From May to October 1988, A. baumannii was isolated from 26 patients in an intensive care unit. During that period, numerous cultures of potential environmental reservoirs were done, and A. baumannii was isolated from the hands of the radiographer, the on-site radiography machine, and the film holder. Therefore, the radiography machine was rigorously cleaned and the film holders were put in a single-use bag before use. At first, these measures seemed to have been effective because A. baumannii was not subsequently isolated from patients in the intensive care unit. In January 1989, however, A. baumannii was again isolated from clinical specimens. This recurrence of infection led to this epidemiologic study.

In this study, the combined use of molecular typing methods and case–control studies was crucial for elucidating the epidemiology of nosocomial A. baumannii infections and for identifying the role played by the use of antimicrobial agents in the persistence of multidrug-resistant clones.

Methods

Top Methods Results Discussion Author & Article Info References

Setting

The Hotel-Dieu Hospital is a 1200-bed, tertiary care teaching hospital that offers medical and surgical acute care services to a population of 600 000 persons. Approximately 50 000 patients are admitted to this hospital each year. The intensive care unit is a 20-bed medical and surgical facility to which about 700 patients are admitted each year. Seventy percent of these patients are admitted from either the emergency department or the emergency operating room, and their mean duration of stay in the intensive care unit is about 10 days.

Selection of Study Participants

A case-patient was defined as any patient in the intensive care unit from whom A. baumannii was isolated between 1 January 1989 and 31 December 1993. To identify cases of A. baumannii infection, we reviewed the records of the microbiology laboratory for positive cultures for A. baumannii from 1 January 1989 to 31 December 1990 (epidemic and endemic case–control studies) and from 1 January 1991 to 31 December 1993 (follow-up case–control study).

A total of 72 controls were selected by simple random sampling from among all patients who were hospitalized in the intensive care unit during the epidemic period and had negative cultures for A. baumannii.

Microbiological Investigation

Acinetobacter baumannii isolates were presumptively identified by using morphology of the colonies, Gram staining, oxydase and catalase reactions, growth at 44°C, and the API-20 NE System (Bio-Merieux, Lyon, France) [7]. Identification as A. baumannii was verified by restriction analysis of the 16S-23S ribosomal RNA intergenic-spacer sequences, as described by Dolzani and colleagues [8].

All isolates were tested by using the disk-diffusion method for susceptibility to ß-lactams (amoxicillin, amoxicillin-clavulanate, ticarcillin, piperacillin, aztreonam, cephalothin, cefotaxime, ceftazidime, and imipenem), aminoglycosides (amikacin, dibekacin, gentamicin, netilmicin, and tobramycin), and fluoroquinolones (ciprofloxacin, norfloxacin, ofloxacin, and pefloxacin) according to the recommendations of the Comite Francais de l'Antibiogramme [9].

Ribotyping

The A. baumannii isolates were grown overnight in Luria broth at 37°C, and DNA was extracted as described elsewhere [10, 11]. Three µg of extracted DNA was digested by EcoR1 and subjected to electrophoresis, and restriction fragments were transferred on nylon membrane H⁺ for hybridization (Boehringer Mannheim France SA, Meylan, France). The riboprobe pKK3535, a derivative of pBR322 containing the rrnB ribosomal RNA operon of Escherichia coli, was labeled with digoxigenin 11 2'-deoxyuridine 5'-triphosphate by random priming done using the DIG DNA labeling kit (Boehringer Mannheim France SA) [12]. Prehybridization, hybridization, and detection were done as described in the DIG System Users Guide to filter hybridization (Boehringer Mannheim France SA). The detection was done with the DIG luminescent detection kit (Boehringer Mannheim France SA) on radiographic film RPN2103 (Amersham France SA, Les Ulis, France).

Repetitive Polymerase Chain Reaction

The oligonucleotide primer ERIC II was used to amplify variable-length regions between the interspersed repetitive elements [13]. Amplification reactions were done in a volume of 50 µL containing 10 mmol of TRIS HCl each per L (pH, 9), 50 mmol of KCl per L, 0.2 µmol per L of deoxynucleotide, 0.3 µmol of primer per L, 3 mmol of MgCl₂ per L, 2.5 U of Taq polymerase (Gibco BRL, Life Technologies SARL, Cergy Pontoise, France), and 10 ng of target DNA. After denaturation was done for 90 seconds at 94°C, amplification mixtures were submitted to 44 cycles that consisted of 15 seconds of denaturation at 94°C, 15 seconds of annealing at 36°C, and 70 seconds of extension at 72°C, with a final extension of 7 minutes at 72°C. Fifteen µL of the polymerase chain reaction (PCR) product were size-fractionated on agarose gel and stained with ethidium bromide.

