Long-term use of selective digestive decontamination in an ICU highly endemic for bacterial resistance

Selective digestive decontamination (SDD) is a prophylactic treatment for critically ill patients that is based on an oropharyngeal paste and enteral suspension containing antimicrobials, usually tobramycin, colistin, and an antifungal as well as an intravenous antibiotic, administered during the first 4 days of intensive care unit (ICU) admission (usually a second-generation cephalosporin). The aim of SDD is to prevent or eradicate, if present, the oropharyngeal and intestinal abnormal carriage of potentially pathogenic microorganisms, such as aerobic gram-negative bacilli (AGNB), methicillin-sensitive Staphylococcus aureus, and yeasts, in patients at risk for nosocomial infections [1, 2]. Once a patient has been successfully decolonized, the unaffected anaerobic flora would offer prevention against new colonization with potential pathogenic microorganisms. In critically ill patients, SDD has been proven to prevent severe infections [1‐3] and to reduce mortality [3, 4], particularly in settings with a low prevalence of multidrug-resistant bacteria. However, the use of SDD is still a matter of debate, largely because of concerns that it may promote the emergence of antibiotic-resistant strains [5, 6]. Also, the effect of SDD in ICUs with endemic circulation of multidrug-resistant gram-negative bacilli MDR-GNB) remains controversial [7, 8]. We investigated whether long-term use of SDD was efficacious in reducing ICU-acquired MDR-GNB infection and also sought to determine its effect, including colistin- and tobramycin-resistant colonization as well as other nosocomial infections and subsequent antibiotic consumption, in a mixed ICU with a high endemic level of multidrug-resistant bacteria (MDRB).

During the 5-year study period, 3948 critically ill patients were admitted to the ICU, and ICU-acquired infection (VAP, CLABSI, secondary BSI, urinary tract infection) was diagnosed in 368 of them (7.8%). Of a total of 994 patients admitted to the ICU between September 2010 and September 2011, 110 patients had ICU-acquired infection in the non-SDD cohort. Of the 3948 patients admitted between October 2011 and September 2015, SDD was administered to 1998 (50.6%), and 258 developed an ICU-acquired infection (SDD cohort) (Fig. 1). No complications related to the use of SDD were recorded.

Results of univariate analysis are shown in Table 1. Demographic data and the distribution of most variables were similar in both cohorts. In the non-SDD cohort, the percentage of patients with chronic obstructive pulmonary disease and CLABSI was significantly lower than in the SDD cohort. However, we observed significantly lower rates of infections caused by MRDB, including Acinetobacter spp., other GNB and ESBL-producing multidrug-resistant bacteria, VAP, and secondary BSI, in the SDD cohort than in the non-SDD cohort. A significantly higher number of patients with CLABSI in the SDD cohort than in the non-SDD cohort was found. ICU-acquired infections caused by C. difficile or vancomycin-resistant Enterococcus spp. did not occur. In the multivariate analysis, SDD was found to be a protective factor against ICU-acquired infections caused by Acinetobacter spp. and MDR-GNB (Table 2). In the multivariate logistic regression model for MDRB infection, renal replacement therapy (OR, 2.130; 95% CI, 1.346–3.372; P = 0.001) was an independent risk factor for MDRB infection, whereas SDD was a protective factor (OR, 0.491; 95% CI, 0.305–0.790; P <  0.001).

Treatment with SDD was associated with a significant reduction of the RR for ICU-acquired infections caused by MDRB, VAP, and secondary BSI (Table 3). The probabilities of acquiring infections caused by MDRB, VAP, and secondary BSI were 69%, 57%, and 66% lower, respectively, in the SDD cohort than in the non-SDD cohort.

The consumption of nine antimicrobial agents commonly used in critically ill patients for treating MDRB, expressed as defined daily dose per 100 bed-days in the ICU, also showed a marked reduction after implementation of the SDD prophylactic strategy (Table 4). During the study period, other maneuvers directed toward reducing the use of antimicrobials were not applied.

