Risk Factors for Isolation of Strains Susceptible Only to Polymyxin among Patients with Pseudomonas aeruginosa Bacteremia

Abstract

We conducted a case-control study to identify risk factors associated with the isolation of Pseudomonas aeruginosa strains susceptible only to polymyxin from blood by comparing data between 16 patients with blood isolates that were susceptible only to polymyxins and 40 patients with blood isolates that were susceptible to carbapenems. The multivariable analysis showed that exposure to carbapenems was associated with the development of P. aeruginosa bacteremia susceptible only to polymyxin (odds ratio, 9.0; 95% confidence interval, 2.4 to 34.3; P = 0.001).

Pseudomonas aeruginosa is an important nosocomial pathogen that causes a wide spectrum of infections and leads to substantial morbidity and mortality (4, 8). Unfortunately, P. aeruginosa has intrinsic resistance to multiple antimicrobial agents, and the emergence of resistance during therapy is a frequently described phenomenon (8). The continuously evolving antimicrobial resistance of P. aeruginosa isolates has led to the recognition of an alarming epidemic of infections due to P. aeruginosa that is susceptible only to polymyxin (1, 3). However, to our knowledge, no study so far has focused on the analysis of risk factors for the development of infections caused by P. aeruginosa isolates susceptible only to polymyxin.

A retrospective case-control study of all patients with bacteremia due to P. aeruginosa during the period between 1 January 2002 and 31 August 2005 was conducted at the Henry Dunant Hospital, a 450-bed tertiary care center in Athens, Greece. The Institutional Review Board of the hospital approved the study. The computerized database of the hospital's bacteriology laboratory was used to identify these patients. For patients with more than one episode of bacteremia due to P. aeruginosa, only data relevant to the first episode were collected and analyzed. Exposure to a specific antimicrobial agent was considered significant enough to be included in our analysis if (i) exposure had occurred only during the hospitalization in which the bacteremia developed and (ii) the antibiotic had been administered for at least three consecutive days prior to the development of bacteremia.

Identification and susceptibility testing of the P. aeruginosa isolates were performed by using an automated broth microdilution method (Vitek II; bioMérieux, Hazelwood, ΜΟ). The breakpoints used for the in vitro susceptibility of the P. aeruginosa blood isolates were those defined by the Clinical and Laboratory Standards Institute (CLSI). Susceptibility to colistin sulfate was tested by means of the E test (AB Biodisk, Solna, Sweden) (isolates were considered susceptible if the MIC was <2 μg/ml) as well as the disk diffusion method with the use of a l0-μg colistin disk (Oxoid, Basingstoke, Hants, England). Isolates were considered susceptible if the inhibition zone was more than or equal to 11 mm and resistant if the inhibition zone was less than or equal to 8 mm. Intermediate sensitivity to a specific antibiotic was considered as resistance, which reflects clinical practice.

Categorical variables were compared by means of χ² or Fisher's exact test. For continuous variables, we used the Student's t test or the Mann-Whitney test for normally and nonnormally distributed variables, respectively. All variables associated with the development of P. aeruginosa bacteremia caused by strains susceptible only to polymyxin that were significant (P < 0.05) in the univariate analysis were included in a backward, stepwise, multivariable logistic regression model. All statistical analyses were performed using SPSS 11.0 (SPSS Inc., Chicago, IL) and S-PLUS 6.1 Professional (Insightful Corporation, Seattle, WA).

There were 70 patients with bacteremia due to P. aeruginosa during the study period. We focused our analysis on the comparison of exposure to various risk factors between patients with P. aeruginosa blood isolates that were susceptible only to polymyxin and those with P. aeruginosa blood isolates susceptible to at least polymyxins and carbapenems. Thus, we excluded 12 patients with bacteremia due to P. aeruginosa strains susceptible to both polymyxins and piperacillin/tazobactan (resistant to carbapenems) and 2 patients with bacteremia due to P. aeruginosa strains susceptible to polymyxins and aminoglycosides (resistant to carbapenems) from our analysis. Subsequently, there were 56 patients with bacteremia secondary to P. aeruginosa (40 intensive care unit [ICU] patients and 16 ward patients) available for inclusion in our analysis. In 16 of these patients, the isolate was susceptible only to polymyxin (28.6%), and in 40 patients (71.4%), the isolate remained susceptible to carbapenems; susceptibility to ertapenem was not tested.

Among patients with bacteremia due to P. aeruginosa strains that were susceptible only to polymyxin, pneumonia was the predominant infection (37.5% [6/16]), followed by urinary tract infection (18.8% [3/16]), intra-abdominal infection (12.5% [2/16]), and catheter-related infection (6.3% [1/16]). In 4 out of the 16 remaining cases (25%), the origin of the bacteremia was not clear. Among patients with bacteremia due to P aeruginosa isolates that were susceptible to carbapenems, pneumonia was again the predominant infection (45% [18/40]), followed by urinary tract infection (12.5% [5/40]) and intra-abdominal infection (5% [2/40]). Fifteen of the 40 cases (37.5%) of bacteremia were not associated with any identifiable site.

Mortality was 62.5% for patients with bacteremia due to P. aeruginosa blood isolates that were susceptible only to polymyxin and 37.5% for patients with bacteremia due to isolates susceptible to carbapenems (P = 0.09). In Table 1, we summarize the results of the univariable analysis showing that exposure to quinolones, metronidazole, glycopeptides, and carbapenems as well as admission to an ICU, tracheotomy, parenteral feeding, and glucocorticoid treatment were associated with isolation of a strain susceptible only to polymyxin among patients with P. aeruginosa bacteremia. The backward, stepwise, multivariable logistic regression model revealed that the only significant risk factor was prior use of carbapenems (odds ratio, 9.0; 95% confidence interval, 2.4 to 34.3; P = 0.001).

