- Email: [email protected]
Evaluation of clinical outcomes in patients with Gram-negative bloodstream infections according to cefepime MIC
Diagnostic Microbiology and Infectious Disease 82 (2015) 165–171
Contents lists available at ScienceDirect
Diagnostic Microbiology and Infectious Disease journal homepage: www.elsevier.com/locate/diagmicrobio
Evaluation of clinical outcomes in patients with Gram-negative bloodstream infections according to cefepime MIC☆ Nathaniel J. Rhodes a,b, Jiajun Liu c, Milena M. McLaughlin a,b, Chao Qi d, Marc H. Scheetz a,b,⁎ a
Department of Pharmacy Practice, Midwestern University, Chicago College of Pharmacy, Downers Grove, IL, USA Department of Pharmacy, Northwestern Memorial Hospital, Chicago, IL, USA Midwestern University, Chicago College of Pharmacy, Downers Grove, IL, USA d Department of Pathology, Clinical Microbiology Division, Northwestern Memorial Hospital, Chicago, IL, USA b c
a r t i c l e
i n f o
Article history: Received 29 July 2014 Received in revised form 26 February 2015 Accepted 3 March 2015 Available online 10 March 2015 Keywords: Cefepime Gram-negative bacteria Bloodstream infection Minimum inhibitory concentration Mortality
a b s t r a c t Predicted and observed failures at higher cefepime MICs have prompted the Clinical and Laboratories Standards Institute (CLSI) to lower the susceptible breakpoint for Enterobacteriaceae to ≤2 mg/L, with dose-dependent susceptibility at 4–8 mg/L, while the susceptibility breakpoint for nonfermentative organisms remain unchanged at ≥8 mg/L. The contribution of increasing cefepime MIC to mortality risk in the setting of aggressive cefepime dosing is not well defined. Patients who were treated with cefepime for Gram-negative blood stream infections (GNBSIs), including both Enterobacteriaceae and nonfermentative organisms, were screened for inclusion in this retrospective cohort study. Demographic and microbiologic variables were collected, including pathogen, cefepime MIC, dosage, and interval. The objective was to define a risk-adjusted mortality breakpoint for cefepime MICs. Secondarily, we looked at time to death and length of stay (LOS) postculture. Ninety-one patients were included in the analysis. Overall, 19 patients died and 72 survived. Classification and Regression Tree analysis identified an inhospital mortality breakpoint at a cefepime MIC between 2 and 4 mg/L for patients with a modified Acute Physiology and Chronic Health Evaluation II score ≤16.5 (4.2% versus 25%, respectively). Multivariate logistic regression revealed increased odds of mortality at a cefepime MIC of 4 mg/L (adjusted odds ratio [aOR] 6.47; 95% confidence interval [CI] 1.25–33.4) and 64 mg/L (aOR 6.54, 95% CI 1.03-41.4). Those with cefepime MICs ≥4 mg/L experienced a greater median intensive care unit LOS for survivors (16 versus 2 days; P = 0.026). Increasing cefepime MIC appears to predict inhospital mortality among patients who received aggressive doses of cefepime for GNBSIs, supporting a clinical breakpoint MIC of 2 mg/L. © 2015 Elsevier Inc. All rights reserved.
1. Introduction In 2010, the Clinical and Laboratories Standards Institute (CLSI) lowered the cephalosporin MIC breakpoints for the Enterobacteriaceae, except for cefepime, a broad-spectrum fourth-generation cephalosporin (CLSI, 2010). The CLSI retained the cefepime resistance breakpoint of 8 mg/L for Enterobacteriaceae and nonfermentative organisms based on data suggesting that adequate doses of cefepime should provide a high likelihood of pharmacokinetic-pharmacodynamic (PK/PD) target attainment (CLSI, 2010; Lee et al., 2007; Nicasio et al., 2009). In contrast, the European Committee on Antimicrobial Susceptibility Testing (EUCAST) susceptibility breakpoint is ≤1 mg/L, while the resistance breakpoint has been ≥4 mg/L for Enterobacteriaceae since 2010 (EUCAST, 2014). Similar to CLSI guidelines, the EUCAST cefepime breakpoint for Pseudomonas spp. remains ≤8 mg/L.
☆ Portions of this paper were presented as a platform at the 53rd Interscience Conference on Antimicrobial Agents and Chemotherapy, Denver, CO, September 10–13, 2013. ⁎ Corresponding author. Tel.: +1-630-515-6116; fax: +1-630-515-6958. E-mail address: [email protected] (M.H. Scheetz). http://dx.doi.org/10.1016/j.diagmicrobio.2015.03.005 0732-8893/© 2015 Elsevier Inc. All rights reserved.
Early reports identified clinical failures at cefepime MICs in the range of 2–16 mg/L (Song et al., 2005; Wi et al., 2009), and an initial clinical breakpoint (defined by a significantly elevated risk of inhospital mortality) of 8 mg/L for Gram-negative (GN) bloodstream infections (BSIs) was determined (Bhat et al., 2007). However, most patients received the dose equivalent of 1–2 g every 12 hours, which is less than the PK/ PD optimized dosing (Bhat et al., 2007; Kuti et al., 2004; Lee et al., 2007; Nicasio et al., 2009). Others have observed worse outcomes with increased cefepime MICs (8–16 mg/L) in the setting of extendedspectrum beta-lactamase (ESBL)–positive GNBSI, but these investigations did not evaluate the impact of cefepime dosing or the effect of incremental increases in cefepime MIC on clinical outcomes (Chopra et al., 2012; Lee et al., 2013; Tumbarello et al., 2007). Given the clinical data, findings from PK/PD modeling and simulation, and the common use of lower cefepime doses, the 2014 update to the CLSI guidelines have included a reduced susceptibility breakpoint for cefepime at ≤2 mg/L with dose-dependent susceptibility defined between 4 and 8 mg/L and resistant defined at ≥16 mg/L for Enterobacteriaceae (CLSI, 2014). Despite the aforementioned studies, clinical outcomes studies of GNBSIs with cefepime MICs between 2 and 8 mg/L are lacking. We sought to identify a clinical mortality breakpoint among patients treated
166
N.J. Rhodes et al. / Diagnostic Microbiology and Infectious Disease 82 (2015) 165–171
with aggressively dosed cefepime for various GNBSI. Secondarily, we sought to examine the clinical breakpoint for other outcomes such as time to death and hospital and intensive care unit (ICU) length of stay (LOS) postculture.
isolates using Vitek 2 cards (bioMérieux, Balmes-les-Grottes, France). Organism susceptibility was interpreted for clinicians according to CLSI guidelines at the time of culture (CLSI, 2013). In cases of mixed GNBSI, the most resistant organism (highest cefepime MIC) was used for analyses.