Of a total of 3948 patients admitted to the ICU during the 4-year period of implementation of the SDD treatment, 285 showed surveillance samples colonized by colistin- or tobramycin-resistant pathogens. As shown in Table 5, there were increases of colonization resistance to colistin and tobramycin at ICU admission. Also, as shown in Table 5, the estimated rates adjusted to 100 patients with SDD decreased in the fourth year for tobramycin-resistant colonization and increased from 1.6 to 1.8 for colistin-resistant colonization in the third and fourth years of the study. The colistin- and tobramycin-acquired increasing rates of colonization resistance in the ICU by 1000 days and adjusted by the rate of resistances at admission were 0.82 (95% CI, 0.56 to 1.95; not statistically significant [NS]) and 1.13 (95% CI, 0.75 to 1.70; NS), respectively. The highest estimated rates of colistin- and tobramycin-resistant colonization by 1000 days in the ICU were 1.2 and 1.1 per 1000 days of ICU stay, respectively (Table 6). A summary of the study findings is shown in Fig. 2.

We found a significant reduction of ICU-acquired infections caused by MDR-GNB following SDD in our ICU with a high level of antibiotic resistance before implementation of the SDD strategy. There is limited information on the effects of SDD in settings with high levels of MDRB. Four observational studies [7, 21‐23] and one small randomized controlled trial [8] have been performed in ICUs where MDR-GNB were endemic or that had an outbreak of certain species of MDR-GNB. In these studies, SDD was applied either as a systematic treatment [21‐23] or as a targeted approach for identified carriers [7, 8]. Most of these previous studies examined the effect of SDD on elimination or persistence of carriage of resistant strains, but ecological outcomes were not reported. Moreover, heterogeneity regarding settings and designs prevented clear interpretation of the findings; in fact, SDD was found to be useful in three studies [7, 21, 23] and failed in two of them [8, 22]. Brun-Buisson et al. [7] reported that SDD reduced colonization or carrier status and infection during an outbreak of ESBL-producing Klebsiella pneumoniae. Our study confirms that SDD can be useful in an environment with high levels of MDR-GNB.

However, the present results are in accord with data of studies carried out in settings with low levels of antibiotic resistance, including findings of systematic reviews of randomized controlled trials [3, 4] and long-term observational studies [24‐27], confirming that SDD does not increase resistance. We also observed a significant reduction of infections caused by ESBL-producing MDRB. Similarly, Saidel-Odes and coworkers [28] reported that SDD decreased intestinal overgrowth of carbapenem-resistant K. pneumoniae. Zandstra et al. [29] also found that SDD is efficacious in controlling colonization with ESBL-producing bacteria, and Tascini et al. [30] showed that oral administration of gentamicin decontaminated the gastrointestinal tract and prevented infections caused by carbapenem-resistant K. pneumoniae strains producing K. pneumoniae carbapenemase (KPC)-type β-lactamase.

We also found a significant reduction of the incidence of infections caused by Acinetobacter baumannii and MDR-GNB. Similarly, in a randomized controlled study of 934 patients admitted to a surgical and medical ICU, of whom 466 were assigned to SDD and 468 to standard treatment (control subjects), colonization with gram-negative bacteria resistant to ceftazidime, ciprofloxacin, imipenem, polymyxin E, or tobramycin occurred in 16% of SDD patients and in 26% in the control group (P = 0.001) [31]. In a crossover study using cluster randomization in 13 ICUs in the Netherlands, the rate of isolation of gram-negative bacteria from rectal swabs was lower with SDD than with selective oropharyngeal decontamination (SOD) [4]. Also, SDD, as compared with standard care, was associated with a reduction of 57% of ICU-acquired bacteremia caused by glucose-nonfermenting gram-negative rods (P. aeruginosa, Stenotrophomonas maltophilia, and Acinetobacter spp.) and of 81% by Enterobacteriaceae, and these reductions were not accompanied by increases in intrinsic MDR-GNB colonization or infection [4]. A further analysis showed that development of ICU-acquired bacteremia caused by highly resistant microorganisms was 59% less frequent with SDD than with standard care and 63% less frequent with SDD than with SOD [32]. Recently, Camus et al. [33] found that the incidence rate of multidrug-resistant AGNB was lower during SDD (1.59 per 1000 patient-days versus preintervention 5.43%; P <  0.001) and also declined with time, concluding that a decontamination regimen did not favor the emergence of multidrug-resistant AGNB. In agreement with other studies, infections caused by C. difficile [31] and vancomycin-resistant Enterococcus spp. [34] were not registered.

