Prospective Multicenter Study of the Impact of Carbapenem Resistance on Mortality in Pseudomonas aeruginosa Bloodstream Infections

Prospective Multicenter Study of the Impact of Carbapenem Resistance on Mortality in Pseudomonas aeruginosa Bloodstream Infections Carmen Peña, a Cristina Suarez, a Mónica Gozalo, b Javier Murillas, c Benito Almirante, d Virginia Pomar, e Manuela Aguilar, f Ana Granados, g Esther Calbo, h Jesús Rodríguez-Baño, i Fernando Rodríguez, j Fe Tubau, a Luis Martínez-Martínez, b and Antonio Oliver c for the Spanish Network for Research in Infectious Diseases (REIPI) Servicio de Enfermedades Infecciosas, Hospital Universitario de Bellvitge-IDIBELL, Barcelona, a Servicio de Microbiología, Hospital Universitario Marqués de Valdecilla- IFIMAV, Santander, b Servicio de Medicina Interna-Infecciosas y Microbiología, Hospital Universitario de Son Espases, Palma de Mallorca, c Servicio de Enfermedades Infecciosas, Hospital Universitario Vall d’Hebrón, Barcelona, d Unidad de Enfermedades Infecciosas, Hospital Santa Creu i Sant Pau, Barcelona, e Servicio de Enfermedades Infecciosas, Hospital Universitario Virgen del Rocío, Seville, f Sección de Enfermedades Infecciosas, Consorci Hospitalari Parc Taulí, Sabadell, g Sección de Enfermedades Infecciosas, Hospital Mutua de Terrasa, Terrasa, h Sección de Enfermedades Infecciosas, Hospital Universitario Virgen Macarena, Seville, i and Servicio de Microbiología Infecciosas, Hospital Universitario Reina Sofía-IMIBIC, Córdoba, j Spain The impact of antimicrobial resistance on clinical outcomes is the subject of ongoing investigations, although uncertainty remains about its contribution to mortality. We investigated the impact of carbapenem resistance on mortality in Pseudomonas aeruginosa bacteremia in a prospective multicenter (10 teaching hospitals) observational study of patients with monomicrobial bacteremia followed up for 30 days after the onset of bacteremia. The adjusted influence of carbapenem resistance on mortality was studied by using Cox regression analysis. Of 632 episodes, 487 (77%) were caused by carbapenem-susceptible P. aeruginosa (CSPA) isolates, and 145 (23%) were caused by carbapenem-resistant P. aeruginosa (CRPA) isolates. The median incidence density of nosocomial CRPA bacteremia was 2.3 episodes per 100,000 patient-days (95% confidence interval [CI], 1.9 to 2.8). The regression demonstrated a time-dependent effect of carbapenem resistance on mortality as well as a significant interaction with the Charlson index: the deleterious effect of carbapenem resistance on mortality decreased with higher Charlson index scores. The impact of resistance on mortality was statistically significant only from the fifth day after the onset of the bacteremia, reaching its peak values at day 30 (adjusted hazard ratio for a Charlson score of 0 at day 30, 9.9 [95% CI, 3.3 to 29.4]; adjusted hazard ratio for a Charlson score of 5 at day 30, 2.6 [95% CI, 0.8 to 8]). This study clarifies the relationship between carbapenem resistance and mortality in patients with P. aeruginosa bacteremia. Although resistance was associated with a higher risk of mortality, the study suggested that this deleterious effect may not be as great during the first days of the bacteremia or in the presence of comorbidities. Pseudomonas aeruginosa is a leading cause of nosocomial infections, which are often severe (26, 43) and difficult to treat because of their limited susceptibility to antimicrobial agents (35) and the frequent emergence of antibiotic-resistant mutants during therapy (8). Factors related to the host, the organism, and the treatment may increase mortality. With regard to the host, the severity of the underlying disease may be synergistic with infection due to resistant organisms; concomitantly, increased virulence could explain the adverse impact of resistant pathogens on clinical outcomes, although this association has not been demonstrated to date. In addition, factors such as decreased antibiotic effectiveness or a delay in the initiation of therapy may contribute to adverse outcomes in patients infected by resistant pathogens (17). The problem of antibiotic-resistant organisms is increasing (18), and the impact of antimicrobial resistance on clinical outcomes is the subject of ongoing investigations (2, 5, 9, 15, 25–28, 32, 33, 40, 41, 43), although its contribution to mortality remains uncertain. Measurements of its impact on patients are, by necessity, derived essentially from observational studies, and therefore, an adequate control of confounding variables is essential. The choice of possible confounders is usually based only on the statistical significance of the association between them and the exposure; thus, the classical criteria of confounding, also based on the relationship with the outcome and the exposure-outcome pathway, are usually ignored (24). Carbapenems are commonly used as first-line agents for nosocomial infections, since their spectrum includes P. aeruginosa, which is high on the list of Gram-negative pathogens frequently isolated (37). Nevertheless, carbapenem-resistant and multidrugresistant P. aeruginosa strains are being increasingly recognized (18, 39), and several studies have suggested higher mortality rates among patients infected by nonsusceptible