2. Materials and methods 2.4. Statistical analysis 2.1. Study design This retrospective, cohort study was conducted at Northwestern Memorial Hospital in Chicago, Illinois. Study methods were reviewed and approved by institutional review boards at Northwestern University and Midwestern University. Patients with at least 1 positive GNBSI treated with cefepime between September 1, 2006, and August 31, 2012, were reviewed for inclusion. Patients were included only once, and the index culture was the first GNBSI. Patients were randomly selected to fill MIC categories. A maximum of 4 bloodstream isolates with a cefepime MIC ≤1 mg/L was included for every isolate with a cefepime MIC ≥2 mg/L to balance power (Ury, 1975). 2.2. Definitions ICU-onset infection was defined as patient ICU status at the time of the GNBSI. Concurrent renal dysfunction was defined as acute or chronic renal dysfunction. Acute renal dysfunction was an increase in serum creatinine of 0.5 mg/dL or 50% from baseline to immediately before the infection (Bellomo et al., 2004). Chronic renal dysfunction was defined as a diagnosis of chronic kidney dysfunction. Hepatic dysfunction was defined as any liver enzyme greater than 3 times the upper normal limit at the time of culture, chronic hepatitis, or documented cirrhosis. Prior and concurrent immunosuppression was defined as steroid, chemotherapy, or immunosuppressant use within the prior 12 months. Patient comorbidities were considered to be present if they were documented in the admission history and physical. The source of the BSI was considered from the attending physician's written diagnosis. Other active therapy in addition to or after cefepime receipt was considered for relevant GN drugs. Cefepime dose intensity was defined according to the recommendations in the approved product labeling as: 2 g every 8 hours, 2 g every 12 hours, 1 g every 12 hours, 500 mg every 12 hours, or the renal dysfunction–adjusted equivalent doses (Hospira, 2013). GN organisms isolated from blood cultures were treated as categorical variables according to genera and lactose fermentative phenotype. 2.3. Data elements Variables were collected from medical, pharmacy, and microbiology records by trained reviewers. Variables collected included demographics (age, gender, and race), ICU versus non-ICU admission, modified Acute Physiology and Chronic Health Evaluation II (APACHE II) component and composite score (Hamilton et al., 2007; Thom et al., 2008), renal dysfunction, hepatic dysfunction, solid organ or hematologic transplant, diabetes, concurrent immunosuppression, hospitalizations within 12 months before culture, hospital LOS postculture, polymicrobial bacteremia (excluding coagulase-negative staphylococci), infecting organism, infection source, the name and dose of all GN antibiotics received, and organism cefepime MIC. Outcomes included inhospital and 30-day all-cause mortality, time to death, days to negative culture, ICU transfer, hospital and ICU LOS for survivors, 30-day readmission, inpatient antimicrobial duration, and disposition postadmission. Mortality at 30 days was assessed using the Social Security Death Master file. Empiric and directed cefepime was dosed prospectively and adjusted by clinical pharmacists according to internal protocols. The impact of dosing was assessed as the ratio of the 24-hour renal dose equivalent divided by the pathogen MIC (dose-to-MIC ratio). GN organisms were identified using the Vitek 2 system (bioMérieux, Balmes-les-Grottes, France). Susceptibility testing was performed on all
The primary outcome (i.e., dependent variable) was the covariateadjusted odds of inhospital mortality. Cefepime MIC was assessed as the primary predictor; however, cefepime dose and dose-to-MIC ratio were also assessed as substitutes for MIC (individually, due to collinearity). Cefepime MIC was handled as an interval variable (log-linear), a categorical variable (at each doubling MIC dilution category from 1 to 64 mg/L), and binned as a dichotomous variable via classification and regression tree (CART). CART was performed using SPSS version 19 and the Decision Trees add-on version 19 (SPSS, Chicago, IL, USA). All relevant covariates (Table S1) were considered for model inclusion. All other analyses were performed using Intercooled Stata version 13.0 (Statacorp, College Station, TX, USA). Secondary outcomes (i.e., additional dependent variables) were time to death and LOS postculture. Continuous variables were evaluated with Student's t test or Wilcoxon rank sum test, as appropriate. Categorical variables were evaluated with the chi-square or Fisher's exact test, as appropriate. Nonnormal variables were normalized using log transformation to permit parametric assessments. Multivariate analyses involved the use of logistic regression, which was performed in a stepwise fashion. Candidate covariates (Table 1) at a bivariate level of significance of P b 0.2 were further assessed as possible independent predictors of the primary outcome (Esterly et al., 2012). Variables were retained in models if the objective function value changed by N3.84 with each iterative addition (n + 1 model). Final models were secondarily checked for optimal parsimony utilizing the Aikaike Information Criterion (AIC) (Hosmer and Lemeshow, 2000). Outcome probabilities, adjusting for comodeled covariates, were calculated from adjusted odds ratios (aORs). Since cefepime MIC was the primary variable of interest, it was the major variable considered for CART splitting; however, as previous studies demonstrated the importance of morbidity indices for CART analyses (Lodise et al., 2007), a tree with 2 splits was created with APACHE II score at the first node and cefepime MIC at the second node. We assessed competing logistic regression models for mortality with cefepime MIC as a predictor using: the CART-derived cefepime MIC breakpoint (model 1), categorical cefepime MIC (model 2), and log2-transformed cefepime MIC (model 3). Modified APACHE II was forced into all models, as it is a known predictor of mortality (Hamilton et al., 2007; Thom et al., 2008). As additional exploratory analyses, time-to-event analyses were conducted by assessing variables identified in multivariate logistic regression models using Cox regression models for the categorical MIC (model 4) and CART-derived breakpoint MIC (model 5). Kaplan–Meier analyses with log-rank tests were conducted for bivariate significance determination and graph generation for all time-to-event analyses. All tests were 2 tailed, with an a priori level of alpha set at 0.05 for statistical significance. 