13%) acquisitions likely to be due to cross-transmission [24]. Similar results were demonstrated even in a higher prevalence situation (ESBL-E carriage at admission: 98/347 (28.2%)), with an acquisition rate of 12.1% (11/91) without any case of cross-transmission [25]. Another study based on pulsed-field gel electrophoresis (PFGE), revealed two cases of suspected cross-transmission for E. coli and two for E. cloacae in a setting with 50/316 (15.8%) of ESBL-E fecal carriage from patients on the liver transplantation, ICU, and hematology/oncology wards. Nevertheless, only one case of suspected cross-transmission (E. cloacae) occurred in ICU [26]. In other studies, cross-transmission has also been shown to be a rare event (5% of ESBL-E acquisition) [27], even in an ICU with no single room [28]. The occurrence of ESBL-E acquisition despite limited cross-transmission suggests limits to hygiene procedure efficacy in controlling ESBL-E dissemination in a non-outbreak situation and other mechanisms of dissemination which we still have to investigate.

A large multicenter study in 13 European ICU testing the effect of rapid screening and isolation of carriers with contact precautions did not find any impact on ESBL-E acquisition, but the study was led in the context of a sustained high level of compliance to hand hygiene and chlorhexidine bathings [29]. In another study, cessation of contact isolation procedures had no independent impact on ESBL-E infections (adjusted OR 1.16, 95% CI 0.38–3.50, p = 0.79) nor on in-ICU death (SHR 1.22, 95% CI 0.93–1.59, p = 0.15) [30]. These results were consistent with a previous study which found that discontinuing contact precautions did not increase ICU-acquired ESBL-E fecal carriage (incidence densities respectively during contact vs standard precautions: 2.7 (95% CI 1.78–3.62) and 2.06 (95% CI 1.27–2.86) per 1000 patient-days; p 0.004 for non-inferiority) in an ICU with single rooms with dedicated equipment, strict application of hand hygiene, medical and paramedical leadership, and good antibiotic stewardship. [31].

A cost-effectiveness analysis showed that an improved compliance with hand hygiene is the most cost-saving strategy to prevent the transmission of ESBL-E. Screening and cohorting had comparable effectiveness but were more expensive; screening and contact precautions were the least effective strategy [32].

To summarize, these results together suggest that cross-transmission does not seem any more to be the main source of ESBL-E acquisition in ICU. Moreover, a universal standard precaution strategy seems to be sufficient to control the risk of cross-transmission. The relevance of a systematic ESBL-E fecal carriage screening policy in ICU to guide hygiene procedures is now questioned (Additional file 1).

Another discussed interest of systematic screening of ESBL-E fecal carriage could be to guide empiric antimicrobial therapy in case of subsequent infection among carriers. The most investigated ICU-acquired infections are VAP and bloodstream infections (BSI) as they are the more frequent ones [33]. Among ESBL-E fecal carriers in ICU, one study found that 10% and 27% of first and second episodes of ICU-acquired infections [34] and another that even 40% of VAP are due to ESBL-E [19].

Regarding BSI, a recent study conducted in China found a proportion of 0.5% of ESBL-E fecal carriers developing subsequent ESBL-E BSI in ICU (42/9015). Independent risk factors associated with subsequent ESBL-E BSI were antibiotic use in the past 72 h (penicillin [OR 12.076; 95% CI 1.397–104.251, p 0.024], cephalosporin [OR 6.900; 95% CI 1.493–31.852, p 0.013], carbapenem [OR: 5422; 95% CI 1.228–23.907, p 0.026]), previous ICU stay (OR 1.041; 95% CI 1.009–1.075, p 0,012), and maximum body temperature (OR 8014; 95% CI 2.408–26.620, p 0.001) [35]. Nevertheless, as a case is diagnosed by a positive blood culture, antimicrobial therapy and maximum body temperature could only be the signal of the on-going infection with empiric treatment initiated as ESBL-E BSI are compared with non-infected ESBL-E fecal carriers.

