Risk Factors for Acquisition of Multidrug-Resistant Pseudomonas aeruginosa Producing SPM Metallo-β-Lactamase

MATERIALS AND METHODS

The study was performed at the Hospital Universitário Clementino Fraga Filho, a 480-bed public tertiary-care teaching hospital in Rio de Janeiro, Brazil, with ∼120,000 patient-days per year. The study was approved by the Ethical Committee of the Hospital. A surveillance study was conducted with the microbiology laboratory database of the hospital to identify all positive cultures for P. aeruginosa between March 1999 and April 2000. P. aeruginosa was recovered from various biological materials, collected according to the discretion of the attending physicians. No routine surveillance culture was performed during this period. MDRPa was defined as an isolate resistant to at least eight of the following antibiotics: amikacin, aztreonam, ceftazidime, cefepime, ciprofloxacin, gentamicin, imipenem, meropenem, piperacillin-tazobactam, and ticarcillin-clavulanate. MDRPa isolates were further characterized as carrying the bla_SPM gene or not, as described under microbiological procedures.

An infectious disease physician evaluated all the patients and classified the P. aeruginosa isolate according to established diagnostic criteria for nosocomial infections (Centers for Disease Control and Prevention) (15). Colonization was defined when these criteria were not fulfilled. When the isolate was determined as causing colonization but the attending physician prescribed an antimicrobial agent, the patient was classified as having a possible infection. The overall use of the different classes of antibiotics did not change during the study period and was equally distributed through all hospital wards.

To identify risk factors associated with the acquisition of MDRPa and particularly MDRPa expressing the bla_SPM gene, we performed a case control study using two different case definitions. In the first analysis, a case was defined as a patient who had at least one isolate of MDRPa carrying the bla_SPM gene (SPM case patient). In the second analysis, a case was defined as a patient who had at least one isolate of MDRPa that did not carry the bla_SPM gene (non-SPM case patient). For each case patient, we selected two controls, defined as patients colonized and/or infected by a non-MDRPa isolate during the same study period and with the closest duration of hospitalization until the isolation of P. aeruginosa, as cases. Only one P. aeruginosa isolate per patient was considered for this analysis. If more than one P. aeruginosa isolate was recovered from the same patient, we selected only one isolate, in the following preferential order: first, an SPM-MDRPa isolate; second, a non-SPM-MDRPa isolate; and third, a non-MDRPa isolate. If a patient had only non-MDRPa isolates, the first P. aeruginosa isolate was considered.

RESULTS

During the study period, we identified 32 MDRPa isolates from 32 patients. In 18 patients (56%), the isolate was classified as causing infection, whereas in 10 patients (31%), the isolates were classified as colonization, and in 4 patients (12%), the isolates were classified as possible infections. A surgical wound was the most frequent source (11 isolates), followed by blood (5 isolates) and urine (3 isolates). All MDRPa isolates were resistant to carbapenems. Fourteen (44%) of the 32 patients with MDRPa were colonized or infected by isolates harboring the bla_SPM gene, all presenting carbapenemase activity (SPM case patients), and 18 patients were colonized or infected by isolates that did not express the bla_SPM gene and did not reveal carbapenemase activity (non-SPM case patients). By PFGE, all SPM isolates had a unique pattern that we called genotype A. The isolates from the 18 non-SPM patients belonged to nine different genotypes, including three isolates with genotype A.

SPM case patients were found across seven hospital wards, but 7 of the 14 patients were hospitalized in the same ward during a 2-month period (January and February 2000). Twelve patients had been hospitalized previously: 7 at the Hospital Universitário Clementino Fraga Filho and 5 in other hospitals. Ten SPM case patients were immunosuppressed: 5 were solid organ transplant recipients (4 renal and 1 liver transplant recipients), 4 had leukemia and had received chemotherapy (1 underwent hematopoietic stem cell transplantation), and 1 had AIDS. In 8 of the 14 SPM case patients, the P. aeruginosa isolate was considered infection; in isolates from 3 patients, the isolate represented colonization; and in isolates from 3 patients, the isolates were considered possible infections.

Table 1 shows the univariate analysis of cases and controls. In the first analysis, the 14 SPM case patients were compared to 28 control (non-MDRPa) patients. Compared to the controls, SPM case patients were more likely to have chronic renal failure, to have received solid organ transplant or dialysis, to be immunosuppressed, to have drains, and to have been hospitalized in the preceding year. In addition, SPM case patients were more frequently exposed to the following antibiotics: narrow-spectrum cephalosporins, cefepime, quinolones, and vancomycin. The median duration of antibiotic use until the isolation of P. aeruginosa was significantly longer in SPM case patients. By multivariate analysis, previous exposure to quinolones (OR = 14.70, 95% CI = 1.70 to 127.34, P = 0.01) was the single variable associated with colonization or infection by MDRPa isolates harboring the bla_SPM gene.