P. aeruginosa strains (2, 25, 26, 28, 40, 41), although the real effect on mortality in P. aeruginosa bacteremia is still unknown. Furthermore, a study performed at one of the participating hospitals (40) showed that patients with carbapenem-resistant P. aeruginosa (CRPA) bacteremia had similar attributable mortality but slow initial mortality compared to patients infected with susceptible strains. This Received 24 October 2011 Returned for modification 20 November 2011 Accepted 6 December 2011 Published ahead of print 12 December 2011 Address correspondence to Carmen Peña, [email protected]. Copyright © 2012, American Society for Microbiology. All Rights Reserved. doi:10.1128/AAC.05991-11 0066-4804/12/$12.00 Antimicrobial Agents and Chemotherapy p. 1265–1272 aac.asm.org 1265 knowledge can guide antibiotic policy away from aggressive empirical therapy and the associated adverse events, increased costs, and increased selective pressure for more antibiotic resistance. We designed a prospective, multicenter, large-scale, observational study of patients with P. aeruginosa bacteremia, with the main objective of investigating the impact of carbapenem resistance on mortality. MATERIALS AND METHODS Setting and design. Ten public hospitals located in four areas of Spain (Andalusia, the Balearics, Cantabria, and Catalonia) participated in the study between January 2008 and December 2009. All adult patients (⬎18 years old) with bacteremia due to P. aeruginosa were recruited prospectively through daily reviews of blood culture results at the participating centers. Episodes with polymicrobial bacteremia were excluded. Patients were monitored for 30 days after the first positive blood culture, and the registered event was death. Bacteremia occurring more than 14 days after the first positive blood culture was considered a new episode. The study was approved by the local ethics committees of the participating centers. Variables and definitions. The following data were recorded: age and sex; comorbidities and severity of underlying diseases, calculated by using the Charlson comorbidity index (10); severity of illness, estimated by the simplified acute physiology score (SAPS II) for intensive care unit (ICU) patients (34); presence of neutropenia (absolute granulocyte count of ⬍500 granulocytes/ml) and use of immunosuppressive therapy (chemotherapy, radiotherapy, and/or immunosuppressive drugs during bacteremia presentation); source of bacteremia (21); severity of acute illness at presentation according to the Pitt bacteremia score (11); presence of septic shock or severe sepsis and multiorgan dysfunction syndrome (MODS) at presentation and at 48 h (3); and antimicrobial treatment received. CRPA was defined as isolates with imipenem and/or meropenem MICs of ⱖ8 mg/liter (Clinical and Laboratory Standards Institute [CLSI] intermediate and resistant categories). Bacteremia was defined as being nosocomially acquired when the infection occurred more than 48 h after admission to the hospital. Episodes of community-acquired bacteremia were further classified as being health care associated according to criteria described previously by Friedman et al. (20). The source of bacteremia was divided into two categories: low risk, including bacteremias of urinary tract and vascular catheter origins, and high risk, including all other sources (27). Empirical antimicrobial therapy was considered adequate when an antimicrobial regimen that included an active antimicrobial was administered within 24 h of the extraction of the blood sample and before susceptibility was known; aminoglycoside monotherapy was considered an inadequate treatment for P. aeruginosa pneumonia. Definitive antimicrobial therapy was considered therapy administered according to susceptibility results. A delay in adequate antimicrobial therapy was defined as the time from bacteremia until the administration of the empirical or definitive adequate therapy. Additional therapy was defined as procedures required to control the source of bacteremia (removal of vascular or urinary catheters as well as surgical or percutaneous drainage of collections). Microbiological studies. The blood isolates were studied at the participating centers and sent to a reference laboratory (Servicio de Microbiología, Hospital Universitario de Bellvitge, Barcelona, Spain). P. aeruginosa strains were identified and tested for their antimicrobial susceptibilities by individual laboratories using standard techniques. In the reference laboratory, the antibiotic susceptibilities of 606 available isolates were confirmed by the disk diffusion method (12). CLSI criteria (13) were used to define susceptibility or resistance to the following antimicrobials tested: piperacillin, ticarcillin, piperacillin-tazobactam, ceftazidime, cefepime, aztreonam, imipenem, meropenem, ciprofloxacin, gentamicin, tobramycin, amikacin, and colistin. Multiresistance was defined according to criteria described elsewhere previously (36). The production of transferable class B carbapenemases (metallo- ␤ - lactamases [MBLs]) was investigated by phenotypic (Etest MBL) and genotypic (PCR and sequencing) testing (23). Additionally, the prevalence of classical chromosomal (mutational) resistance mechanisms was recently analyzed with a sample of 190 isolates