Carbapenem-Resistant Pseudomonas aeruginosa Bacteremia: Risk Factors for Mortality and Microbiologic Treatment Failure

ABSTRACT

We reviewed 37 patients treated for bacteremia due to carbapenem-resistant (CR) Pseudomonas aeruginosa. Although 65% of isolates were multiple-drug resistant, therapeutic options were available, as all were susceptible to ≥1 antibiotic. A total of 92% of patients received active antimicrobial therapy, but only 57% received early active therapy (within 48 h). Fourteen-day mortality was 19%. Microbiologic failure occurred in 29%. The Pitt bacteremia score (P = 0.046) and delayed active therapy (P = 0.027) were predictive of death and microbiologic failure, respectively.

TEXT

Infections due to multidrug-resistant (MDR) Gram-negative bacteria have emerged as a health care crisis across the globe (1–3). Carbapenems are a last line of defense against many drug-resistant bacterial infections. Unfortunately, infections due to carbapenem-resistant (CR) pathogens are on the rise (4). Carbapenem-resistant Enterobacteriaceae (CRE) have garnered particular attention in the popular press and scientific literature, in part due to several high-profile outbreaks of infection (5, 6). At our center and others, however, CR Pseudomonas aeruginosa is recovered more commonly from patients than CRE (7).

Mechanisms of carbapenem resistance vary considerably between CRE and CR P. aeruginosa. Carbapenem resistance in CRE is mediated by carbapenem-hydrolyzing enzymes, including Ambler class A (e.g., KPC), B (e.g., IMP, VIM), and D (e.g., OXA) beta-lactamases (8). Plasmid-mediated β-lactamases typically are present along with aminoglycoside modifying enzymes (AMEs) and other resistance determinants (9). Accordingly, CRE generally are resistant to most classes of antimicrobials except for salvage agents such as polymyxins and tigecycline (9, 10); unfortunately, the use of polymyxins and tigecycline is limited by high rates of toxicity and suboptimal pharmacokinetics, respectively (11). Ceftazidime-avibactam has activity against CRE, but clinical experience is limited. In addition, recent studies have demonstrated that resistance to ceftazidime-avibactam can develop during or after treatment with this agent (12). Carbapenem resistance in P. aeruginosa stems from a combination of β-lactamases (especially AmpC), porin mutations, MexA-MexB-OprM efflux pump overexpression, and/or penicillin-binding protein alterations (13, 14). Combinations of mechanisms confer reduced susceptibility to carbapenems, but other β-lactam agents and aminoglycosides may retain in vitro activity (15, 16). Our objectives were to define the antibiogram of CR P. aeruginosa bloodstream isolates at our center and to identify factors associated with CR P. aeruginosa bacteremia treatment failure.

We conducted a retrospective cohort study of consecutive patients with CR P. aeruginosa bacteremia at the University of Pittsburgh Medical Center (UPMC) between 1 January 2009 and 30 September 2014. Antimicrobial susceptibility testing was performed by the UPMC clinical microbiology laboratory according to the Clinical and Laboratory Standards Institute (CLSI) methods (15). Carbapenem resistance was defined by current CLSI reference breakpoints for meropenem; isolates classified as intermediate or resistant were considered resistant (15). Multidrug-resistant (MDR) P. aeruginosa was defined by resistance to ≥1 agent in ≥3 antimicrobial classes (17). All patients with bacteremia due to CR P. aeruginosa were included unless (i) they died within 48 h of identification of positive blood cultures or (ii) they did not receive antibiotic treatment for CR P. aeruginosa bacteremia. Prior carbapenem therapy was defined as any treatment with a carbapenem for ≥3 days within 3 years prior to CR P. aeruginosa bacteremia. Early and delayed antimicrobial therapies were defined as receipt of ≥1 agent active against the respective CR P. aeruginosa isolate within and after 48 h of bacteremia onset, respectively. Combination therapy was defined as the use of ≥2 active agents. Recurrent CR P. aeruginosa infection was defined as isolation of CR P. aeruginosa and associated new signs and symptoms of infection within 90 days of initial therapy for CR P. aeruginosa bacteremia; for an infection to be included as a recurrence, the patients must have initially experienced clinical success. Microbiologic failure was defined as persistent bacteremia due to CR P. aeruginosa despite ≥5 days of active antimicrobial therapy or as recurrent CR P. aeruginosa infections. Patients who died within 5 days of CR P. aeruginosa bacteremia onset were included in risk factor analysis for mortality but not for microbiologic failure. The primary outcome was 14-day mortality. Secondary outcomes were 30-day mortality and microbiologic response.