Regarding risk factors for VAP, ESBL-producing Klebsiella pneumoniae was found to be an independent risk factor ICU-acquired ESBL-producing K. pneumoniae pneumonia (OR 60.6; 95% CI 56.33–578.73) [36]. Other identified independent risk factors for ESBL-E pneumonia among ESBL-E fecal carriers were SAPS II at admission > 43 [OR 2.81 (1.16–6.79)] and colonization with Enterobacter sp. or K. pneumoniae species [OR 10.96 (2.93–41.0)] whereas receipt of > 2 days of amoxicillin/clavulanic acid during the ICU stay was protective [OR 0.24 (0.08–0.71)] [37]. Despite these identified risk factors, individual prediction of subsequent ESBL-E VAP among ESBL-E fecal carriers remains difficult and prognostic values will be discussed thereafter.

Regarding ESBL-E pulmonary colonization, ESBL-E fecal carriage has an excellent negative predictive value (NPV) (99.2%, 95% CI [98.7,99.6] for ≤ 5 days and 93.4%, 95% CI [91.9,95.0] for > 5 days) despite a poor positive predictive value (PPV) (14.5% [95% CI 12.8, 16.3] and 34.4% [95% CI [31.4, 37.4]), for the early and late groups respectively [38].

Nevertheless, pulmonary colonization does not mean infection. A monocentric prospective study found 111 (13%) patients among 843 ESBL-E fecal carriers who developed ICU-acquired pneumonia of whom 48 (43%) had ESBL-E pneumonia (6% of carriers). Patients with ESBL-PE pneumonia in this study had a higher SOFA score (p = 0.037) and more frequent septic shock at pneumonia onset (p = 0.047) than patients with pneumonia due to another germ [37]. Even if this article also identified risk factors, it remains very difficult to predict which patient will suffer from a ESBL-E VAP or not. Enhancing this result, an inception cohort of the prospective database OUTCOMEREA showed that infectious-related ventilator-associated complications among ESBL-E fecal carriers mostly reflect non-VAP events (18/361, 13%) with only 18/361 (5%) ESBL-E VAP but that they are a major driver of carbapenem consumption [39]. Regarding predictive values, a retrospective monocentric study showed a 41.5% PPV and a 99.4% NPV [19]. Another study confirmed prior ESBL-E fecal carriage to be the only independent risk factor for subsequent ESBL-E VAP [OR 23 (95% CI 10–55)] with a PPV of 43.6% and a NPV of 97.3%. Duration of mechanical ventilation, length of ICU stays, and mortality rates (55.8% vs 50%, p = 0.48) were similar in ESBL-E VAP, compared with VAP due to other bacteria [40]. Because of these excellent NPV, some authors have suggested that systematic ESBL-E fecal carriage screening could help to limit the use of carbapenems, but this assertion is now contradicted.

Jalalzaï et al. assessed carbapenem consumption after cessation of screening for intestinal carriage of ESBL-E during two consecutive 1-year period (with and without systematic screening with respectively 524 and 545 patients) [30]. An admission during the no-systematic screening period exerted no independent impact on the hazards of ESBL-E infections and in-ICU death. The exposure to carbapenems in patients without ESBL-E infection even decreased between the systematic screening and no-systematic screening periods (75 versus 61 carbapenem-days per 1000 patient-days, p = 0.01). These results are consistent with a prospective multicentric study revealing that ESBL-E infections are rather infrequent in carriers and that carbapenem exposure was increased among ESBL-E carriers without infection. ESBL-E carriers even without infections also had a delayed discharge, thereby amplifying the selective pressure and the colonization pressure in ICU [41].

To summarize, the link between ESBL-E colonization and subsequent ESBL-E infection seems to be real as consistently observed by several different teams. Nevertheless, subsequent ESBL-E infections are a rare event which is almost unpredictable. An empirical antimicrobial therapy guided on ESBL-E carriage status leads to an overconsumption of carbapenems without a clinical benefit. A systematic screening policy of ESBL-E fecal carriage in ICU to guide empirical antimicrobial therapy is now questioned. Further studies are needed to better understand the link between colonization and infection and to assess if we can improve the prediction of subsequent ESBL-E infections. Disturbances of gut microbiota could be part of the explanation, and so selective decontamination has been evoked as a tool to modulate the gut microbiota and to eradicate multi-drug resistant bacteria fecal carriage [21, 42].

Several studies aimed to identify reliable risk factors to predict subsequent ESBL-E infections. In non-ICU patients, severity at admission and colonization with Enterobacter sp. or K. pneumoniae; referral from a medical ward, nursing home, or rehabilitation center; previous fluoroquinolone treatment; extracorporeal membrane oxygenation; and the absence of prior positive ESBL-E rectal swab culture were identified as risk factors [62]. Despite these identified risk factors, it remains difficult to predict which patient is infected with an ESBL-E or not, especially in ICU [39]. The effect of BLI use prior to infection among ESBL-E fecal carriers remains unclear [37, 63]. Available data focus on VAP and BSI as they are the most frequent ICU-acquired infections and as the distinction between urinary tract infection and urinary colonization in catheterized patients can be seriously challenging.