Case-Control Methods

To assess potential risk factors associated with A. baumannii infections, we did three case–control studies. In the first (the epidemic case–control study), the 11 case-patients whose isolates either shared the ribotyping profile R1 or (if the isolates had not been kept) were resistant to ticarcillin were compared with the 72 controls. In the second case–control study (the endemic case–control study), the 17 case-patients whose isolates differed by ribotyping profile (R2, R3, R4, R5, R6) or antimicrobial susceptibility profile were compared with the 72 controls. In the third case–control study (the follow-up study), the 17 case-patients who developed A. baumannii infection between 1 January 1991 and 31 December 1993 were compared with the 51 controls who were randomly selected during the same period. We compared hospitalization of case-patients with that of controls whose hospitalizations lasted at least 14 days in the epidemic and endemic case–control studies and at least 17 days in the follow-up study (14 days and 17 days were the mean durations of case-patient hospitalization before collection of the first positive A. baumannii culture). The following data were abstracted from the medical records of case-patients and controls: age, sex, admission diagnosis (medical or surgical), severity of illness at admission (according to the simplified acute physiologic score [SAPS]), outcome, duration of stay in the hospital and the intensive care unit, exposure to peripheral or central intravenous catheters, use of arterial catheters, hyperalimentation, tracheotomy, mechanical ventilation, antimicrobial agents received after hospital admission, and radiologic and surgical procedures within or outside of the intensive care unit [14].

Statistical Analysis

Data were analyzed by using the Epi-Info computer database (version 5, Centers for Disease Control and Prevention, Atlanta, Georgia). The Fisher two-tailed exact test and the Student t-test were used to test for the significance of association. Biologically plausible risk factors for A. baumannii infection were evaluated by logistic regression analyses; factors significant at the level of P = 0.1 were retained in the model (SAS, version 6.12, SAS Institute, Inc., Cary, North Carolina).

Results

Top Methods Results Discussion Author & Article Info References

Description of Case-Patients

From 1 January 1989 to 31 December 1990, A. baumannii isolates were recovered from 28 patients. These 28 case-patients ranged in age from 13 to 81 years (mean age ±SD, 40.8 ± 20.7 years; median age, 38.5 years); 19 (68%) were male. Fourteen (50%) were surgical patients, and 1 was immunocompromised. The mean SAPS was 15 ± 4.6 (median SAPS, 14 [range, 7 to 24]); 26 case-patients (93%) were intubated and ventilated, and 27 (96%) had perfusion through a central venous line. The mean duration of stay in the intensive care unit was 28.9 ± 21 days (median, 26 days [range, 3 to 88 days]), and the mean interval between intensive care unit admission and first positive culture was 13.6 ± 8.9 days (median, 12 days [range, 0 to 34 days]). Acinetobacter baumannii was isolated from urine (n = 10), the lower respiratory tract (n = 5), blood (n = 2), wounds (n = 8), cerebrospinal fluid (n = 2), and the sinus (n = 1). No screening for colonization was done, and all case-patients were considered infected. Seven case-patients (25%) died during their stay in the intensive care unit, but their deaths were not due to A. baumannii infection.

Microbiological Investigation

Identification of the isolates as A. baumannii was retrospectively confirmed by restriction analysis of the 16S-23S ribosomal RNA intergenic-spacer sequences. Because comparison of the isolates done by using simple phenotypic markers (such as the biotypes or the susceptibility profiles) showed some heterogeneity among the isolates, we decided to ribotype the 19 isolates that had been kept before beginning the clinical investigation.

Six profiles were obtained after digestion by EcoR1: Seven isolates had profile R1, 6 had profile R2, 1 had profile R3, 2 had profile R4, 2 had profile R5, and 1 had profile R6 (Figure 1). All R1 isolates were resistant to ticarcillin, whereas all of the other isolates were susceptible to this antimicrobial agent. Thus, susceptibility to ticarcillin was used as a marker to classify the strains that had not been kept and could not be typed by molecular methods. As a result, the R1 strains and the ticarcillin-resistant strains constituted the first group (epidemic), and the strains with other ribotyping profiles and the ticarcillin-susceptible strains made up the second group (endemic). The banding patterns obtained by repetitive PCR with ERIC II confirmed the results of the ribotyping by classifying the isolates in the same groups (Figure 2).