Dichotomous and continuous variables were compared using Fisher's exact chi-square and Mann-Whitney U tests, respectively. Variables associated with death at 14 days or microbiologic failure at P values of <0.20 in univariate analysis were included in the multivariate logistic regression model using backward selection procedures. P values of ≤0.05 (two-tailed) were considered statistically significant.

Thirty-seven patients with CR P. aeruginosa bacteremia were included; the median age was 61 years (range, 18 to 90). Forty-nine percent (18/37) of patients were men, and 41% (15/37) were solid-organ transplant recipients. Sixty-eight percent (25/37) of patients resided in the intensive care unit (ICU) at the onset of CR P. aeruginosa bacteremia. The median Charlson comorbidity index was 5 (range, 1 to 9); the median acute physiology and chronic health evaluation II (APACHE II) score and Pitt bacteremia score were 19 (range, 4 to 32) and 3 (range, 0 to 12), respectively. Fifty-seven percent (21/37) of patients received a carbapenem prior to CR P. aeruginosa bacteremia (median of 16 days [range, 3 to 114 days] before onset). Twenty-seven percent (10/37) of patients had primary bacteremia. Secondary bacteremia resulted from sites in the abdomen (27%, 10/37), respiratory tract (27%, 10/37), skin/soft tissue (8%, 3/37), urinary tract (5%, 2/37), and left-ventricular assist device driveline (3%, 1/37); the portal of entry could not be determined in one patient.

No patient had polymicrobial bacteremia. Rates of CR P. aeruginosa susceptibility to other antimicrobial agents ranged from 22% to 92% (Fig. 1). A total of 100% of isolates were susceptible to at least one of the agents shown; 65% (24/37) of isolates were MDR.

FIG 1.

Percentage of Carbapenem-resistant P. aeruginosa bloodstream isolates susceptible to tested antibiotics. *, susceptibility rates were defined by CLSI interpretive criteria for P. aeruginosa. All 37 isolates were tested against each antibiotic listed.

Ninety-two percent (34/37) and 8% (3/37) of patients received definitive antimicrobial monotherapy and combination therapy, respectively. Fifty-four percent (20/37) and 92% (34/37) of patients were treated with active empirical and active definitive therapy, respectively. Early active therapy was administered in 57% (21/37) of patients; the median time to active therapy was 23.5 h (range, 0 to 164).

Fourteen- and 30-day mortality rates were 19% (7/37) and 30% (11/37), respectively. By univariate analysis, death at 14 days was more common among patients with secondary bacteremia from the respiratory tract than among patients with secondary bacteremia from other sources (P = 0.009). Death was not impacted by the time to initiation of active therapy or treatment with an active agent (Table 1). Mortality rates (14 day) were similar among patients infected with MDR P. aeruginosa (21%, 5/24) or non-MDR P. aeruginosa (15%, 2/13) (P = 1.00). By multivariate analysis, the Pitt bacteremia score was the sole predictor of death at 14 days (odds ratio [OR] = 2.39, 95% confidence interval [CI] = 1.01 to 5.63; P = 0.046). Microbiologic failures occurred in 29% (10/35) of patients who survived ≥5 days, and all were due to recurrent CR P. aeruginosa infections (one patient with infective endocarditis also had persistent CR P. aeruginosa bacteremia). Microbiologic failures were independently associated with a delay in active therapy for >48 h (OR = 10.51, 95% CI = 1.31 to 84.10; P = 0.027).

