The ecological effects of selective decontamination of the digestive tract (SDD) on antimicrobial resistance: a 21-year longitudinal single-centre study

This study has a single-centre longitudinal observational cohort design. The patient data are registered in a local database since 1997 and retrospectively extracted for this analysis. The study is conducted in a Dutch 20-bed adult mixed medical, surgical and cardiac surgery tertiary intensive care unit. All consecutive patients admitted to the ICU between January 1997 and July 2017 were included, and clinical data were extracted. All available cultures and patient data were used. Patients without cultures taken during ICU treatment could not be counted in the culture data but are included in the other analyses. Data on baseline characteristics of all included patients were prospectively recorded in the ICU database and extracted for this analysis.

The local medical ethical review board (ACWO OLVG) approved the study and waived informed consent due to its retrospective design in accordance with Dutch and European legislation (study no. WO 13.063).

Data on all cultures taken during ICU stay of the included patients were over the years registered in the hospital database and retrospectively retrieved from this database. Both surveillance cultures (throat, rectum, tracheal aspirate) and cultures of organ sites and blood taken during ICU stay were included for analysis. Routinely, cultures from the throat and rectum were taken twice a week and tracheal aspirate three times a week for surveillance in patients treated with SDD. For non-intubated patients, the tracheal aspirate was substituted for sputum when available. All culture samples taken in the context of surveillance were plated on an unselective blood agar and four specific agars selecting for gram-positive bacteria, gram-negative bacteria, yeast and vancomycin-resistant enterococci. Gram-negatives were cultured on McConkey agar; Staphylococcus aureus was detected using both unselective blood agar and Columbia aztreonam polymyxin agar. For yeasts, Sabouraud maltose agar was used. Cultured gram-negative bacteria were tested for antimicrobial susceptibility using agar disk diffusion during the entire study period. Cutoff values for resistance were set following the guidelines of the ‘Clinical and Laboratory Standards Institute’ (CLSI) [13] until July 2011 and ‘the European Committee on Antimicrobial Susceptibility Testing’ (EUCAST) from July 2011 until 2017 [14]. Cutoff values for resistance for colistin were performed according to CLSI ranges, for the whole study period.

Patients were treated with SDD when the expected ICU stay was more than 24 h, with or without mechanical ventilation. This relatively liberal use of SDD is based on the concept that acquisition of potential pathogenic microorganisms starts in the first day of ICU stay [15, 16]. SDD has been consistently used since 1986 following the same protocol. The SDD regimen consists of four times daily Orabase® oral paste with 2% polymyxin B, amphotericin B and tobramycin. In addition, 10 ml of a suspension containing 500 mg amphotericin B, 100 mg polymyxin B and 80 mg tobramycin is administered four times daily in the gastric tube or swallowed in patients without a gastric tube. Cefotaxime is administered four times daily 1 g i.v. for 4 days in all SDD-treated patients but is administered longer in case of active infection. For active infection, the coverage of gram-negatives can be extended with ciprofloxacin i.v. or tobramycin i.v. when needed. In case of peritonitis, metronidazole is added as well. Other i.v. antimicrobials can be given based on previous culture results. Penicillins are carefully avoided whenever possible due to their negative effects on the normal aerobic and anaerobic intestinal flora which could lead to a loss of the protective effect to invading pathogens (colonization resistance). Methicillin-resistant staphylococci (MRSA) colonization and infection were consistently treated with vancomycin i.v. and 2% vancomycin added in the oral paste and gastric suspension. When a combined polymyxin and tobramycin resistance was present in AGNB from surveillance cultures, then co-trimoxazole 2% was added in the oral paste and twice daily 960 mg in the enteral suspension when appropriate and until decontamination was obtained [15].

Cultures with aerobic gram-negative potentially pathogenic bacteria (AGNB) were selected. Among others, these AGNB are Acinetobacter spp., Citrobacter spp., Escheria coli, Enterobacter spp., Klebsiella spp., Morganella spp., Proteus spp., Pseudomonas aeruginosa, and Serratia marcescens [16]. To obtain the resistant strains for analysis, we selected in every patient the AGNB with antimicrobial resistance for tobramycin/gentamicin, polymyxin B/colistin, cephalosporins or ciprofloxacin and counted the first culture with this particular resistance. The strains were defined as resistant when they were tested intermediate or resistant in vitro. Strains that were resistant on admission were marked as not ICU-acquired and were excluded for the calculation of the incidence rates but they were analyzed separately to calculate the background resistance rates on admission. Resistant pathogens were considered ICU-acquired if these pathogens were absent in the first 48 to 72 h after admission but present in subsequent cultures. In case a patient enters the ICU with resistant bacteria and subsequently acquires other resistant bacteria, they are counted in both groups. Because of variability in time of admission and time that the culture was taken, the cultures taken on day three are considered within 48 to 72 h after admission. As a consequence, all resistant pathogens from cultures taken on day of admission (day one), day two or day three were considered resistant on admission and not ICU acquired. This is consistent with the definition of ICU-acquired infection in ‘the ECDC HAIICU protocol for healthcare associated infections in intensive care units’ [17].