3. Results 3.1. Baseline characteristics A total of 223 patients were screened for inclusion, and 91 patients met inclusion criteria during the study period. CART defined a clinical MIC inhospital mortality breakpoint when controlling for modified APACHE II score on day 0. Splits were identified at a modified APACHE II score of ≤16.5 and N16.5 and at a cefepime MIC of ≤2 mg/L and ≥4 mg/L (Fig. 1). Patients were then stratified according to MIC breakpoint category (≤2 mg/L or ≥4 mg/L) with baseline demographics displayed in Table 1. Predictors were relatively similar between
N.J. Rhodes et al. / Diagnostic Microbiology and Infectious Disease 82 (2015) 165–171
167
Table 1 Baseline characteristics by CART-derived breakpoint group and inhospital mortality outcome. Cefepime MIC ≤2 mg/L
Cefepime MIC ≥4 mg/L
Parameter
n = 55
n = 36
Age in years, mean (SD) Male, n (%) Race White, n (%) Black, n (%) Asian, n (%) Hispanic, n (%) Other, n (%) Modified APACHE II on day of culture, mean (SD) Admit from Emergency department, n (%) Outside hospital, n (%) Outpatient clinic, n (%) Community, n (%) Serum creatinine on day of culture, median (IQR) ICU at culture, n (%) ANC b500 cells/mm3 during admission, n (%) Mechanical ventilation in prior 12 months, n (%) History of leukemia, n (%) History of lymphoma, n (%) History of myeloma, n (%) Renal dysfunction within 5 days of culture, n (%) Diabetes, n (%) Receipt of immune suppressants within 12 months, n (%) Days to positive culture, median (IQR) Days to cefepime therapy, median (IQR) Received other active therapy in addition to cefepime, n (%) Source known, n (%) Source, n (%) Urine (n = 10) Respiratory (n = 1) Catheter tip (n = 7) GI or skin (n = 11) Central line (n = 21) Gram-negative genera, n (%) Acinetobacter baumannii (n = 3) Acinetobacter spp. (n = 1) Citrobacter spp. (n = 4) Delftia spp. (n = 1) Enterobacter spp. (n = 5) Escherichia coli (n = 32) Klebsiella spp. (n = 15) Pseudomonas spp. (n = 30) Polymicrobial infection, n (%) ESBL positive, n (%)
62.1(14.0) 26 (47.3)
59.9 (13.9) 24(66.7)
40 (72.7) 7 (12.7) 0 (0) 4 (7.2) 4 (7.2) 17.3 (4.32)
18 (50) 11 (30.5) 2 (5.6) 3 (8.3) 2 (5.6) 16.1 (5.12)
24 (43.6) 6 (10.9) 8 (14.6) 17 (30.9) 0.88 (0.7–1.5) 12 (21.8) 41 (74.6) 9 (16.4) 11 (20) 8 (14.6) 8 (14.6) 22 (40) 14 (25.5) 44 (80) 3.7 (1.6–14.6) 0.09 (0.03–0.44) 54 (98.2) 33 (60)
20 (55.6) 9 (25) 3 (8.3) 4 (11.1) 1.02 (0.77–1.33) 13 (36.1) 20 (55.6) 12 (33.3) 8 (22.2) 7 (19.4) 3 (8.3) 17 (47.2) 6 (16.7) 25 (69.4) 7.2 (1.9–18.3) 0.24 (0–0.97) 32 (88.9) 17 (47.2)
7 (21.2) 0 (0) 4 (12.1) 11 (33.3) 11 (33.3)
3 (17.7) 1 (5.88) 3 (17.7) 0 (0) 10 (58.8)
0 (0) 1 (1.82) 3 (5.45) 1 (1.82) 3 (5.45) 26 (47.3) 10 (18.2) 11 (20) 18 (32.7) 3 (5.45)
3 (8.33) 0 (0) 1 (2.78) 0 (0) 2 (5.56) 6 (16.7) 5 (13.9) 19 (52.8) 13 (36.1) 13 (36.1)
P-value
0.45 0.07 0.06 0.03 0.04 0.15 N0.99 N0.99 0.22 0.052 0.27 0.08 0.52 0.04 0.44 0.14 0.06 0.07 0.80 0.54 0.52 0.50 0.32 0.25 0.35 0.64 0.08 0.23 0.019 N0.99 0.34 0.67 0.009 0.084 0.002 0.06 N0.99 N0.99 N0.99 N0.99 0.003 0.59 0.001 0.74 b0.001
Survived inhospital
Died inhospital
n = 72
n = 19
61 (14.7) 39 (54.2)
62.1 (10.6) 11 (57.9)
48 (66.7) 13 (18.1) 2 (2.8) 6 (8.3) 3 (4.2) 16.6 (4.8)
10 (52.6) 5 (26.3) 0 (0) 1 (5.3) 3 (15.8) 17.5 (4.2)
34 (47.2) 10 (13.9) 10 (13.9) 18 (25) 0.90 (0.77–1.30) 12 (16.7) 46 (63.9) 19 (26.8) 12 (16.7) 12 (16.7) 9 (12.5) 30 (41.7) 19 (26.4) 51 (70.8) 2.5 (1.7–12.5) 0.19 (0.02–0.76) 67 (93.1) 46 (63.9)
10 (52.6) 5 (26.3) 1 (5.3) 3 (15.8) 1.15 (0.57–1.70) 13 (68.4) 15 (79.0) 2 (10.5) 7 (36.8) 3 (15.8) 2 (10.5) 9 (47.4) 1 (5.26) 18 (94.7) 15.7 (6.5–28.8) 0.04 (0–0.44) 19 (100) 4 (21.1)
9 (19.6) 1 (2.17) 6 (13.0) 10 (21.7) 20 (43.5)
1 (25) 0 (0) 1 (25) 1 (25) 1 (25)
2 (2.78) 1 (1.39) 3 (4.17) 1 (1.39) 3 (4.17) 27 (37.5) 12 (16.7) 23 (31.9) 23 (31.9) 13 (18.1)
1 (5.26) 0 (0) 1 (5.26) 0 (0) 2 (10.53) 5 (26.3) 3 (15.8) 7 (36.8) 8 (42.1) 3 (15.8)
P-value
0.77 0.77 0.31 0.26 0.42 N0.99 N0.99 0.10 0.45 0.42 0.68 0.19 0.45 0.55 0.82 b0.001 0.28 0.22 0.054 N0.99 N0.99 0.66 0.06 0.035 0.006 0.28 0.58 0.001 0.90 N0.99 N0.99 0.46 N0.99 0.63 0.81 0.51 N0.99 N0.99 N0.99 0.28 0.36 N0.99 0.69 0.41 N0.99
Abbreviation: ANC = absolute neutrophil count.
breakpoint groups with the following exceptions: infection source (P = 0.019), type of GN pathogen (P = 0.002), and the presence of confirmed ESBL-producing pathogen (P b 0.001), as shown in Table 1. Likewise,
predictors of inhospital death were relatively similar with the following exceptions: being in the ICU at time of culture (P b 0.001), the number of days to positive culture (P = 0.006), prior immunosuppression history
Study Population ( n = 91 )
Modified APACHE II ≤16.5 [ n = 6/44 (13.6%) ]
Modified APACHE II >16.5 [ n = 13/47 (27.7%) ]
P=0.125
MIC ≤ 2 mg/L [ n = 7/31 (22.6%) ]
MIC ≤ 2 mg/L [ n = 1/24 (4.2%) ] P=0.077
P=0.28 MIC ≥ 4 mg/L [ n = 5/20 (25%) ]
MIC ≥ 4 mg/L [ n = 6/16 (37.5%) ]
Fig. 1. Incidence of inhospital death: classification and regression tree determined clinical MIC breakpoint for cefepime.