from this cohort (7). In order to determine clonal relatedness, chromosomal DNAs from these 190 isolates were processed for analysis by pulsed-field gel electrophoresis (PFGE) (42). Statistical analysis. The initially required sample size (560 samples) was calculated based on the results of a retrospective study (40), in order to detect a risk ratio (RR) between carbapenem resistance and mortality of at least 1.4, with ␣ and ␤ probabilities of 5% and 20%, respectively; a carbapenem-susceptible P. aeruginosa bacteremia (nonexposed group) mortality rate of 35%; and an expected CRPA/carbapenem-susceptible P. aeruginosa (CSPA) ratio of 1 to 3. The analysis performed after having exceeded the number of patients required showed a smaller difference in the effect than expected; therefore, the recruitment of patients was stopped due to logistical difficulties in achieving the recalculated sample size (1,644 patients). The nosocomial incidence rate of P. aeruginosa bacteremia was adjusted by the number of stays and is expressed as the rate per 100,000 patient-days. Incidence estimations and confidence intervals (CIs) were calculated with the exact (Poisson) method. As no population census data were available, population incidence rates were not calculated. Survival curves were constructed by means of the Kaplan-Meier method, and the log rank test was used for comparisons between groups. Cox hazard regression was used to estimate the most unbiased measure of the effect of resistance (exposure) on the time to mortality (outcome) (24). The outcome evaluated was death, the patient admission was the unit of analysis, and the date of collection of the initial P. aeruginosa isolate was considered to be time zero. Cases were censored on day 30; those lost to follow-up were censored on the last follow-up day that they were known to be alive. As the final model did not satisfy the proportional hazard assumptions, the variables resistance and Pitt index were modeled as time-dependent variables. Thus, the final model was a nonproportional hazard model, in which time-dependent variables had different effects depending on the time point considered (14). To assess effect modifications, possible interaction terms between the exposure variable (resistance) and covariates were examined and maintained in the model depending on the results of the significance test (P⬍ 0.05). To control for confounding, variables were first screened to meet three classical criteria (38). In the crude analysis, variables that were associated (P⬍0.20) with exposure (resistance), those that were associated (P⬍0.20) with the outcome (mortality), and those that were not considered to be an intermediate variable between resistance and mortality were candidates for multivariate analysis. Variables and subsets of variables which led to substantial confounding if removed (10% change or more in the coefficient estimate) were maintained in the model, together with exposure (resistance). Results of the final model were expressed as the adjusted hazard ratio (aHR) for carbapenem resistance given different values of the interaction variables at different time points. All statistical tests were two tailed; a Pvalue of ⱕ0.05 was considered significant. All statistical analyses were performed with the SPSS package, version 15.0. RESULTS A total of 638 episodes of P. aeruginosa bacteremia from 10 participating centers were entered into the study. Six episodes were excluded: three episodes with patients younger than 18 years of age, one episode that did not have complete information, one episode with a blood specimen that was obtained only through a vascular catheter without signs or symptoms of infection, and one remaining episode that was finally identified as polymicrobial bacteremia in the reference laboratory. Therefore, 632 episodes were available for evaluation, 487 (77%) CSPA and 145 (23%) CRPA episodes. In addition, nine patients were lost to follow-up: seven patients with CSPA and two with CRPA bacteremias. Epidemiological and clinical characteristics of the cohort are shown in Table Peña et al. 1266 aac.asm.org Antimicrobial Agents and Chemotherapy 1. A recurrence of P. aeruginosa bacteremia within 14 days after the initial episode occurred in 3 episodes of CSPA infection; the 3 pairs of initial and subsequent P. aeruginosa strains isolated from the same patients were found to be clonally identical and showed identical susceptibility patterns. The numbers of P. aeruginosa episodes in each hospital are shown in Fig. 1. Among the nosocomial episodes, the median incidence density of P. aeruginosa bacteremia was 7.9 episodes per 100,000 patient-days (95% CI, 7.1 to 8.8), and that of CRPA bacteremia was 2.3 episodes per 100,000 patient-days (95% CI, 1.9 to 2.8). Microbiology results. A total of 23% of the isolates were nonsusceptible to imipenem, 21.5% were nonsusceptible to meropenem, and 21% were resistant to both carbapenems. Just under one-fifth (19.6%) of the isolates met the established criteria for multiresistance (36); only amikacin (96.9%) and colistin (100%) showed conserved activity in most of the isolates. Six of the isolates (1%), from four different hospitals, were found to produce MBLs (5 VIM-2 isolates and 1 VIM-20 isolate). In contrast, an analysis of a subset of 