As described for the risk of transmission, K. pneumoniae could be at increased risk of subsequent infections in colonized patients compared with other Enterobacteriaceae [55, 56, 64]. Moreover, no significant difference in hospital mortality in non-ICU patients has been found between E. coli and K. pneumoniae (ESBL E. coli 23.8% vs ESBL K. pneumoniae 27.1%, p = 0.724) [65]. Once again, the role of the bacteria itself or of the host’s condition is still unclear [56]. ESBL-E carriage, a fortiori for K. pneumoniae, could be a major reflection of the host frailty. The fact that adequate empirical therapy does not have a clear impact on ESBL-E infections contrary to the host’s condition supports this hypothesis [66‐69]. The more recent study on the field is the only randomized controlled trial available and did not manage to prove the non-inferiority of piperacillin-tazobactam compared with meropenem for the treatment of BSI due to 3CG-resistant E.coli or K. pneumoniae [69]. Nevertheless, despite the quality of this study, some limitations apply to the conclusions of this study which should be interpreted cautiously [70]. Further studies are urgently needed to address this issue. Therefore, systematic carbapenem use among subsequently infected ESBL-E carriers could be questioned. In that case, main alternative candidates would be combinations such as piperacillin-tazobactam, new beta-lactams as ceftolozane-tazobactam or ceftazidime-avibactam, and combination with aminoglycosides [71‐74]. Those strategies need to be validated but would allow carbapenems spare without the need of systematic ESBL-E fecal carriage screening.

Selective decontamination is an attractive approach but, as previously described, its efficacy for ESBL-E decontamination still needs to be proven. Moreover, some concerns are already raised about the emergence of resistance to used antimicrobials such as colistin and tobramycin [75‐77]. One of these studies even showed the emergence of ESBL-E when 3CG are used for selective decontamination [77]. Regarding a non-ICU setting, the major study was a single-center trial with 54 ESBL-E carriers involved which assessed colistin and neomycin plus nitrofurantoin to successfully decolonize patients during the treatment but not 7 days after treatment cessation. Furthermore, no clinical outcome was reported [78].

Another approach to eradicate ESBL-E fecal carriage would be to cope with the effects of critical illness and antimicrobial therapy on gut. In fact, major gut microbiota modifications occur during hospitalization, a fortiori in ICU, and some of these modifications are associated ESBL-E fecal carriage [79, 80]. Cases of FMT to eradicate ESBL-E carriage have been reported [81, 82]. This approach is now being investigated in non-ICU patients. A proof-of-principle study included 15 patients to receive one FMT (among whom 7 received a second one) and suggested that FMT could be an effective treatment in patients carrying ESBL-E with several possible factors of response to therapy, such as donor-recipient microbiota match and number of FMTs [83]. A randomized, open-label, superiority trial in four tertiary care centers showed a non-significant decolonization success of 1.7 [95% CI 0.4–6.4] for a 5-day course of oral antibiotics followed by FMT, but the study failed to achieve the planned sample size for logistical and regulatory reasons making firm conclusions regarding efficacy difficult [84]. Moreover, fecal transplantation is not always well accepted by the patients and their siblings and concerns exist about gut permeability among critically ill patients [85]. A “soft decolonization” approach by probiotics could be a more feasible option. Nevertheless, this field has to be considered with caution and remains controversial, since there are studies with some positive effects, but also many negative studies. A major limit of previous studies is that used probiotics were “ready-to-use.” A promising but still experimental approach is to tailor the probiotics by indication or even by patient. A first step forward is the identification of a four-bacteria consortium which eradicates gut colonization with vancomycin-resistant Enterococcus faecium in a murine model [86]. Regarding ESBL-E colonization, a recent study identified Bacteroïdes uniformis as a possible candidate [87]. FMT and “soft decolonization” by tailored probiotics have to be further investigated and shown to be effective and safe among ICU patients before it can be used in routine. Therefore, systematic ESBL-E fecal carriage screening cannot be indicated to guide non-validated therapies.