168
N.J. Rhodes et al. / Diagnostic Microbiology and Infectious Disease 82 (2015) 165–171
(P = 0.035), and those with a known source of infection (P = 0.001), as shown in Table 1. Clinical outcomes by the CART-derived breakpoint are shown in Table 2. Inhospital death was numerically greater in the cefepime MIC ≥4 mg/L group (30.6% versus 14.6%; P = 0.066). 3.2. Cefepime dosing intensity The dosing intensity of patients is shown in Table 3. Overall, 82.4% of patients received the renal equivalent of 2 g every 8 hours (n = 75), 16.5% of patients received the renal equivalent of 2 g every 12 hours (n = 15), and 1.1% of patients received the renal equivalent of 1 g every 12 hours (n = 1). The incidence of maximal dosing intensity (i.e., the equivalent of 2 g every 8 hours) was similar between CARTderived breakpoint groups (P = 0.71) and between survivors and those who died (P N 0.99). Cefepime dose did not improve multivariate predictions of mortality. Similarly, dose-to-MIC ratios were collinear with the observed cefepime MICs and did not significantly improve the model estimates of mortality (data not shown), thus allowing evaluation the contribution of the more clinically relevant parameter of cefepime MIC without regard to dose in all subsequent models.
In the survival analysis, controlling for log10-normalized days to positive culture and modified APACHE II score on day 0 (model 4), cumulative risk of mortality increased in MIC categories of 4 mg/L (hazard ratio [HR] 3.75, 95% CI 1.18–12.0; P = 0.025) and 64 mg/L (HR 4.61, 95% CI 1.31–16.2; P = 0.017). Similar findings were noted in 30-day mortality analyses (data not shown). In survival analyses using the CART identified MIC categories (i.e. ≤2 versus ≥4 mg/L) (model 5) and controlling for log10 days to positive culture and modified APACHE II score N16.5 on day 0, hospital death increased at MICs ≥4 mg/L (HR 2.54, 95% CI 1.01–6.37; P = 0.048; Figure S1). Relatively similar results were noted for 30-day mortality. Overall, median hospital LOS postculture for survivors was numerically but not statistically shorter for patients with cefepime MICs ≤2 versus ≥4 mg/L (16 days, interquartile range [IQR] 9–38 days versus 24 days, IQR 12–40 days; P = 0.19). Median ICU LOS postculture for survivors was significantly shorter for patients with a cefepime MIC ≤2 versus ≥4 mg/L (2 days, IQR 2–3 days versus 16 days, IQR 4–27 days; P = 0.026). All-cause 30-day readmission rates and antibiotic days postculture did not differ significantly between patients (Table 2). 3.4. Contribution of time to cefepime therapy and other active therapies administered
3.3. Multivariate models of survival according to cefepime MIC The multivariate models for mortality according to cefepime MIC are shown in Tables 4 and 5. Model 1 evaluated the influence of the CARTderived categorical MIC breakpoint and controlled for the CART-derived modified APACHE II breakpoint of N16.5 and days to positive culture (log10 normalized) on the outcome of inhospital mortality. Cefepime MIC ≥4 mg/L independently predicted increased mortality (aOR 3.21, 95% confidence interval [CI] 1.02–10.2; P = 0.047). Model 2 evaluated the influence of each cefepime MIC category (compared to a referent category of an MIC of 1 mg/L) adjusting for modified APACHE II score and days to positive culture (log10 normalized) on the outcome of inhospital mortality. The cefepime MIC categories of 4 mg/L and 64 mg/L independently predicted increased mortality (MIC = 4 mg/L: aOR 6.47, 95% CI 1.25–33.4; P = 0.026, and MIC = 64 mg/L: aOR 6.54 95% CI 1.03–41.4; P = 0.046). The influence of cefepime MIC on the actual and predicted odds of mortality is depicted in Fig. 2. Log-linear transformations of MIC (model 3) did not result in improved mortality predictions (Table S2). There was no significant effect of pathogen type on multivariate mortality estimates (Table S2).
Time to receipt of cefepime and receipt of other active agents were also assessed as potential predictors of mortality. All patients received cefepime therapy within 4 days of the index blood culture, and days to initiation of cefepime therapy did not predict increased mortality (P = 0.49 by log-rank test, Fig. 3). Receipt of other active agents (administered concomitantly or serial fashion) as directed therapies was common as 91% (83/91) of patients within the study population received at least 1 other active agent in addition to or after cefepime including the following classes: fluoroquinolones (27.7%), carbapenems (24.1%), piperacillin-tazobactam (16.9%), alternate cephalosporins (15.7%), aminoglycosides (8.4%), polymyxins (2.4%), tigecycline (2.4%), and trimethoprim-sulfamethoxazole (2.4%). Receipt of other active agents in addition to or after cefepime did not predict increased mortality at the bivariate level (93.1% [n = 67] among survivors versus 100% [n = 19] among those who died; P = 0.60) nor did it improve multivariate models (data not shown). Antecedent carbapenem use was rare with only 3 cases. As such, mortality rates were not significantly different between subjects receiving a carbapenem prior to cefepime therapy and those who did not (0% versus 21.6%; P N 0.99). Antecedent
Table 2 Clinical outcomes by CART-derived breakpoint group and inhospital mortality outcome. MIC ≤2 mg/L
MIC ≥4 mg/L
Parameter
n = 55
n = 36
Died, n (%) ICU transfer postculture, n (%) Days to negative culture, median (IQR) Days to death postculture (n = 19), median (IQR) Hospital LOS postculture for survivors (n = 72), median (IQR) ICU LOS postculture for survivors (n = 22), median (IQR) Readmission within 30 days (n = 72), n (%) Discharge from ICU, n (%) Discharge from general medicine, n (%) Disposition, n (%) Death (n = 19) Discharged home (n = 53) Long-term acute care (n = 1) Skilled nursing facility (n = 4) Inpatient rehabilitation (n = 13) Temporary housing (n = 1) Inpatient days of antibiotics postculture, median (IQR)
8 (14.6) 11 (30.6) 19 (34.6) 13 (36.1) 2.06 (1.47–2.40) n = 52 2.00 (1.63–3.10) n = 33 8.38 (1.82–11.5) 8.21 (1.83–17.8)
0.066 0.88 0.56 0.97
19 (100) 19 (26.4) 13 (68.4) 1.99 (1.46–2.51) n = 71 2.22 (1.83–3.08) n = 14 8.2 (1.8–14.1)
10 (6.22–18.4)
10.3 (7.08–29.0)
0.30
10.2 (6.46–20.9)
2 (2–3)
16 (4–27)
0.026
2.91 (2.00–12.5)
14 (29.8), n = 47 5 (9.09) 50 (90.9)
10 (40), n = 25 10 (27.8) 26 (72.2)
0.38 0.019
24 (33.3), n = 72 1 (1.39) 71 (98.6)
8 (14.6) 38 (69.1) 0 (0) 0 (0) 8 (14.6) 1 (1.82) 9.06 (6.13–15.3)
11 (30.6) 15 (41.7) 1 (2.78) 4 (11.1) 5 (13.9) 0 (0) 10.2 (6.32–20.1)
P-value Survived inhospital n = 72
0.007 0.066 0.009 0.40 0.022 N0.99 N0.99 0.27
Died inhospital
P-value
n = 19
14 (73.7) 5 (26.3)
0.001 0.19
b0.001 b0.001
19 (100) 53 (73.6) 1 (1.39) 4 (5.56) 13 (18.1) 1 (1.39) 9.13 (6.33–17.0)
9.10 (1.29–12.6)
0.11
N.J. Rhodes et al. / Diagnostic Microbiology and Infectious Disease 82 (2015) 165–171
169
Table 3 Cefepime renal dose equivalents received according to product labeling. MIC ≤2 mg/L
MIC ≥4 mg/L
Parameter
n = 55
n = 36
Cefepime dosing 2 g every 8 hours 2 g every 12 hours 1 g every 12 hours 500 mg every 12 hours
46 (83.6) 9 (16.4) 0 (0) 0 (0)
29 (80.6) 6 (16.7) 1 (2.78) 0 (0)