In the second analysis, 18 patients with MDRPa isolates that did not harbor the bla_SPM gene (non-SPM case patients) were compared to 36 control patients with non-MDRPa isolates. There were no differences between cases and controls with regard to demographics, underlying conditions, or coexisting exposures. Regarding the exposure to antibiotics, case patients were more likely to have received cefepime, carbapenems, and vancomycin. Multivariate analysis showed that the use of cefepime was associated with non-SPM-MDRPa isolates (OR = 8.50, 95% CI = 1.51 to 47.96, P = 0.01).

DISCUSSION

This is the first attempt to characterize the clinical and epidemiological context of colonization and invasive disease by P. aeruginosa isolates expressing the bla_SPM gene, which is widely disseminated among Brazilian hospitals. We observed that bla_SPM isolates accounted for a substantial number of MDRPa cases (44%), and most of these isolates were clinically relevant, since 57% of the patients were considered to have invasive disease.

Half of the SPM case patients were clustered in a 2-month period in a single ward. In addition, most of the patients had been previously admitted to our hospital (seven patients) or to other institutions (five patients). In a previous paper, we identified patients with colonization or invasive disease by bla_SPM from three other hospitals in Rio de Janeiro during the same period as our cases (34). Taken together, these data suggest that cross-transmission between patients and the transfer of patients between institutions may have played a major role in the dissemination of bla_SPM. Indeed, after the identification of the problem, infection control measures were implemented (basically, educational aspects); in a subsequent analysis of 107 isolates between October 2002 and August 2003, no SPM-MDRPa isolate was identified (Pellegrino et al., submitted).

In the present study, we used as controls patients colonized and/or infected with susceptible isolates of P. aeruginosa (non-MDRPa isolates) because we sought to investigate the probability that a patient with an isolate of P. aeruginosa had an MDRPa isolate. This information would be a useful guide for selection of empirical antibiotic treatment and early institution of contact precautions. By univariate analysis, we observed that bla_SPM colonization and/or infection was associated with immunosuppression, receipt of hemodialysis, and hospitalization in the preceding year. On the other hand, for non-SPM-MDRPa isolates, there were no differences among cases and controls regarding underlying conditions or invasive procedures. By multivariate analysis, risk factors for SPM and non-SPM-MDRPa differed but in both situations involved the use of antibiotics: quinolones for the acquisition of SPM-MDRPa (bla_SPM) and cefepime for non-SPM-MDRPa. It is possible that quinolones could select for other resistance mechanisms present in Mbla-producing isolates and thus indirectly select for SPM-MDRPa. On the other hand, to our knowledge, the use of cefepime has not been identified as a risk factor for resistant P. aeruginosa. Exposure to various antibiotics has been associated with MDRPa, including carbapenems (4, 5, 18, 20, 38), piperacillin-tazobactam (18, 20), vancomycin (18), aminoglycosides (18, 20), cephalosporins (18), ceftazidime (36), and quinolones (3, 8, 31, 39). The mechanisms of resistance of MDRPa in these studies were not defined.

The mechanism by which the use of a specific antibiotic predisposes to the development of infection by resistant organisms is not well known. One explanation for the association between antibiotic exposure and MDRPa is that the antibiotic itself has a potential to induce or select for resistance. It is well known that previous exposure to a particular antibiotic is associated with the acquisition of resistance to itself (10) or to antibiotics belonging to a different class (7, 25). This is particularly true for P. aeruginosa, which is a pathogen harboring multiple mechanisms of resistance (24). On the other hand, it would be expected that in the case of colonization or infection by bla_SPM, antibiotic pressure could play a lesser role, since horizontal transmission would be the main mechanism of acquisition (32). However, it is possible that exposure to a particular antibiotic increases the burden of colonization by a clonal strain, increasing the possibility of horizontal transmission. This phenomenon has been reported with vancomycin-resistant enterococci and the use of antianaerobic antibiotics (9).

Our study has some limitations. First, since we did not perform surveillance cultures, we cannot rule out the possibility that some of the control subjects might have been colonized by MDRPa and should have been classified as case patients instead of controls. Second, since our controls were those with susceptible strains of P. aeruginosa, they were much less likely to have been exposed to antibiotics that are active against this organism. Therefore, it is possible that the OR is overmagnified in this analysis, as discussed by Harris et al. (17, 19). However, since our main interest was to evaluate the probability that a patient with P. aeruginosa had an MDRPa isolate, the most appropriate control group seemed to be non-MDRPa patients. Another limitation of our study is that since most cases were clustered in place, it is possible that other variables associated with breakdowns in infection control measures played a role in the acquisition of bla_SPM. Also, because of the small number of patients with bla_SPM, we cannot rule out the possibility of a beta error in some analyses. On the other hand, our study seems to be the first to analyze by multivariate analysis the clinical scenario associated with a highly prevalent metallo-β-lactamase-producing P. aeruginosa strain.