View larger version (84K): [in this window] [in a new window]   Figure 1. Ribotype patterns of Acinetobacter baumannii strains after EcoR1 digestion and hybridization with the labeled pKK3535 riboprobe. Lanes A, B, C, and D: Raoul 1 size marker; lanes 1 through 7: pattern R1 (epidemic strains); lanes 8 through 14: pattern R2 (endemic strains from the second and third case–control studies); and lanes 15 and 16: sporadic strains. Kb = kilobase pairs.

View larger version (83K): [in this window] [in a new window]   Figure 2. Repetitive polymerase chain reaction patterns of Acinetobacter baumannii after digestion with ERIC II. Lanes A, B, and C: 1-kilobase DNA ladder size marker. Lanes 1 through 16: Isolates are shown in the same order as in Figure 1. Bp = base pairs.

All isolates were resistant to amoxicillin, amoxicillin-clavulanate, and cefoxitin and were susceptible to imipenem. The 11 strains of the epidemic group were all susceptible to ceftazidime and amikacin and were resistant to ticarcillin-clavulanate, piperacillin, ciprofloxacin, tobramycin, and netilmicin. Nine percent were susceptible to aztreonam. Fifty-three percent of the isolates from the endemic group were susceptible to netilmicin, 24% were susceptible to amikacin and tobramycin, 18% were susceptible to ciprofloxacin and piperacillin, 12% were susceptible to ceftazidime, and 6% were susceptible to cefotaxime.

Clinical Investigation

Analysis of the molecular typing and susceptibility profiles led us to perform two case–control studies to separately assess the risk factors for each group of A. baumannii infections in patients in an intensive care unit. The same set of controls was used for both studies.

Epidemic Case-Control Study

Univariate analysis showed that case-patients were more likely than controls to have had a traffic accident (54% [6 of 11] compared with 23% [17 of 72]; odds ratio [OR], 3.8 [95% CI, 0.9 to 17.3]; P = 0.06), to be a surgical patient (81.9% [9 of 11] compared with 34.7% [23 of 72]; OR, 8.5 [CI, 1.5 to 63]; P = 0.006), to have undergone surgery in the emergency operating room before admission to the intensive care unit (63% [7 of 11] compared with 23% [17 of 72]; OR, 5.7 [CI, 1.3 to 27]; P = 0.01), and to be receiving arterial pressure monitoring (36% [4 of 11] compared with 7% [5 of 72]; OR, 7.7 [CI, 1.3 to 46]; P = 0.02).

The following variables were considered to be biologically plausible risk factors and were evaluated in a logistic regression model: age; sex; immunodepression; SAPS; having had a traffic accident; procedures undergone before admission to the intensive care unit, such as intubation, insertion of central venous and urinary catheters, and surgery in the emergency operating room; and procedures done during the stay in the intensive care unit, such as surgery and arterial pressure monitoring. Only one independent risk factor was identified: any surgical procedure (OR, 8 [CI, 1.5 to 41]; P = 0.01). However, we suspected from the start of the investigation that the emergency operating room was involved in the outbreak because 7 of the 9 epidemic case-patients who underwent a surgical procedure had it done in the emergency operating room. This room was used for surgery in contaminated and infected patients, as well as emergency and trauma patients, and it was impossible to isolate the infected patients. In an attempt to identify more-specific risk factors for use in implementing prevention strategies, we constructed a second model, excluding the variable "any surgical procedure." Only one independent risk factor was then identified: surgery in the emergency operating room (OR, 5 [CI, 1.2 to 19]; P = 0.007).

Endemic Case-Control Study

When case-patients were compared with controls by univariate analysis, only one factor was significantly associated with infections: previous receipt of a fluoroquinolone (pefloxacin, ofloxacin, norfloxacin, or ciprofloxacin) (53% [9 of 17] compared with 12.3% [9 of 73]; OR, 8 [CI, 2 to 31]; P < 0.001).

Initial Control Measures

The emergency operating room was closed in October 1990, before completion of the first epidemiologic study. It was therefore impossible to assess the mode of transmission of the epidemic strain of A. baumannii in this operating room. However, no epidemic strains have been isolated from patients in the intensive care unit since the new emergency operating room opened (Figure 3). Multiple samples were obtained from the new emergency operating room and from operating room personnel, and none of these samples grew A. baumannii.

View larger version (21K): [in this window] [in a new window]   Figure 3. Epidemic curve.Acinetobacter baumannii infections in an intensive care unit, 1989-1993. OR = operating room.