fluoroquinolone use during the admission of the GNBSI was also uncommon with only 13 cases. Mortality rates were also not significantly different between subjects who received a fluoroquinolone prior to cefepime therapy and those who did not (15.4% versus 21.8%; P = 0.73). 3.5. Microbiology of infections The distribution of pathogens according to CART-derived clinical MIC breakpoint and inhospital mortality outcome is shown in Table 1. BSIs due to Acinetobacter spp. and Pseudomonas spp. more frequently had MICs ≥4 mg/L (8.3% versus 0%; P = 0.059 and 52.8% versus 20%; P = 0.001, respectively), while BSIs due to Escherichia coli more frequently had MICs ≤2 mg/L (47.3% versus 16.7%; P = 0.003). Organism category was not significantly associated with death in univariate analyses (Table 1) nor did it significantly improve mortality predictions in multivariate analyses (Table S2). The addition of GN genera in place of categorical cefepime MIC was less explanatory than the categorical treatment of cefepime MIC (AIC = 84.7 versus 98.5) (Table S2). 4. Discussion Elevated cefepime MICs predicted inhospital mortality among patients with GNBSI. The association of higher MICs with worse outcomes is supported by the difference in outcomes observed in the clinical breakpoint model, the categorical MIC model, and the time-to-death model. The near homogenous cefepime dosing administered to our patients allowed exploration of MIC as a mortality predictor, which is important since this is the most accessible parameter available to clinicians and is used by the CLSI to classify susceptible, susceptible dose dependent, and resistant. This study is unique in that 1) most patients received aggressive doses (the renal equivalent of 2 g every 8 hours) of cefepime as per our institutional practice for patients with severe infections; 2) our patients represented a highly comorbid population, with a majority of patients having had neutropenia (61/91, 67%); and 3) a variety of organisms with cefepime MICs between 2 and 8 mg/L were reasonably represented. Our data support a clinical MIC breakpoint of 2 mg/L for various GNBSI, which is similar to the new CLSI recommendation (CLSI, 2014) and the EUCAST breakpoint for Enterobacteriaceae (EUCAST, 2014) but lower than the CLSI and EUCAST breakpoint for nonfermentative organisms like Pseudomonas spp. Our data also suggest caution in treating MICs 4–8 mg/L with cefepime, though additional study is necessary before formal recommendations are possible. Table 4 Multivariate model of the contribution of cefepime MIC ≥4 mg/L to the odds of inhospital mortality. Parameter
Log10 days to positive culture Cefepime MIC ≥4 mg/L Modified APACHE II (day 0) N16.5 Abbreviation: OR = odds ratio.
Cefepime MIC ≥4 mg/L (model 1) Univariate analysis OR (95% CI)
Multivariate analysis OR (95% CI)
4.99 (1.67–14.9) 2.59 (0.92–7.25) 2.42 (0.83–7.08)
5.38 (1.67–17.3) 3.21 (1.02–10.2) 3.06 (0.93–10.1)
P-value
0.63 0.71 0.97 0.40
Survived inhospital
Died inhospital
n = 72
n = 19
59 (81.9) 12 (16.7) 1 (1.39) 0 (0)
16 (84.2) 3 (15.8) 0 (0) 0 (0)
P-value
N0.99 N0.99 N0.99 N0.99
Our results are generally consistent with others studies. Bhat et al. (2007) studied 204 episodes of GNBSIs in patients receiving cefepime 1–2 g intravenously every 12 hours. They found higher 28-day mortality at a cefepime MIC of ≥8 mg/L than at a cefepime MIC of b8 mg/L (54.8% versus 24.1%; P = 0.001). Although immune compromise and neutropenia in our population was higher than Bhat et al. (2007), neutropenia status was evaluated and did not influence our findings. Chopra et al. (2012) evaluated the extent to which empiric or directed use of cefepime or a carbapenem for ESBL-producing GNBSI impacted clinical outcomes. When empiric cefepime use was evaluated in the multivariate model, there was no association with increased mortality. Lee et al. (2013) studied patients treated with cefepime for ESBL-producing GNBSI in Taiwan. The authors observed increased odds of 30-day mortality and a lower survival rate when cefepime monotherapy was utilized. In our limited dataset, we did not observe a significant association of ESBL status, drug dose, or receipt of combination therapy on inpatient mortality in univariate or multivariate analyses. Limitations must be considered. First, this was a single-center, retrospective cohort using a sample size of convenience and is subject to inherent biases. Second, though pharmacokinetic and clinical studies have shown higher likelihood of target attainment (Kuti et al., 2003; Nicasio et al., 2009) and improved mortality (Lodise et al., 2007) by utilizing more frequent or prolonged infusions, the majority of patients in this study received maximal standard infusions, and only 2 (2.2%) received continuous infusion, both of whom survived. Though one may criticize an MIC-only analysis, PK data (serum drug monitoring) were not available within our cohort, and thus, a PK/PD analysis could not be undertaken. However, as MIC values represent the most commonly available clinical indicators of resistance upon which to base drug selection, the impact of this variable on mortality was evaluated. Third, many
Table 5 Multivariate model of the contribution of increasing categorical cefepime MIC relative to a referent category of 1 mg/L to the odds of inhospital mortality. Parameter
Cefepime MICc 1 mg/L 2 mg/L 4 mg/L 8 mg/L 16 mg/L 32 mg/L 64 mg/L Modified APACHE II (day 0) Log10 days to positive culture a
Cefepime MIC as a categorical variable (model 2) Univariate analysis OR (95% CI)a
Multivariate analysis OR (95% CI)b
0.58 (0.20–1.65) 0.31 (0.04–2.55) 4.79 (1.22–18.8) 0.82 (0.16–4.17)
0.79 (0.08–7.88) 6.47 (1.23–33.4) 1.58 (0.25–10.0)
d
d
b
d
d
6.13 (1.24–30.3) 1.04 (0.94–1.16) 4.99 (1.67–14.9)
6.54 (1.03–41.4) 1.07 (0.94–1.23) 4.85 (1.41–16.7)
Univariate referent category was all other MIC categories for each MIC. Multivariate referent category was MIC = 1 mg/L. To test the importance of the inclusion of cefepime MIC in the multivariate model of mortality, we conducted a likelihood ratio test of the final model 2 to a model without categorical cefepime MIC (referent category of 1 mg/L). We found that model 2 with categorical cefepime MIC was significantly more informative than a model without categorical cefepime MIC (AIC [9 df] = 85 versus AIC [3 df] = 86.5; P = 0.036). d Parameter estimates were poor at MICs of 16 and 32 mg/L, and subjects with MICs in these categories were excluded from the model as the outcome did not vary in these categories. b c
N.J. Rhodes et al. / Diagnostic Microbiology and Infectious Disease 82 (2015) 165–171
0.2
0.4
Death
0.6
0.8
Yes
No
0
Probability of Death
1.0
170
1
2
4
8
16
32
64
Cefepime MIC (mg/L) Mean Predicted Death (%)
Individual Predicted Death (%)
95% CI Probability of Death (%)
Observed Death (Yes/No)
Fig. 2. Adjusted probabilities of death⁎ and observed inhospital mortality according to cefepime MIC category.⁎⁎. ⁎Probabilities of death adjusted for regression covariates as shown in Table 3, model 2. Each MIC category was regressed against the outcome of death while holding modified APACHE II score and the log10-normalized LOS prior to culture at their mean values (16.9 and 5.6 days, respectively). ⁎⁎Individual predicted death and observed death are jittered for visual appreciation of the count at each MIC.