TABLE 1.

Univariate analysis of risk factors associated with bacteremia caused by P. aeruginosa isolates susceptible only to polymyxin

Variablea	Patients with strains susceptible only to polymyxin (n = 16)	Patients with strains susceptible to carbepenems (n = 40)	P value
Demographic
Age (yr) (mean ± SD)	64.8 ± 15.3	66.8 ± 15.0	0.55
No. male (%)	11 (68.7)	20 (50)	0.20
Comorbidity [no. of patients (%)]
Heart dysfunction	10 (62.5)	20 (50)	0.40
Malignancy	5 (31.25)	17 (42.5)	0.44
Lung dysfunction	4 (25)	16 (40)	0.29
Diabetes mellitus	7 (43.7)	9 (22.5)	0.19
Chronic renal failure	5 (31.25)	12 (30)	1.00
Liver dysfunction	3 (18.75)	7 (17.5)	1.00
Hematological disorders	3 (18.75)	11 (27.5)	0.73
Neurological disorders	8 (50)	10 (25)	0.07
Prior hospitalization	10 (62.5)	26 (65)	0.86
Prior surgery	8 (50)	14 (35)	0.29
Abdominal	3 (18.75)	3 (7.5)	0.34
Central nervous system	2 (12.5)	4 (10)	1.00
Other surgery	3 (18.75)	7 (17.5)	1.00
Admitted to ICU	15 (93.75)	25 (62.5)	0.023
Mechanical ventilation	11 (68.75)	20 (50)	0.20
APACHE II score on ICU admission (mean ± SD)	17.0 ± 7.0	15.1 ± 5.0	0.51
Tracheotomy	7 (43.75)	6 (15)	0.035
CSF drainage	2 (12.5)	3 (7.5)	0.62
Abdominal drainage	3 (18.75)	2 (5)	0.13
Prior antibiotic use or treatment [no. of patients (%)]
Antipseudomonal penicillins	6 (37.5)	11 (27.5)	0.53
Expanded-spectrum cephalosporins	1 (6.25)	7 (17.5)	0.42
Broad-spectrum cephalosporins	5 (31.25)	15 (37.5)	0.66
Aminoglycosides	6 (37.5)	10 (25)	0.51
Quinolones	9 (56.25)	11 (27.5)	0.043
Metronidazole	5 (31.25)	3/40 (7.5)	0.035
Clindamycin	4 (25)	10 (25)	1.00
Glycopeptides	12 (75)	16 (40)	0.018
Carbapenems	12 (75)	10 (25)	0.001
Antineoplastic chemotherapy	3 (18.75)	7 (17.5)	1.00
Glucocorticoid treatment	10 (62.5)	10 (25)	0.008
Blood transfusion	12 (75)	24 (60)	0.29
Trauma	2 (12.5)	9 (22.5)	0.48
Invasive procedures or devices [no. of patients (%)]
Central line	14 (87.5)	27 (67.5)	0.19
Arterial line	14 (87.5)	26 (65)	0.11
Foley catheter	13 (81.25)	30 (75)	0.74
Nasogastric tube	11 (68.75)	17 (42.5)	0.08
Foreign material in the body	3 (18.75)	6 (15)	0.70
Colostomy	0 (0)	3 (7.5)	0.55
Gastrostomy	0 (0)	1 (2.5)	1.00
Parenteral feeding	7 (43.75)	4 (10)	0.008

^aCSF, cerebrospinal fluid.

The rapid emergence of imipenem resistance among susceptible P. aeruginosa strains is well known and adds credibility to our findings. Moreover, mutational overexpression of P. aeruginosa genes associated with multidrug efflux systems has been shown to promote acquired multidrug resistance (7). Thus, we can hypothesize that in some settings, carbapenem use per se can induce the spread of multidrug-resistant P. aeruginosa even without prior use of other classes of antibiotics. Our study implies that a similar scenario may be possible in the emerging epidemic of infection by P. aeruginosa that is susceptible only to polymyxin.

Our study has several limitations. It is a small, retrospective, case-control study with all the inherent problems related to this study design (2). Also, the fact that our control group consisted of patients with carbapenem-susceptible strains may have led us to overestimate the odds ratio for carbapenem use as a risk factor for the acquisition of strains susceptible only to polymyxin (5, 6). In addition, we excluded patients with isolates susceptible to polymyxins and antibiotics of other classes but resistant to carbapenems from our analysis because we wanted to focus on a direct comparison between blood isolates susceptible only to polymyxin and those susceptible to carbapenem. Also, we did not screen our patients for the presence of P. aeruginosa by active surveillance culture and did not perform molecular epidemiologic analysis. Thus, it is not possible to distinguish whether a particular case was due to newly acquired resistance of a previously susceptible strain or due to horizontal transmission of a resistant strain. Finally, we did not examine the effect of various pharmacokinetic and pharmacodynamic parameters of carbapenem administration, such as insufficient dosing, dosing interval, and duration of infusion, on the development of antimicrobial resistance.

In conclusion, despite the previously mentioned limitations, our study provides insight into the rising problem of bacteremia caused by strains of P. aeruginosa that are susceptible only to polymyxin by highlighting prior carbapenem use as an independent risk factor for its development.