TABLE 1.

Risk factors for death at 14 days and microbiologic failure among all carbapenem-resistant P. aeruginosa isolates

Factor	Mortality at 14 days (n = 37)		Univariate P value	Multivariate P value	OR (95% CI)a	Microbiologic outcome (n = 35)		Univariate P value	Multivariate P value	OR (95% CI)
	Alive (n = 30)	Dead (n = 7)				Success (n = 25)	Failure (n = 10)
Median age in yrs (range)	60.5 (18–90)	65 (50–72)	0.34			61 (18–90)	56.5 (20–67)	0.27
Prior hospitalization within previous 90 days, n (%)	26 (87)	7 (100)	0.57			22 (88)	9 (90)	1.00
Surgery within previous 31 days, n (%)	13 (43)	2 (29)	0.68			11 (44)	4 (40)	1.00
Neutropenia, n (%)	2 (7)	1 (14)	0.48			3 (12)	0 (0)	0.54
Immunocompromised, n (%)b	12 (40)	4 (57)	0.44			13 (52)	2 (20)	0.13	0.053
Median Charlson comorbidity index score (range)c	5 (1–9)	7 (4–9)	0.07	0.199		5 (1–9)	4.5 (2–7)	0.62
Concomitant CR P. aeruginosa infection, n (%)	20 (67)	6 (88)	0.649			18 (72)	6 (60)	0.69
Lung portal of entry, n (%)d	5 (17)	5 (71)	0.009	0.170		5 (20)	3 (30)	0.66
i.v.e catheter portal of entry, n (%)	8 (27)	1 (14)	0.66			6 (24)	3 (30)	0.69
Median Pitt bacteremia score (range)	2 (0–9)	7 (6–12)	0.0007	0.046	2.39 (1.01–5.63)	3 (0–12)	2.5 (0–9)	0.82
Late active therapy (>48 h), n (%)	14 (47)	2 (29)	0.65			8 (32)	7 (70)	0.06	0.027	10.51 (1.31–84.10)
Susceptible to empirical regimen, n (%)	16 (53)	4 (57)	1.00			16 (64)	3 (30)	0.132	0.996
Median time to active therapy (range)	23.5 (0–164)	9 (1–72)	0.34			9 (0–128)	70 (8–164)	0.02	0.423
Bacteremia due to MDR P. aeruginosa, n (%)	19 (63)	5 (71)	0.43			14 (56)	8 (80)	0.26

^aOR, odds ratio; CI, confidence interval.

^bAn immunocompromised state was defined as the presence of neutropenia or receipt of steroids or other immunosuppressive agents in the 31 days prior to the infection.

^cVariables were defined according to the standard Charlson comorbidity index.

^dOf the patients with lung portal of entry, 3 had a history of double lung transplantation, 2 had chronic obstructive pulmonary disease, 1 had chronic bronchitis, 1 had 5-aminosalicylic acid (5-ASA)-induced lung fibrosis, and 2 did not have any history of prior lung disease.

^ei.v., intravenous.

A striking finding of this study was that CR P. aeruginosa bloodstream isolates at our center, despite their MDR phenotypes, retained sensitivity to aminoglycosides, antipseudomonal beta-lactams, and fluoroquinolones (Fig. 1). As a result, 92% of patients were treated with at least one agent that was active against CR P. aeruginosa, without a need to resort to salvage agents such as colistin. As in our experience, amikacin susceptibility rates of >90% have been previously described among CR P. aeruginosa bloodstream isolates (18, 19). Notably, rates of susceptibility of our CR P. aeruginosa isolates to tobramycin (91%), piperacillin-tazobactam (57%), and cefepime (68%) were higher than reported elsewhere (62%, 43%, and 24%, respectively) (19). A total of 65% of our CR P. aeruginosa bloodstream isolates were classified as MDR, which is consistent with prior studies (60% to 62%) (19, 20). Discrepancies between studies may reflect differences in antimicrobial prescribing practices and in antibiotic resistance mechanisms among isolates from different centers. Importantly, the MDR phenotype observed here is clearly different from that seen with the “carbapenem resistant only P. aeruginosa” (CROPA) isolates that are uniformly susceptible to ciprofloxacin, antipseudomonal penicillins, and cephalosporins (21). Accordingly, only 57% of our patients received active empirical therapy. The rate of active empirical therapy in our study was similar to the 60% reported by Joo and colleagues (20) but was higher than those reported by Suárez et al. and Jeong et al. (33% to 38%) (22, 23). Nevertheless, 92% of our patients ultimately received active definitive therapy.