0.75 0.25
0.50
P=0.49
0.00
Proportion on Cefepime
1.00
patients in our study were neutropenic, which qualifies for maximal dosing of cefepime at our institution. Results may be less applicable to healthier patients. Fourth, we did not observe significant increases in mortality in the cefepime MIC categories of 8, 16, or 32 mg/L; however, sample size was constrained within these MIC categories, and multivariate predictions were poor at MICs of 16 and 32 mg/L. Despite the less robust fit, the log-linear assessment was consistent with our categorical MIC assessment, supporting higher mortality with higher cefepime MICs. Additionally, sample size constraints limited our ability to identify whether specific GN genera were more closely associated with in hospital mortality. Fifth, the receipt of other active agents in addition to or after cefepime administration represented a confounding variable. However, the evaluation of this effect in the multivariate analysis mitigated the impact of this limitation for assessing the impact of cefepime MIC on mortality outcomes. Sixth, MICs were obtained using automated susceptibility testing by the Vitek 2 system as opposed to agar dilution or broth microdilution; however, Vitek 2 is a Food and Drug Administration–approved method and frequently used for clinical decision making and was the only source of MICs available during the study period. We have shown that there is clinical evidence to support lower cefepime susceptibility breakpoints (≤2 mg/L) for various GNBSI when using Vitek 2 for susceptibility testing. Although the impact of increasing MIC requires confirmation in larger and more robust studies, we recognize that quantifying outcomes for patients with more resistant organisms and for
0
.5
1
1.5
2
2.5
3
3.5
4
Time in Days Died = No
Died = Yes
Fig. 3. Time to cefepime initiation stratified by outcome of inhospital mortality.
those treated with cefepime monotherapy is challenging. Until a larger study is conducted with a greater representation of MICs between 8 and 32 mg/L and a more diverse representation or organisms at each MIC, we believe that it is prudent to consider GNBSIs in highly comorbid patients with an MIC ≥4 mg/L as potentially clinically resistant. Acknowledgments We would like to acknowledge the following individuals who supported the development of this manuscript and provided reviews of the data presented herein: Sonia Rao, PharmD, and Sheila Wang, PharmD. We thank Dr. Steven J.H. Samuels at the University at Albany, State University of New York, and Dr. Nicholas J. Cox at Durham University, United Kingdom, for their kind input regarding the Stata code for CIs of marginal effects. Funding No external funding was received. Transparency declaration All authors: No conflicts of interest. Appendix A. Supplementary data Supplementary data to this article can be found online at http://dx. doi.org/10.1016/j.diagmicrobio.2015.03.005. References Bellomo R, Ronco C, Kellum JA, Mehta RL, Palevsky P, Acute Dialysis Quality Initiative, w. Acute renal failure—definition, outcome measures, animal models, fluid therapy and information technology needs: the Second International Consensus Conference of the Acute Dialysis Quality Initiative (ADQI) Group. Crit Care 2004;8:R204–12. Bhat SV, Peleg AY, Lodise TP, Shutt KA, Capitano B, Potoski BA, et al. Failure of current cefepime breakpoints to predict clinical outcomes of bacteremia caused by gramnegative organisms. Antimicrob Agents Chemother 2007;51:4390–5. Chopra T, Marchaim D, Veltman J, Johnson P, Zhao JJ, Tansek R, et al. Impact of cefepime therapy on mortality among patients with bloodstream infections caused by extended-spectrum-beta-lactamase-producing Klebsiella pneumoniae and Escherichia coli. Antimicrob Agents Chemother 2012;56:3936–42. Clinical and Laboratory Standards Institute. Performance standards for antimicrobial susceptibility testing. M100-S20; 2010 [Wayne, PA]. Clinical and Laboratory Standards Institute (CLSI). Performance standards for antimicrobial susceptibility testing. M100-S23; 2013 [Wayne, PA]. Clinical and Laboratory Standards Institute (CLSI). Performance standards for antimicrobial susceptibility testing. M100-S24; 2014 [Wayne, PA]. Esterly JS, Wagner J, McLaughlin MM, Postelnick MJ, Qi C, Scheetz MH. Evaluation of clinical outcomes in patients with bloodstream infections due to Gram-negative bacteria according to carbapenem MIC stratification. Antimicrob Agents Chemother 2012;56: 4885–90. Hamilton KW, Bilker WB, Lautenbach E. Controlling for severity of illness in assessment of the association between antimicrobial-resistant infection and mortality: impact of calculation of Acute Physiology and Chronic Health Evaluation (APACHE) II scores at different time points. Infect Control Hosp Epidemiol 2007;28:832–6. Hosmer DW, Lemeshow S. Applied logistic regression. 2nd ed. New York: Wiley; 2000. Hospira. MAXIPIME (cefepime hydrochloride) IV package insert. Lake Forest, IL: Hospira, Inc.; 2013. Kuti JL, Dandekar PK, Nightingale CH, Nicolau DP. Use of Monte Carlo simulation to design an optimized pharmacodynamic dosing strategy for meropenem. J Clin Pharmacol 2003;43:1116–23. Kuti JL, Nightingale CH, Nicolau DP. Optimizing pharmacodynamic target attainment using the MYSTIC antibiogram: data collected in North America in 2002. Antimicrob Agents Chemother 2004;48:2464–70. Lee SY, Kuti JL, Nicolau DP. Cefepime pharmacodynamics in patients with extended spectrum beta-lactamase (ESBL) and non-ESBL infections. J Infect 2007;54:463–8. Lee NY, Lee CC, Huang WH, Tsui KC, Hsueh PR, Ko WC. Cefepime therapy for monomicrobial bacteremia caused by cefepime-susceptible extended-spectrum beta-lactamase-producing Enterobacteriaceae: MIC matters. Clin Infect Dis 2013;56: 488–95. Lodise Jr TP, Lomaestro B, Drusano GL. Piperacillin-tazobactam for Pseudomonas aeruginosa infection: clinical implications of an extended-infusion dosing strategy. Clin Infect Dis 2007;44:357–63. Nicasio AM, Ariano RE, Zelenitsky SA, Kim A, Crandon JL, Kuti JL, et al. Population pharmacokinetics of high-dose, prolonged-infusion cefepime in adult critically ill patients with ventilator-associated pneumonia. Antimicrob Agents Chemother 2009;53: 1476–81. Song W, Moland ES, Hanson ND, Lewis JS, Jorgensen JH, Thomson KS. Failure of cefepime therapy in treatment of Klebsiella pneumoniae bacteremia. J Clin Microbiol 2005;43: 4891–4.