Because previous receipt of a fluoroquinolone was identified as the only risk factor and because most isolates were resistant to fluoroquinolones, a first attempt to reduce the use of these agents was made.

Follow-up Study

Because of the continuous occurrence of nosocomial A. baumannii infections or colonizations among patients in the intensive care unit (seven cases of pneumonia, three cases of bacteremia, four cases of urinary tract infection, and three cases of skin colonization) (mean interval between intensive care unit admission and first positive culture, 18.7 ± 10.7 days; median, 17 days [range, 6 to 39 days]) and the death from a lower respiratory tract infection in one patient, we did a third epidemiologic study in December 1993. When compared with 51 controls, case-patients were significantly associated (by univariate analysis) with previous receipt of a fluoroquinolone (64.7% [11 of 17] compared with 33.3% [17 of 51]; OR, 8.6 [CI, 2.1 to 36]; P < 0.001), alcoholism (47% [8 of 17] compared with 17.6% [9 of 51]; OR, 4 [CI, 1.1 to 16]; P = 0.02), and having a femoral catheter (35.3% [6 of 17 compared with 9.8% {5 of 51}; OR, 5 [CI, 1.1 to 24]; P = 0.02). The following variables were considered to be biologically plausible risk factors and were evaluated in a logistic regression model: age, sex, SAPS, having a femoral catheter, alcoholism, and previous receipt of a fluoroquinolone. By multivariate analysis, only previous receipt of a fluoroquinolone (OR, 8.6 [CI, 2.5 to 30]; P < 0.001) and alcoholism (OR, 4 [CI, 0.96 to 16]; P = 0.045) were independent risk factors for A. baumannii infection or colonization. Ribotyping and repetitive PCR were again done on all of the A. baumannii isolates that had been kept. Four strains from the follow-up study shared the same profile and were identical to R2 strains isolated from 1989 to 1991. All 17 isolates were resistant to fluoroquinolones.

Analysis of the relation between fluoroquinolone use and the incidence of A. baumannii infection showed a parallelism between these two events. Between 1991 and 1992, the use of fluoroquinolones decreased 42.5%, from 402 g to 231.3 g, and the incidence of A. baumannii infections per admission concomitantly decreased significantly, from 2.5% (16 of 644) to 0.8% (6 of 716) (relative risk, 0.32 [CI, 0.16 to 0.64]; P = 0.03) (Figure 4). The efficacy of the attempt to decrease fluoroquinolone use was short-lived; the amount of fluoroquinolones used increased from 231.3 g in 1992 to 438.8 g in 1993 while, during the same period, the incidence of A. baumannii infection per admission increased twofold, from 0.8% (6 of 716) to 1.6% (11 of 696). Subsequently, the incidence decreased again when the number of doses of fluoroquinolones given in the intensive care unit declined after implementation of a policy that prohibited the use of fluoroquinolones for empirical antimicrobial therapy and allowed their use only for the treatment of microbiologically documented infections. The assessment of the relation between the mode of administration of fluoroquinolones and the incidence of A. baumannii infection showed that the parallelism seen between these two events was due to the intravenous use of fluoroquinolones (Figure 5). From 1993 to 1994, the incidence of A. baumannii infection per admission decreased 12% (1.6% compared with 1.41%) when the amount of parenteral fluoroquinolones used decreased 32% (184.2 g compared with 125.6 g). At the same time, the number of oral fluoroquinolone units administered increased 11.5% (255 g compared with 284.2 g).

View larger version (26K): [in this window] [in a new window]   Figure 4. Amount of fluoroquinolones given (white bars) compared with incidence of Acinetobacter baumannii infections per admission (solid line) in an intensive care unit, 1991-1994.

View larger version (36K): [in this window] [in a new window]   Figure 5. Mode of fluoroquinolone administration compared with incidence of Acinetobacter baumannii infections per admission (solid line) in patients in an intensive care unit, 1991-1994. White bars represent oral administration of fluoroquinolone; striped bars represent intravenous administration of fluoroquinolone.

Discussion

Top Methods Results Discussion Author & Article Info References

In the past decade, A. baumannii has emerged as an important nosocomial pathogen that is often resistant to numerous antimicrobial agents and that causes life-threatening infections in patients with altered host-defense mechanisms. In addition, it has a tendency toward cross-transmission, especially in intensive care units, where numerous outbreaks have occurred [1, 15-17]. Previous studies have identified various risk factors for A. baumannii infection or colonization, including factors related to the host (severity of illness or history of infection), to hospitalization (long stay in the intensive care unit or hospital), to treatments (previous administration of third-generation cephalosporins), or to procedures (intubation, mechanical ventilation, indwelling urinary catheter placement, or neurosurgery) [4, 6, 18-20].