190 isolates from the collection revealed a high prevalence of chromosomal resistance mechanisms: up to 39% of the isolates overexpressed the ␤ -lactamase AmpC or one of the major multidrug efflux pumps, and all tested imipenem-resistant isolates were deficient in the carbapenem porin OprD (7). PFGE analysis showed a notable clonal diversity, with up to 152 different genotypes detected among the isolates investigated. Nevertheless, a significant clonal dissemination of multiresistant isolates was also shown: among 35 multiresistant P. aeruginosa isolates studied, one genotype prevailed (20/35 isolates [57%]), which was found mainly in hospitals in Catalonia (18/20 [90%]). Patient characteristics according to carbapenem susceptibility. The demographic and clinical characteristics of patients with P. aeruginosa bacteremia according to carbapenem susceptibility are shown in Table 2. Among the 145 patients with CRPA bacteremias, acquisition was mainly nosocomial; they were more frequent in ICU patients (40% versus 23% [P⬍0.001]), and the median number of days between hospital admission and the bacteremia episode was significantly higher for these patients than for patients with CSPA infections. Patients in the CSPA group were more likely than those in the CRPA group to have bacteremia from a high-risk source (272 [56%] CSPA patients versus 69 [48%] CRPA patients [P⫽0.08]). Mortality. The 30-day mortality cumulative incidence rates TABLE 1 Epidemiological and clinical characteristics of the 632 episodes of P. aeruginosa bacteremia Variable a n% Median age (yr) (IQR) 68 (55.5–77.5) No. of male patients 433 69 Carbapenem resistance 145 23 Type of acquisition Nosocomial 362 57 Community 33 5 Health care 237 38 ICU admission 145 23 Mean SAPS II (SD) 42.60 (17.9) Median Charlson index (IQR) 2 (1–4) Underlying condition 534 85 Solid malignancy 178 28 Diabetes 164 26 Immunosuppression 157 25 Heart disease 115 18 Chronic renal failure 100 16 Hematologic malignancy 93 15 Chronic lung disease 94 15 Neutropenia 67 11 Chronic neurologic disease 43 7 Cirrhosis 34 5 AIDS 7 1 High-risk/low-risk bacteremia 341/291 54/46 Origin of bacteremia Urinary tract 170 27 Unknown 170 27 Vascular catheter 121 19 Respiratory tract 73 12 Intra-abdominal 64 10 Soft tissue 20 3 Other 14 2 Clinical presentation of: Pitt score of ⱖ2 271 43 Initial shock/MODS 160 25 Shock/MODS at 48 h 135 21 Antibiotic therapy Empirical 582 92 Adequate empirical 396 68 Additional 199 31.5 Outcome 30-day mortality 183 29 48-h mortality 64 10 a IQR, interquartile range. FIG 1 Episodes of bacteremia due to P. aeruginosa at participating hospitals. Pseudomonas aeruginosa Bloodstream Infections March 2012 Volume 56 Number 3 aac.asm.org 1267 TABLE 2 Baseline characteristics of patients with carbapenem-susceptible and carbapenem-resistant P. aeruginosa bacteremias Characteristic a Value for group RR 95% CI P CSPA (n⫽487 [77%]) CRPA (n⫽145 [23%]) Median age (yr) (IQR) 68.5 (56–79) 67.5 (57–76) 0.48 No. (%) of male patients 322 (66) 111 (77) 1.44 1.05–1.99 0.018 No. (%) with nosocomial/health care acquisition 454 (93) 145 (100) 1.07 1.05–1.1 ⬍0.001 Median no. of days from admission until bacteremia (IQR) 16 (7–28) 23.5 (13.5–43) ⬍0.001 No. (%) with ICU stay 112 (23) 58 (40) 1.7 1.3–2.2 ⬍0.001 Mean SAPS (SD) 42 (32–54) 39 (25–56) 0.71 Median Charlson index (IQR) 3 (1–4) 0.23 No. (%) with underlying condition Any 410 (84) 124 (85.5) 1.1 0.7–1.7 0.69 Solid malignancy 14 33 0.76 0.5–1.1 0.10 Diabetes 128 36 0.9 0.7–1.3 0.73 Immunosuppressive therapy 109 48 1.5 1.1–2.0 0.009 Cardiopathy 86 29 1.1 0.8–1.7 0.52 Chronic renal failure 73 27 1.2 0.8–1.9 0.29 Hematologic malignancy 73 20 0.9 0.6–1.5 0.72 Neurologic disease 33 10 1.0 0.5–2.0 0.96 Neutropenia 54 13 0.8 0.5–1.4 0.5 Cirrhosis 22 12 1.8 0.9–3.6 0.08 AIDS 5 2 1.3 0.3–6.9 0.66 No. (%) with origin of bacteremia High risk 272 (56) 69 (48) 0.9 0.7–1.0 0.08 Urinary tract 126 44 1.2 0.9–1.6 0.29 Vascular catheter 89 32 1.2 0.8–1.7 0.3 Unknown 140 30 0.7 0.5–1.0 0.05 Respiratory tract 52 21 1.4 0.8–2.2 0.21 Abdominal/biliar 52 12 0.8 0.4–1.4 0.4 Soft tissue 16 4 0.8 0.3–2.5 0.75 Other 12 2 0.6 0.2–2.2 0.44 No. (%) with clinical presentation of: Pitt score of ⱖ2 200 (41) 71 (49) 1.2 1.0–1.4 0.09 Initial shock/MODS 115 (24) 45 (31) 1.3 1.0–1.8 0.07 Shock/MODS at 48 h 95 (20) 40 (28) 1.4 1.0–1.9 0.04 No. (%) on antimicrobial empirical therapy 445 (91) 137 (94.5) 1.0 1.0–1.1 0.22 No. (%) with type of empirical treatment of: Carbapenems 94 (19) 52 (36) 1.9 1.4–2.5 ⬍0.001 Fluoroquinolones 59 (12) 12 (8) 0.7 0.4–1.2 0.20 Piperacillin-tazobactam 134 (28) 42 (29) 1.1 0.8–1.4 0.73 Antipseudomonal cephalosporins 110 (20.5) 26 (18) 0.9 0.8–1.0 0.23 Aminoglycosides 71 (15) 38 (26) 1.8 1.3–2.5 0.001 Colistin 6 (1) 6 (4) 3.4 1.1–10.3 0.04 No. (%) with adequate empirical therapy 338 (76) 58 (42) 0.6 0.5–0.7 ⬍0.001 No. (%) with adequate therapy b 438 (94) 131 (94) 1 1–1 1 No. (%) with delay of adequate treatment of ⬍24 h 286 (65) 52 (40) 0.6 0.5–0.8 ⬍0.001 No. (%) with additional therapy 142 (29) 57 (39) 1.3 1.1–1.7 0.021 No. (%) with outcome of: 30-day mortality 132 (27) 51 (35) 1.3 1.0–1.7 0.06 48-h mortality 63 (13) 17 (12) 0.9 0.5–1.5 0.70 a IQR, interquartile range. b Adequate empirical and/or definitive treatment (excluding 24 patients who did not receive