N.J. Rhodes et al. / Diagnostic Microbiology and Infectious Disease 82 (2015) 165–171 The European Committee on Antimicrobial Susceptibility Testing (EUCAST) 2014. Breakpoint tables for interpretation of MICs and zone diameters. Version 4.0, vol. 2014. ; 2014. Thom KA, Shardell MD, Osih RB, Schweizer ML, Furuno JP, Perencevich EN, et al. Controlling for severity of illness in outcome studies involving infectious diseases: impact of measurement at different time points. Infect Control Hosp Epidemiol 2008;29:1048–53. Tumbarello M, Sanguinetti M, Montuori E, Trecarichi EM, Posteraro B, Fiori B, et al. Predictors of mortality in patients with bloodstream infections caused by extended-
171
spectrum-beta-lactamase-producing Enterobacteriaceae: importance of inadequate initial antimicrobial treatment. Antimicrob Agents Chemother 2007;51:1987–94. Ury HK. Efficiency of case-control studies with multiple controls per case: continuous or dichotomous data. Biometrics 1975;31:643–9. Wi YM, Kang CI, Cheong HS, Chung DR, Jung CW, Ko KS, et al. Failure of cefepime therapy in neutropenic patients with extended-spectrum beta-lactamase-producing Gramnegative bacteraemia. Int J Antimicrob Agents 2009;33:384–6.
Contents lists available at ScienceDirect
Diagnostic Microbiology and Infectious Disease journal homepage: www.elsevier.com/locate/diagmicrobio
Evaluation of clinical outcomes in patients with Gram-negative bloodstream infections according to cefepime MIC☆ Nathaniel J. Rhodes a,b, Jiajun Liu c, Milena M. McLaughlin a,b, Chao Qi d, Marc H. Scheetz a,b,⁎ a
Department of Pharmacy Practice, Midwestern University, Chicago College of Pharmacy, Downers Grove, IL, USA Department of Pharmacy, Northwestern Memorial Hospital, Chicago, IL, USA Midwestern University, Chicago College of Pharmacy, Downers Grove, IL, USA d Department of Pathology, Clinical Microbiology Division, Northwestern Memorial Hospital, Chicago, IL, USA b c
a r t i c l e
i n f o
Article history: Received 29 July 2014 Received in revised form 26 February 2015 Accepted 3 March 2015 Available online 10 March 2015 Keywords: Cefepime Gram-negative bacteria Bloodstream infection Minimum inhibitory concentration Mortality
a b s t r a c t Predicted and observed failures at higher cefepime MICs have prompted the Clinical and Laboratories Standards Institute (CLSI) to lower the susceptible breakpoint for Enterobacteriaceae to ≤2 mg/L, with dose-dependent susceptibility at 4–8 mg/L, while the susceptibility breakpoint for nonfermentative organisms remain unchanged at ≥8 mg/L. The contribution of increasing cefepime MIC to mortality risk in the setting of aggressive cefepime dosing is not well defined. Patients who were treated with cefepime for Gram-negative blood stream infections (GNBSIs), including both Enterobacteriaceae and nonfermentative organisms, were screened for inclusion in this retrospective cohort study. Demographic and microbiologic variables were collected, including pathogen, cefepime MIC, dosage, and interval. The objective was to define a risk-adjusted mortality breakpoint for cefepime MICs. Secondarily, we looked at time to death and length of stay (LOS) postculture. Ninety-one patients were included in the analysis. Overall, 19 patients died and 72 survived. Classification and Regression Tree analysis identified an inhospital mortality breakpoint at a cefepime MIC between 2 and 4 mg/L for patients with a modified Acute Physiology and Chronic Health Evaluation II score ≤16.5 (4.2% versus 25%, respectively). Multivariate logistic regression revealed increased odds of mortality at a cefepime MIC of 4 mg/L (adjusted odds ratio [aOR] 6.47; 95% confidence interval [CI] 1.25–33.4) and 64 mg/L (aOR 6.54, 95% CI 1.03-41.4). Those with cefepime MICs ≥4 mg/L experienced a greater median intensive care unit LOS for survivors (16 versus 2 days; P = 0.026). Increasing cefepime MIC appears to predict inhospital mortality among patients who received aggressive doses of cefepime for GNBSIs, supporting a clinical breakpoint MIC of 2 mg/L. © 2015 Elsevier Inc. All rights reserved.
1. Introduction In 2010, the Clinical and Laboratories Standards Institute (CLSI) lowered the cephalosporin MIC breakpoints for the Enterobacteriaceae, except for cefepime, a broad-spectrum fourth-generation cephalosporin (CLSI, 2010). The CLSI retained the cefepime resistance breakpoint of 8 mg/L for Enterobacteriaceae and nonfermentative organisms based on data suggesting that adequate doses of cefepime should provide a high likelihood of pharmacokinetic-pharmacodynamic (PK/PD) target attainment (CLSI, 2010; Lee et al., 2007; Nicasio et al., 2009). In contrast, the European Committee on Antimicrobial Susceptibility Testing (EUCAST) susceptibility breakpoint is ≤1 mg/L, while the resistance breakpoint has been ≥4 mg/L for Enterobacteriaceae since 2010 (EUCAST, 2014). Similar to CLSI guidelines, the EUCAST cefepime breakpoint for Pseudomonas spp. remains ≤8 mg/L.
☆ Portions of this paper were presented as a platform at the 53rd Interscience Conference on Antimicrobial Agents and Chemotherapy, Denver, CO, September 10–13, 2013. ⁎ Corresponding author. Tel.: +1-630-515-6116; fax: +1-630-515-6958. E-mail address: [email protected] (M.H. Scheetz). http://dx.doi.org/10.1016/j.diagmicrobio.2015.03.005 0732-8893/© 2015 Elsevier Inc. All rights reserved.