In addition, the development of molecular typing methods has given the clinical microbiology laboratory powerful tools, thus providing the means for a better knowledge of the epidemiology of bacterial infections. Many typing systems have been used to study the epidemiology of A. baumannii infections. They have shown the molecular heterogeneity of A. baumannii isolated within a hospital or service with the coexistence of epidemic and sporadic patterns, but no clear explanation has been offered for this phenomenon [15, 18, 20-30]. Furthermore, results obtained from the typing of isolates have seldom been used to conduct epidemiologic studies. Another reason why most epidemiologic studies have not been conclusive may be the complexity of the taxonomy of the genus Acinetobacter, this complexity requires the use of genotypic methods, such as those used in our study, to correctly identify the isolates [31]. In our study, this and the typing of the isolates were essential for comprehension of the epidemiology of A. baumannii infections because they led us to perform several case–control studies instead of one and to increase the sensitivity and specificity of the epidemiologic study.

Because not all isolates were kept, we were unable to type all isolates by using molecular methods. In our study, however, the combination of phenotypic and genotypic markers proved to be highly discriminating and showed the coexistence of different epidemiologic features [3]. The results of the first case–control study suggested that epidemic strains of A. baumannii may have been acquired by a patient before admission to the service where the infection was diagnosed; this may explain why it was difficult to control the outbreak. Unfortunately, because our study was retrospective, it was impossible for us to use observation of surgical procedures or environmental cultures to assess the mechanisms by which the epidemic strain may have been acquired in the emergency operating room. However, despite the limitations of the use of logistic regression analysis in such a small data set, the disappearance of the epidemic strain from the intensive care unit after the destruction of the operating room suggests that the operating room may have been the primary source of (or reservoir for) the epidemic strain and that secondary transmission may have occurred within the intensive care unit.

The use of antimicrobial agents is considered to be a predisposing factor for the emergence of resistant bacteria; until now, however, few data have supported the hypothesis of a causal relation between administration of and resistance to antimicrobial agents. Both the development of resistance de novo and the stepwise selection of mutants showing high-level resistance to fluoroquinolones during in vitro or in vivo exposure to these drugs have been reported and provide consistent biological support for the hypothesis of a causal relation between fluoroquinolone use and acquisition of fluoroquinolone resistance by previously susceptible organisms [32-35]. Our study showed that previous receipt of a fluoroquinolone was an independent risk factor for endemic A. baumannii infection and that the selection pressure caused by the indiscriminate use of fluoroquinolones was responsible for the persistence of multidrug-resistant clones over at least 5 years. The parallelism between the amount of fluoroquinolones prescribed and the number of cases of A. baumannii infection clearly shows a dose-response gradient: the greater the consumption of fluoroquinolones, the stronger the selection pressure.

Our study also showed that the selection pressure caused by fluoroquinolone use was much stronger when the antimicrobial agents were given intravenously than when they were given orally. The comparison of intravenous with oral administration of fluoroquinolones indicates marked differences in pharmacokinetics that may explain why the selection pressure may be stronger when fluoroquinolones are given intravenously. In studies done in healthy volunteers, 11% to 15% of ciprofloxacin given intravenously and 60% (185 to 2220 µg/g) of ciprofloxacin given orally was recovered in feces [36-38]. In contrast, only 110 to 220 µg of pefloxacin per g was found in the feces of healthy volunteers after intravenous administration of this antimicrobial agent [39]. The amount of intraluminal drug is much higher after oral administration than after intravenous administration; therefore, the likelihood of selection of mutants with minimal inhibitory concentrations of 32 µg/g or more colonizing the gastrointestinal tract is increased after intravenous administration and may explain why the parallelism between fluoroquinolone use and the incidence of A. baumannii infections was associated with intravenous use of these antimicrobial agents. These results suggest that fluoroquinolones should be used intravenously with caution or that fluoroquinolone use should be discontinued in units or services where fluoroquinolone-resistant, gram-negative bacilli are prevalent.

In summary, we longitudinally analyzed the epidemiology of infections caused by A. baumannii in a medical and surgical intensive care unit for almost 5 years. The use of molecular typing methods early in the course of the epidemiologic study was crucial to the success of the investigation. The case–control studies done by using the classification of patients obtained with the molecular typing of the isolates showed that infections caused by an epidemic strain of A. baumannii coexisted with other infections that were favored by the selection pressure of intravenous fluoroquinolones.