either empirical or definitive treatment). Peña et al. 1268 aac.asm.org Antimicrobial Agents and Chemotherapy were 27% for the CSPA group (132 of 487 patients) and 35% for the CRPA group (51 of 145 patients). Figure 2 shows the 30-day survival curves and a comparison between the groups (P⫽0.10). Although the mortality rates were similar for the groups at the end of the first 72 h (63 [13%] patients in the CSPA group died, versus 17 [12%] in the CRPA group [P⫽0.7]), the Kaplan-Meier survival curves (Fig. 2) showed a higher mortality rate for the CSPA group in the first 2 days than for the CRPA group (12% versus 7% [the Pvalue was not significant]). The unadjusted analyses of the association of the cohort characteristics with mortality are shown in Table 3. The results of the crude analysis showed that carbapenem resistance did not have a significant impact on mortality (HR, 1.3; 95% CI, 0.9 to 1.8; P⫽0.11). An adjusted analysis with Cox regression controlled for the source of bacteremia (high/low risk), the Pitt index, and a delay of less than 24 h to receive adequate antimicrobial therapy showed a significant interaction between carbapenem resistance and the Charlson index (Table 4). At all time points considered, the highest risk of mortality associated with carbapenem resistance was observed for the patients with the lowest Charlson index scores. The risk fell steadily as the number of comorbidities rose. Carbapenem resistance significantly increased the risk of mortality from the fifth day after the onset of bacteremia (data not shown). However, this association was found to be not significant during the first 4 days, independent of the Charlson value. DISCUSSION The incidence of carbapenem resistance in P. aeruginosa strains has increased in recent years due to the growing resistance of this organism to antibiotics. This increase has been reported all over the world, although there are regional differences in rates (39). In fact, the nosocomial rates of carbapenem resistance in our study show a strong geographical variation inside Spain. Differences in antimicrobial consumption or epidemiological features in particular areas may contribute to these differences, although molecular analysis suggests that to some extent, the cross-transmission of carbapenem-resistant strains in the hospital and even in the interhospital setting may be responsible. This report is the largest prospective cohort study to include a detailed clinical investigation of the impact of carbapenem resistance on mortality in patients with P. aeruginosa bacteremia. Our data show that carbapenem resistance was associated with significantly increased 30-day mortality rates but that the relationship depended on the degree of severity of the underlying disease (19, 30). In fact, our study describes an interaction between carbapenem resistance and underlying conditions, with an impact on mortality that is significantly higher for less severely ill patients. In spite of the belief that infections caused by antibioticresistant organisms result in higher mortality rates, the studies that have evaluated this problem have produced conflicting results. There are many possible reasons for these discrepancies, since the data are highly dependent on the methods used (6). The differences observed may result, in part, from limitations in the design of retrospective studies (2, 9, 25, 26, 28, 40, 41), usually performed to identify the predictors of mortality (2, 25, 26, 28, 40, 41), and an inability to properly control for illness severity and underlying disease (16). Another potential confounder is the different pathogenic potentials of the causative pathogens under study (5, 27, 33). In addition, the types of infection, particularly those with poor prognoses, may differ or may not be balanced between the groups being compared (6). Finally, many studies had smaller numbers of patients with P. aeruginosa infection, thereby decreasing their statistical power. FIG 2 Kaplan-Meier survival curves of carbapenem-resistant and carbapenem-susceptible P. aeruginosa bacteremias (P⫽0.10 by log rank test). Pseudomonas aeruginosa Bloodstream Infections March 2012 Volume 56 Number 3 aac.asm.org 1269 Few studies have quantified the direct impact of antibiotic resistance (9, 31, 32). Carmeli et al. (9) showed previously that resistance to P. aeruginosa was not associated with increased mortality. In another study (31), despite the increases in the proportion of resistant organisms observed, no significant differences were found for the relative risk of death due to a resistant phenotype of the target pathogen. Finally, a recent cohort study (32) showed that the additional effect of antimicrobial resistance on mortality related to health care-associated infections was comparatively low. Patients with a greater severity of underlying disease may have many reasons for increased mortality, and the addition of bacteremia due to antimicrobial resistance to their life-threatening problems may be inconsequential. Thus, it is essential to select an adequate methodology (6, 17, 29) and adjust for the severity of the illness and comorbidities before infection, as these factors can have a significant effect on the measurement of outcomes. We feel that we used the best methodology available to