The use of SDD resulted in a significant reduction of VAP, which is consistent with previous observations. In a systematic review of randomized controlled trials of antibiotic prophylaxis in 6914 ICU patients collected from 36 trials, there was a significant reduction of respiratory tract infections in the treated group (OR, 0.28; 95% CI, 0.65 to 0.87) [3]. Also, in a study of 4945 mechanically ventilated patients admitted between 2005 and 2013, the incidence of VAP per 1000 ventilator days declined significantly from 4.38 ± 1.64 before to 1.64 ± 0.43 after introduction of SOD/SDD in December 2010 (P = 0.007) [35]. Implementation of SDD as the standard of care in ICUs is thus effective in preventing VAP.

A further remarkable finding of the study was a significant reduction of secondary BSI associated with the use of SDD. In a randomized study involving 16 Dutch ICUs, the proportion of ICU-acquired bacteremia by Enterobacteriaceae was lower for SDD than for SOD (OR, 0.38; 95% CI, 0.26 to 0.55; P <  0.001) [1]. In a systematic review of 51 randomized controlled trials conducted between 1987 and 2005, comprising 4079 patients treated with SDD and 3986 control subjects, SDD was associated with a reduction of overall and gram-negative BSIs of 27% and 61%, respectively, without affecting gram-positive BSIs [2]. Furthermore, prophylactic treatment with SDD was a protective factor for infections caused by MRDB. In a systematic review and meta-analysis of 64 studies assessing the effect of SDD and SOD on antimicrobial resistance, no differences were found in the prevalence of colonization or infection with gram-positive antimicrobial-resistant pathogens (MRSA, vancomycin-resistant enterococci) and gram-negative bacilli resistant to aminoglycosides and fluoroquinolones [36]. However, there was a reduction in polymyxin-resistant and third-generation cephalosporin-resistant gram-negative bacilli in recipients of SDD compared with those who did not receive the intervention. According to these data, the perceived risk of long-term harm related to SDD cannot be justified. The authors also conclude that the effect of SDD on ICU-level antimicrobial resistance rates is probably understudied. However, emergence of antimicrobial resistance is still a main objection to the widespread use of SDD in ICUs [5, 6, 8].

Also, there is a controversy regarding the emergence of an increased resistance to colistin and tobramycin used as part of SDD. We found low rates of colistin- and tobramycin-resistant colonization in cultures of surveillance samples during the 4-year SDD. It is known that there may be nosocomial transmission of highly resistant microorganisms from one patient infected to another, with or without SDD, and that this can increase the number of patients with GNB-resistant colonization [37]. As shown in Table 5, there are increases of colonization resistance to colistin and tobramycin at ICU admission. Also, the estimated rates adjusted to 100 patients with SDD decreased in the fourth year for tobramycin-resistant colonization and showed a small increase from 1.6 to 1.8 for colistin-resistant colonization in the third and fourth years of the study. The colistin- and tobramycin-acquired increasing rates of colonization resistance in the ICU by 1000 days and adjusted by the rate of resistances at admission were 0.82 (95% CI, 0.56 to 1.95; NS) and 1.13 (95% CI, 0.75 to 1.70; NS), respectively. These findings mean that although there were increases in the rates of colistin- and tobramycin-resistant colonization, these increases could not be associated with SDD and may have been linked to the progressive rise of MDR-GNB at ICU admission over the 4 years of the study and also may have been due to a higher degree of nosocomial transmission of highly resistant microorganisms among ICU patients. The highest estimated rates of colistin- and tobramycin-resistant colonization by 1000 days at risk were 1.2 and 1.1 per 1000 days, respectively (Table 6).