Retained antimicrobial susceptibility and the ability to devise active antimicrobial treatment regimens distinguished our experience with CR P. aeruginosa bacteremia from our previously reported studies of CRE bacteremia (9, 24–26). Rates of resistance to gentamicin and cefepime among CRE bloodstream infections at our institution were 53% and 100%, respectively, which were higher than the 27% and 32% rates reported here for CR P. aeruginosa (27). Therefore, a significant minority of our CRE infections were effectively untreatable, even with salvage agents. It will be important to study the impact of new drugs such as ceftazidime-avibactam and ceftolozane-tazobactam, which have activity against carbapenem-resistant pathogens, on treatment and outcomes of CR P. aeruginosa and CRE infections. Ceftolozane-tazobactam evades common mechanisms of β-lactam resistance in CR P. aeruginosa, including efflux systems, OprD loss, and at least some β-lactamases that are inhibited by tazobactam (28). Ceftazidime-avibactam, on the other hand, demonstrates broad activity against β-lactamase-producing Gram-negative pathogens, including those that produce extended-spectrum β-lactamases (ESBLs), AmpC cephalosporinases, and serine-based carbapenemases such as KPC (29). In our previous study, we found that the majority of CR P. aeruginosa isolates at our center were susceptible to both agents, which provide much-needed additions to a depleted antimicrobial armamentarium (30).

Despite the use of treatment regimens that were active against CR P. aeruginosa, mortality rates in this study were still 19% and 30% at 14 and 30 days, respectively. Thirty-day mortality rates in this study (30%) were comparable to those reported from prior studies of carbapenem-susceptible and CR P. aeruginosa bacteremia (18). Indeed, it is unclear if outcomes are worse among patients with CR P. aeruginosa bacteremia than among those with carbapenem-susceptible P. aeruginosa bacteremia, as data from multiple studies are conflicting (23, 31). In one recent prospective, multicenter study, 30-day mortality was greater for CR P. aeruginosa bacteremia than for carbapenem-susceptible P. aeruginosa bacteremia, but the difference diminished as the number of comorbidities increased (18). Along these lines, severity of illness was an independent predictor of death in our series. An association between delayed active antimicrobial therapy and microbiologic failures, such as we describe, is consistent with previous reports for a wide range of infectious diseases (32–34). This finding highlights the need for new diagnostic tests that rapidly detect MDR and carbapenem-resistant pathogens.

Our study was limited to a small number of cases and experience from a single center. Thus, findings may not be applicable to centers with different patient populations or antibiograms. Our findings highlight important differences in treatment options against CRE infections, for which salvage regimens are commonly employed, and CR P. aeruginosa infections, for which standard agents remain useful. Understanding local antibiograms against CR P. aeruginosa will be particularly important as new broad-spectrum antibiotics reach the clinic in order to preserve their utility and to devise the most rational treatment paradigms. Regardless of the choice of specific agents, active treatment against CR P. aeruginosa should be instituted as quickly as possible and, at worst, within 48 h of bacteremia onset. Research is needed into optimized antibiotic dosing regimens based on pharmacokinetic-pharmacodynamic data and the role of combination therapy in improving outcomes and preventing further resistance.