us; certain research questions can be answered only by use of observational studies, due to obvious ethical constraints. By using a predetermined uniform endpoint of mortality at 30 days, we avoided the potential bias in the assignment of a cause of death, which is particularly problematic for patients with many comorbid conditions. Regarding the type of infection, microbiologically documented bacteremia represents the optimal model for investigating the pathogenic significance of resistance. Finally, we used a standardized underlying disease index, the Charlson comorbidity score, a valid method for estimating the risk of death from comorbid disease, and we measured comorbidities before the development of bacteremia. Finally, our study suggests that the increase in mortality explained by carbapenem resistance is not evident during the first days after the onset of bacteremia, as a recent single-center study indicated (40). These findings may be consistent with in vitro ev- TABLE 3 Crude analysis of effects of carbapenem resistance and other variables on mortality Covariate a Value for group HR b 95% CI PCensored Died Total no. (%) 449 (71) 183 (29) No. (%) with carbapenem resistance 94 (21) 51 (28) 1.3 0.9–1.8 0.11 Median age (yr) (range) 65.5 (54.5–76.5) 71.5 (60.5–79.5) 1.1 1.0–1.0 0.002 No. (%) of male patients 311 (69) 122 (67) 1.1 0.8–1.5 0.60 No. (%) with nosocomial/health care acquisition 424 (94) 175 (96) 1.1 0.5–2.2 0.80 Median Charlson index (IQR) 2 (1–4) 3 (2–5) 1.1 1.1–1.2 ⬍0.001 No. (%) with high-risk bacteremia 207 (46) 134 (73) 2.7 2.0–3.8 ⬍0.001 No. (%) with origin of bacteremia Urinary tract 142 (32) 28 (15) 0.4 0.3–0.7 ⬍0.001 Vascular catheter 100 (22) 21 (11) 0.5 0.3–0.8 0.002 Unknown 99 (22) 71 (39) 2.0 1.5–2.7 ⬍0.001 Respiratory tract 34 (7.5) 39 (21) 2.5 1.8–3.6 ⬍0.001 Abdominal/biliar 54 (12) 10 (5.5) 0.5 0.2–0.9 0.02 Soft tissue 11 (2.5) 9 (5.5) 1.6 0.8–3.1 0.17 Other 9 (2) 5 (3) 1.3 0.5–3.1 0.60 No. (%) with clinical presentation of: Pitt score of ⱖ2 139 (30) 132 (72) 4.5 3.3–6.3 ⬍0.001 Initial shock/MODS 68 (15) 92 (50) 4.2 3.1–5.6 ⬍0.001 Shock/MODS at 48 h 38 (9) 97 (56) 7.7 5.7–10.4 ⬍0.001 No. (%) with adequate empirical antimicrobial therapy 290 (70) 106 (64) 0.8 0.6–1.1 0.19 No. (%) with delay of adequate treatment of ⬍24 h 241 (57) 97 (67) 1.5 1.1–2.1 0.02 No. (%) with additional therapy 166 (37) 33 (18) 0.4 0.3–0.6 ⬍0.001 a IQR, interquartile range. b HR, hazard ratio. TABLE 4 Adjusted a effect of carbapenem resistance on mortality b Day Charlson index aHR c 95% CI P 2 0 1.7 0.8–3.3 0.16 1 1.5 0.8–3.0 0.24 2 1.1 0.6–1.9 0.83 3 0.8 0.5–1.6 0.67 4 0.6 0.3–1.0 0.55 5 0.6 0.3–1.1 0.09 7 0 2.2 1.2–4.1 0.01 1 2.0 1.1–3.8 0.01 2 1.4 0.9–2.3 0.12 3 1.3 0.8–2.1 0.36 4 0.8 0.5–1.3 0.31 5 0.7 0.4–1.3 0.31 30 0 9.9 3.3–29.4 ⬍0.001 1 8.7 2.9–26.2 ⬍0.001 2 5.8 2.0–16.4 0.001 3 6.7 1.8–24.3 0.004 4 3.5 1.1–11.0 0.04 5 2.6 0.8–8.0 0.1 a Adjusted by the source of bacteremia (high risk/low risk), the Pitt score, and a delay of less than 24 h to receive adequate antibiotic therapy. b Shown is the interaction between carbapenem susceptibility and the Charlson index at different time points. c aHR, adjusted hazard ratio. Peña et al. 1270 aac.asm.org Antimicrobial Agents and Chemotherapy idence showing that resistance genes or mutations can alter the fitness of microorganisms (4), making them less virulent and reducing their capacity to generate deleterious host inflammatory responses (22). Nevertheless, the complex interplay between resistance mechanisms, cost-compensatory mutations, fitness, and virulence in the clinical setting has yet to be fully elucidated (1). The present investigation may have limitations. Unfortunately, because of the lack of statistical power to detect true but small differences, the inferences that can be made from our data may be limited despite the large sample size. In addition, a system to score the severity of acute illness, which is well validated for predicting bacteremia prognoses (11), was assessed at the time of bacteremia presentation. If the severity of the illness was an intermediate variable in the causal pathway between carbapenem resistance and mortality, adjustment by that variable could have led to an underestimation of the magnitude of the effect (16); however, no such association was demonstrated. In our opinion, this multi-institutional study reflects a broad spectrum of patients, and given the fact that we derived our cohort from a multicenter study, we believe that these results are applicable to other settings. In conclusion, to our knowledge, this is the first prospective multicenter study to investigate the adjusted impact of antimicrobial resistance on the mortality of P. aeruginosa bloodstream infections. We have documented the distribution of this epidemiological problem in Spain. The hospitals studied showed marked diversity in the occurrence of these strains, and this heterogeneity is due, at least in part, to clonal dissemination. The study clarifies the relationship between carbapenem-resistant P. aeruginosa bacteremia and mortality, although the impact was observed mainly for patients with a lower severity of underlying disease. The lesser effect of resistance during the first days after the onset of bacteremia may have implications for the design of empirical