Colistin- and tobramycin-resistant colonization rates in our study were lower than 2.5/1000 patients days at risk, as shown in the study by Oostdijk et al. [38]. Using two large cohorts of ICU patients, Oostdijk et al. demonstrated that the prolonged use of colistin, as part of SDD and SOD, was not associated with increased acquisition of colistin-resistant GNB in the respiratory tract. Moreover, acquisition rates of colistin-resistant GNB in the intestinal tract during SDD ranged from 1.2 to 3.2 per 1000 patient-days at risk. The overall conversion rate from colistin susceptibility to resistance in the intestinal tract was below 1 conversion per 1000 patient-days at risk. During SDD, though, these conversion rates ranged from 3.2 to 5.4 per 1000 days of colonization with GNB and from 15.5 to 12.6 per 1000 days of colonization with tobramycin-resistant GNB. Also, the use of meropenem appeared to be strongly associated with the development of meropenem resistance in P. aeruginosa with an adjusted HR of 11.1 (95% CI, 2.4–51.5), corresponding to 23 events of resistance acquisition per 1000 patient-days at risk. [39]. On the basis of these findings, we concluded, as Oostdijk et al. [38] did, that the rates of resistance acquisition for frequently used antibiotics were considerably higher than for acquisition of colistin resistance during topical use of this agent.

Our findings differ from those of previous studies showing no increase in acquisition of resistant flora to these agents over a 5-year period [24] or no increases in the prevalence of resistance against colistin and tobramycin among gram-negative isolates during a mean of 7 years of SDD or SOD use [40]. Noteboom et al. [41] also observed that the percentages of antibiotic resistance with SDD and standard care were similar.

However, in a short course of SDD with colistin and gentamicin during an outbreak due to a KPC-2-producing K. pneumoniae strain, development of secondary resistance to colistin (19% increase in resistance rate) and gentamicin (45% increase) was found [8]. Halaby et al. [5] reported a significant relationship between use of SDD and tobramycin resistance as well as resistance to colistin among ESBL-producing pathogens. Brink et al. [6] showed the emergence of KPC in Enterobacteriaceae and the selection of strains resistant to colistin. Of note, Silvestri et al. [42], regarding data reported by Brink et al. [6], argued that an inadequate dose of enteral antimicrobials in the SDD protocol was responsible for the failure of K. pneumoniae to decolonize and eventually become resistant to colistin. Failure associated with subtherapeutic doses of SDD may cause overgrowth of MDR-GNB, with increased spontaneous mutation leading to polyclonality and resistance [43].

Associations between prolonged intravenous colistin use and development of colistin resistance have been reported from settings with high levels of carbapenemase-producing GNB [44, 45]. In contrast to facilitating resistance, SDD has been used successfully as a control measure in outbreak situations with ESBL-producing GNB [7, 46]. High intraluminal levels of topical antibiotics exceed minimum inhibitory concentrations of resistant pathogens, leading at least to temporary suppression, which reduces the risk of overgrowth and cross-transmission. However, there are several factors aside from SDD that produce GNB-resistant colonization. We did not find any MDR-GNB susceptible only to colistin in our study. Also, we observed decreased ICU global mortality over the course of the 4-year application of SDD.

Nevertheless, we think that SDD must be accompanied by careful monitoring of tobramycin and colistin resistance in GNB. We do so, as described in our protocol. We recommended screening weekly throughout the ICU stay.

SDD in an ICU setting with a high level of resistance was associated with a clinically relevant reduction of infections caused by MDRB, with low rates of colistin- and tobramycin-resistant colonization and a nonsignificant increasing rate of ICU colonization resistance by 1000 days, adjusted by the rate of resistance at ICU admission. SDD was also a protective factor against MDRB infection. Furthermore, VAP and secondary BSI were significantly decreased after SDD. Notably, a decrease in antimicrobial consumption was also observed.