The incidence rates for ICU-acquired resistance were calculated for each year for second- or third-generation cephalosporins (including cefotaxime, ceftriaxone, cefuroxim and ceftazidim), polymyxin B/colistin, tobramycin/gentamicin and ciprofloxacin separately. Microorganisms that were intrinsically resistant to one of the antimicrobial agents were excluded for the analysis of that specific agent. For third-generation cephalosporins, we excluded Serratia marcescens, Acinetobacter spp., Citrobacter freundii, Enterobacter spp. and Morganella morgannii, and for polymyxinB/colistin, Serratia marcescens, Proteus spp. and Morganella spp. were excluded for this calculation. Incidence rates were calculated instead of cumulative incidence because of the dynamic population [18]. The incidence rate was defined as:

[Number of ICU-acquired resistant PPMs/sum of the person-time of the population at risk] × 1000.

Population at risk was defined as all patients admitted to the ICU during the study period without resistant strains on admission. Patients who were treated less than 48 h in the ICU were excluded from the denominator as they could not have acquired new (multi)resistant strains in this short period. The person-time, calculated in patient-days per year, per patient at risk is defined as the number of days between ICU admission and discharge from the ICU. The number of days was calculated without correcting for the time of admission or discharge. The day of admission is ‘day one’ and so forth. In case of acquisition of a resistant potentially pathogenic microorganism (PPM), person-time was defined as the number of patient-days between ICU admission and acquisition of this resistant PPM. The total number of patient-days for patients at risk was calculated for every antimicrobial agent separately. Incidence rates were calculated for patients with an ICU length of stay more than 48 h and for all patients (Additional files 1, 2, 3 and 4), for medical and surgical patients (Additional files 1, 2, 3 and 4).

To determine background resistance rates, the prevalence of resistance on admission in AGNB was calculated for each year separately. The prevalence was calculated by dividing the number of resistant strains on admission by the number of admissions in that particular year. All data were analyzed using Statistical Package of Social Science (SPSS 24.0 Chicago, IL, USA).

Between January 1, 1997, and July 29, 2017, 15,051 patients were admitted to the ICU. These patients had 17,511 admissions with cultures taken during ICU stay. Baseline characteristics are shown in Table 1. In the second column, all patients with an ICU length of stay less than 48 h are excluded as they could not acquire new colonization with resistant strains in this short period. Over time, the case mix did not change substantially with around half of the population having cardiovascular disease and the other mixed diseases including sepsis and abdominal surgery. The distribution of sample sites is also shown in Table 1. The majority of the cultures were surveillance cultures from the throat or rectum (41.9%) or cultures retrieved from tracheal aspirate (23.3%).

Figure 1 shows the number of cultures and number of resistant strains. A total of 20,593 isolates of AGNB were identified. These included Proteus spp. (N = 2857), Klebsiella spp. (N = 1634), Morganella spp. (N = 759), Enterobacter spp. (N = 1571), Citrobacter spp. (N = 529), Serratia marcescens (N = 1081), Pseudomonas spp. (N = 5269), Acinetobacter spp. (N = 485) and E. coli (N = 6409).

For the second- or third-generation cephalosporins, the number of isolates that were tested for resistance increased from 78.4% in 1997 to 99.2% in 2017. A higher test rate leads to a better detection of resistant microorganisms. Susceptibility for ciprofloxacin was tested for 78.4% in 1997 and 99.2% in 2017. In 65.8% of isolates, information on susceptibility to polymyxin B/colistin was available. The number of tested isolates for polymyxin B/colistin rose from 22.7% in 1997 to 99.0% in 2017. Information on susceptibility for aminoglycosides was available in 92.9% of all isolates. This number varied between 78.8 and 99.2% over the years. More information on susceptibility testing per year for the different antimicrobial agents is shown in Additional files 1, 2, 3 and 4.

The incidence rate per 1000 patient-days per year is shown in Fig. 2. These data reflect the population of ICU patients with a length of stay more than 48 h. In Additional files 1, 2, 3 and 4, the data including the patients with a shorter length of stay are shown. The average incidence rate for ICU-acquired resistance to third-generation cephalosporins was 2.03 per 1000 patient-days (range 0.55/1000 patient-days in 2001 to 4.3/1000 patient-days in 2012), 0.51 per 1000 patient-days for polymyxin B/colistin (range 0.00/1000 patient-days in 1997 to 2.12/1000 patient-days in 2010), 2.59 per 1000 patient-days for tobramycin (range 1.09/1000 in 2004 to 4.08/1000 patient-days in 2012) and 2.20 per 1000 patient-days for ciprofloxacin (range 0.26/1000 patient-days in 1998 to 4.94 per 1000 patient-days in 2009. The results of the negative binomial regression analysis are listed in Table 2.

The incidence rates for medical and surgical patients have been calculated separately. Additional files 3 and 4 Medical and Surgical show the incidence rates over time for these two groups. Basically, it is shown that in both groups the incidence rates are low with some variability over the years.

The prevalence rates of resistant aerobic gram-negative PPMs on admission per year are presented in Fig. 3, and this represents the ‘background’ prevalence. Overall prevalence rate of resistance to cephalosporins was 6.7 per 100 ICU admissions. This was 0.8/100 admissions for polymyxin B/colistin, 5.6/100 admissions for tobramycin and 7.6/100 admissions for ciprofloxacin. Figure 3 shows an increase in background resistance for cephalosporins, ciprofloxacin and tobramycin but for polymyxin B/colistin this trend is not seen.