antibiotic strategies. Better risk scores should be developed to distinguish patients at a low risk of mortality due to infections with antimicrobial-resistant pathogens from those for whom the broadest-spectrum therapy could be truly beneficial. ACKNOWLEDGMENTS This work was supported by National Health Service grant FIS 08/0276 from the Fondo de Investigación Sanitarias and was also supported by the Ministerio de Ciencia e Innovación, Instituto de Salud Carlos III, cofinanced by European Regional Development Fund, A Way To Achieve Europe; by the Spanish Network for the Research in Infectious Diseases (REIPI RD06/0008); and by the Ciber de Enfermedades Respiratorias (CB06/06/0037). We thank other participants from the REIPI group: Mercedes Gurgui (Hospital Universitari de la Santa Creu i Sant Pau, Barcelona, Spain); Roger Sorde and Nieves Larrosa (Hospital Universitario Vall d’Hebrón); Cecilia Martín (Hospital Universitario Virgen del Rocío, Seville, Spain); Dionisia Fontanals (Consorci Hospitalari Parc Taulí, Sabadell, Barcelona, Spain); Marina de Cueto and Maria Dolores Navarro (Hospital Universitario Virgen Macarena, Seville, Spain); and Julian Torre-Cisneros, Manuel Casal, Rosario Lara, Clara Natera, and Antonio Rivero (Hospital Universitario Reina Sofía-IMIBIC, Córdoba, Spain). B.A. has served as a speaker for Gilead Sciences, Merck, Sharp and Dohme, Pfizer, and Novartis and has received research support from Gilead Sciences and Pfizer. J.R.-B. has been a consultant for Wyeth, Merck, and Pfizer; has served as speaker for Wyeth, Merck, Pfizer, Astra- Zeneca, and GlaxoSmithKline; and has received research support from Merck and Wyeth. L.M.-M. has been a consultant for Wyeth and Pfizer; has served as a speaker for Wyeth, Merck, Pfizer, and Janssen-Cilag; and has received research support from Merck, Wyeth, and Janssen-Cilag. A.O. has received research support from Janssen-Cilag. All other authors have no conflicts. REFERENCES 1. Abdelraouf K, Kabbara S, Ledesma KR, Poole K, Tam VH. 2011. Effect of multidrug resistance-conferring mutations on the fitness and virulence of Pseudomonas aeruginosa. J. Antimicrob. Agents 66:1311–1317. 2. Aloush V, Navon-Venezia S, Seigman-Igra Y, Cabili S, Carmeli Y. 2006. Multidrug-resistant Pseudomonas aeruginosa: risk factors and clinical impact. Antimicrob. Agents Chemother. 50:43–48. 3. American College of Chest Physicians/Society of Critical Care Medicine. 1992. American College of Chest Physicians/Society of Critical Care Medicine Consensus Conference: definitions for sepsis and organ failure and guidelines for use of innovative therapies in sepsis. Crit. Care. Med. 20:864–874. 4. Andersson DI. 2006. The biological cost of mutational antibiotic resistance: any practical conclusions? Curr. Opin. Microbiol. 9:461–465. 5. Blot S, Vandewoude K, De Bacquer D, Colardyn F. 2002. Nosocomial bacteremia caused by antibiotic-resistant gram-negative bacteria in critically ill patients: clinical outcome and length of hospitalization. Clin. Infect. Dis. 34:1600–1606. 6. Blot S, Depuydt P, Vandewoude K, De Bacquer D. 2007. Measuring the impact of multidrug resistance in nosocomial infection. Curr. Opin. Infect. Dis. 20:391–396. 7. Cabot G, et al. 2011. Overexpression of AmpC and efflux pumps in Pseudomonas aeruginosa isolates from bloodstream infections: prevalence and impact on resistance in a Spanish multicenter study. Antimicrob. Agents Chemother. 55:1906–1911. 8. Carmeli Y, Troillet N, Eliopoulos GM, Samore MH. 1999. Emergence of antibiotic-resistant Pseudomonas aeruginosa: comparison of risks associated with different antipseudomonal agents. Antimicrob. Agents Chemother. 43:1379–1382. 9. Carmeli Y, Troillet N, Karchmer AW, Samore MH. 1999. Health and economic outcomes of antibiotic resistance in Pseudomonas aeruginosa. Arch. Intern. Med. 159:1127–1132. 10. Charlson ME, Pompei P, Ales KL, MacKenzie CR. 1987. A new method of classifying prognostic comorbidity in longitudinal studies: development and validation. J. Chronic Dis. 40:373–383. 11. Chow JW, et al. 1991. Enterobacter bacteremia: clinical features and emergence of antibiotic resistance during therapy. Ann. Intern. Med. 115: 585–590. 12. Clinical and Laboratory Standards Institute. 2009. Performance standards for antimicrobial disk susceptibility test; approved standard, 10th ed. CLSI document M02-A10. Clinical and Laboratory Standards Institute, Wayne, PA. 13. Clinical and Laboratory Standards Institute. 2010. Performance standards for antimicrobial susceptibility testing: 20th informational supplement. CLSI document M100-S20. Clinical and Laboratory Standards Institute, Wayne, PA. 14. Collet D. 2003. Model checking in the Cox regression model, p 111–150. In Collet D (ed), Modelling survival data in medical research, 2nd ed. Oxford University Press, London, United Kingdom. 15. Cosgrove SE, Kaye KS, Eliopoulous GM, Carmeli Y. 2002. Health and economic outcomes of the emergence of third-generation cephalosporin resistance in Enterobacter species. Arch. Intern. Med. 162:185–190. 16. Cosgrove SE, Carmeli Y. 2003. The impact of antimicrobial resistance on health and economic outcomes. Clin. Infect. Dis. 36:1433–1437. 17. Cosgrove SE. 2006. The relationship between antimicrobial resistance and patient outcomes: mortality, length of hospital stay and health care costs. Clin. Infect. Dis. 42(Suppl 2):S82–S89. 18. European Antimicrobial Resistance Surveillance System/National Institute for Public Health and the Environment. 2009. European Antimicrobial Resistance Surveillance System annual report, 2008. European Antimicrobial Resistance Surveillance System/National Institute for Public Health and the Environment, Bilthoven, Netherlands. 19. Figuereido S. 2008. Impact of antimicrobial resistance on the treatment and outcome of patients with sepsis. Shock 30(Suppl 1):S23–S29. 20. Friedman ND, et al. 2002. Health care-associated bloodstream infections in adults: a reason to change the accepted definition of communityacquired infections. Ann. Intern. Med. 137:791–797. 21. Garner JS, Jarvis WR, Emori TG, Horan TC, Hughes JM. 1988. CDC Pseudomonas aeruginosa Bloodstream Infections March 2012 Volume 56 Number 3 aac.asm.org 1271 definitions for nosocomial infections, 1988. Am. J. Infect. Control 16: 128–140. 22. Giamarellos-Bourboulis EJ, et al. 2004. Stimulation of innate immunity by susceptible and multidrug-resistant Pseudomonas aeruginosa an in vitro and in vivo study. Clin. Exp. Immunol. 135:240–246. 23. Gutiérrez O, et al. 2007. Molecular epidemiology and mechanisms of carbapenem resistance in Pseudomonas aeruginosa isolates from Spanish hospitals. Antimicrob. Agents Chemother. 51:4329–4335. 24. Harbarth S, Nobre V, Pittet D. 2007. Does antibiotic selection impact patient outcome? Clin. Infect. Dis. 44:87–93. 25. Joo EJ, et al. 2011. Risk factors for mortality in patients with Pseudomonas aeruginosa bacteremia: clinical impact of antimicrobial resistance on outcome. Microb. Drug Resist. 17:305–312. 26. Kang CI, et al. 2003. Pseudomonas aeruginosa bacteremia: risk factors for mortality and influence or delayed receipt of effective antimicrobial therapy on clinical outcome. Clin. Infect. Dis. 37:745–751. 27. Kang CI, et al. 2005. Bloodstream infections caused by antibioticresistant gram-negative bacilli: risk factors for mortality and impact of inappropriate initial antimicrobial therapy on outcome. Antimicrob. Agents Chemother. 49:760–766. 28. Kang CI, et al. 2005. Risk factors for antimicrobial resistance and influence of resistance on mortality in patients with bloodstream infection caused by Pseudomonas aeruginosa. Microb. Drug Resist. 11:68–74. 29. Kaye KS, Engemann JJ, Mozaffari E, Carmeli YY. 2004. Reference group choice and antibiotic resistance outcomes. Emerg. Infect. Dis. 10:1125– 1128. 30. Kim PW, et al. 2005. Risk of mortality with a bloodstream infection is higher in the less severely ill at admission. Am. J. Respir. Crit. Care Med. 171:616–620. 31. Klevens RM, Edwards JR, Gaynes RP, and the National Nosocomial Infectious Surveillance System. 2008. The impact of antimicrobialresistant health care-associated infections on mortality in the United States. Clin. Infect. Dis. 47:927–930. 32. Lambert ML, et al. 2011. Clinical outcomes of health-care-associated infections and antimicrobial resistance in patients admitted to European intensive-care units: a cohort study. Lancet Infect. Dis. 11:30–38. 33. Lautenbach E, Patel JB, Bilker WB, Edelstein PH, Fishman NO. 2001. Extended-spectrum ␤ -lactamase-producing Escherichia coli and Klebsiella pneumoniae: risk factors for infection and impact of resistance on outcomes. Clin. Infect. Dis. 32:1162–1171. 34. Le Gall JR, et al. 1984. A simplified acute physiologic score for ICU patients. Crit. Care Med. 12:975–977. 35. Livermore DM. 2002. Multiple mechanisms of antimicrobial resistance in Pseudomonas aeruginosa: our worst nightmare? Clin. Infect. Dis. 34:634– 640. 36. Magiorakos AP, et al. 7 May 2011, posting date. Multidrug resistant, extensively drug-resistant and pandrug-resistant bacteria: an international expert proposal for interim standard definitions for acquired resistance. Clin. Microbiol. Infect. [Epub ahead of print.] doi:10.1111/j.1469- 0691. 37. National Nosocomial Infections Surveillance System. 2004. System report data summary from January 1992 through June 2004, issued October 2004. Am. J. Infect. Control 32:470–485. 38. Rothman KJ, Greenland S. 2008. Introduction to stratified analysis, p 258–282. In Rothman KJ, Greenland S, Lash TL (ed), Modern epidemiology, 3rd ed. Little Brown, Boston, MA. 39. Souli M, Galani I, Giamarellou H. 2008. Emergence of extensively drugresistant and pandrug-resistant Gram-negative bacilli in Europe. Euro Surveill. 13(47):pii⫽19045. http://www.eurosurveillance.org/ViewArticle .aspx?ArticleId⫽19045. 40. Suarez C, et al. 2010. Influence of carbapenem resistance on mortality and the dynamics of mortality in Pseudomonas aeruginosa bloodstream infection. Int. J. Infect. Dis. 14S:e73–e78. 41. Tam VH, et al. 2010. Impact of multidrug-resistant Pseudomonas aeruginosa bacteremia on patient outcomes. Antimicrob. Agents Chemother. 54:3717–3722. 42. Tenover FC, et al. 1995. Interpreting chromosomal DNA restriction patterns produced by pulsed-field gel electrophoresis: criteria for bacterial strain typing. J. Clin. Microbiol. 33:2233–2239. 43. Vidal F, et al. 1996. Epidemiology and outcome of Pseudomonas aeruginosa bacteremia, with special emphasis on the influence of antibiotic treatment. Arch. Intern. Med. 156:2121–2126. Peña et al. 1272 aac.asm.org Antimicrobial Agents and Chemotherapy