Early reports identified clinical failures at cefepime MICs in the range of 2–16 mg/L (Song et al., 2005; Wi et al., 2009), and an initial clinical breakpoint (defined by a significantly elevated risk of inhospital mortality) of 8 mg/L for Gram-negative (GN) bloodstream infections (BSIs) was determined (Bhat et al., 2007). However, most patients received the dose equivalent of 1–2 g every 12 hours, which is less than the PK/ PD optimized dosing (Bhat et al., 2007; Kuti et al., 2004; Lee et al., 2007; Nicasio et al., 2009). Others have observed worse outcomes with increased cefepime MICs (8–16 mg/L) in the setting of extendedspectrum beta-lactamase (ESBL)–positive GNBSI, but these investigations did not evaluate the impact of cefepime dosing or the effect of incremental increases in cefepime MIC on clinical outcomes (Chopra et al., 2012; Lee et al., 2013; Tumbarello et al., 2007). Given the clinical data, findings from PK/PD modeling and simulation, and the common use of lower cefepime doses, the 2014 update to the CLSI guidelines have included a reduced susceptibility breakpoint for cefepime at ≤2 mg/L with dose-dependent susceptibility defined between 4 and 8 mg/L and resistant defined at ≥16 mg/L for Enterobacteriaceae (CLSI, 2014). Despite the aforementioned studies, clinical outcomes studies of GNBSIs with cefepime MICs between 2 and 8 mg/L are lacking. We sought to identify a clinical mortality breakpoint among patients treated
166
N.J. Rhodes et al. / Diagnostic Microbiology and Infectious Disease 82 (2015) 165–171
with aggressively dosed cefepime for various GNBSI. Secondarily, we sought to examine the clinical breakpoint for other outcomes such as time to death and hospital and intensive care unit (ICU) length of stay (LOS) postculture.
isolates using Vitek 2 cards (bioMérieux, Balmes-les-Grottes, France). Organism susceptibility was interpreted for clinicians according to CLSI guidelines at the time of culture (CLSI, 2013). In cases of mixed GNBSI, the most resistant organism (highest cefepime MIC) was used for analyses.
2. Materials and methods 2.4. Statistical analysis 2.1. Study design This retrospective, cohort study was conducted at Northwestern Memorial Hospital in Chicago, Illinois. Study methods were reviewed and approved by institutional review boards at Northwestern University and Midwestern University. Patients with at least 1 positive GNBSI treated with cefepime between September 1, 2006, and August 31, 2012, were reviewed for inclusion. Patients were included only once, and the index culture was the first GNBSI. Patients were randomly selected to fill MIC categories. A maximum of 4 bloodstream isolates with a cefepime MIC ≤1 mg/L was included for every isolate with a cefepime MIC ≥2 mg/L to balance power (Ury, 1975). 2.2. Definitions ICU-onset infection was defined as patient ICU status at the time of the GNBSI. Concurrent renal dysfunction was defined as acute or chronic renal dysfunction. Acute renal dysfunction was an increase in serum creatinine of 0.5 mg/dL or 50% from baseline to immediately before the infection (Bellomo et al., 2004). Chronic renal dysfunction was defined as a diagnosis of chronic kidney dysfunction. Hepatic dysfunction was defined as any liver enzyme greater than 3 times the upper normal limit at the time of culture, chronic hepatitis, or documented cirrhosis. Prior and concurrent immunosuppression was defined as steroid, chemotherapy, or immunosuppressant use within the prior 12 months. Patient comorbidities were considered to be present if they were documented in the admission history and physical. The source of the BSI was considered from the attending physician's written diagnosis. Other active therapy in addition to or after cefepime receipt was considered for relevant GN drugs. Cefepime dose intensity was defined according to the recommendations in the approved product labeling as: 2 g every 8 hours, 2 g every 12 hours, 1 g every 12 hours, 500 mg every 12 hours, or the renal dysfunction–adjusted equivalent doses (Hospira, 2013). GN organisms isolated from blood cultures were treated as categorical variables according to genera and lactose fermentative phenotype. 2.3. Data elements Variables were collected from medical, pharmacy, and microbiology records by trained reviewers. Variables collected included demographics (age, gender, and race), ICU versus non-ICU admission, modified Acute Physiology and Chronic Health Evaluation II (APACHE II) component and composite score (Hamilton et al., 2007; Thom et al., 2008), renal dysfunction, hepatic dysfunction, solid organ or hematologic transplant, diabetes, concurrent immunosuppression, hospitalizations within 12 months before culture, hospital LOS postculture, polymicrobial bacteremia (excluding coagulase-negative staphylococci), infecting organism, infection source, the name and dose of all GN antibiotics received, and organism cefepime MIC. Outcomes included inhospital and 30-day all-cause mortality, time to death, days to negative culture, ICU transfer, hospital and ICU LOS for survivors, 30-day readmission, inpatient antimicrobial duration, and disposition postadmission. Mortality at 30 days was assessed using the Social Security Death Master file. Empiric and directed cefepime was dosed prospectively and adjusted by clinical pharmacists according to internal protocols. The impact of dosing was assessed as the ratio of the 24-hour renal dose equivalent divided by the pathogen MIC (dose-to-MIC ratio). GN organisms were identified using the Vitek 2 system (bioMérieux, Balmes-les-Grottes, France). Susceptibility testing was performed on all
The primary outcome (i.e., dependent variable) was the covariateadjusted odds of inhospital mortality. Cefepime MIC was assessed as the primary predictor; however, cefepime dose and dose-to-MIC ratio were also assessed as substitutes for MIC (individually, due to collinearity). Cefepime MIC was handled as an interval variable (log-linear), a categorical variable (at each doubling MIC dilution category from 1 to 64 mg/L), and binned as a dichotomous variable via classification and regression tree (CART). CART was performed using SPSS version 19 and the Decision Trees add-on version 19 (SPSS, Chicago, IL, USA). All relevant covariates (Table S1) were considered for model inclusion. All other analyses were performed using Intercooled Stata version 13.0 (Statacorp, College Station, TX, USA). Secondary outcomes (i.e., additional dependent variables) were time to death and LOS postculture. Continuous variables were evaluated with Student's t test or Wilcoxon rank sum test, as appropriate. Categorical variables were evaluated with the chi-square or Fisher's exact test, as appropriate. Nonnormal variables were normalized using log transformation to permit parametric assessments. Multivariate analyses involved the use of logistic regression, which was performed in a stepwise fashion. Candidate covariates (Table 1) at a bivariate level of significance of P b 0.2 were further assessed as possible independent predictors of the primary outcome (Esterly et al., 2012). Variables were retained in models if the objective function value changed by N3.84 with each iterative addition (n + 1 model). Final models were secondarily checked for optimal parsimony utilizing the Aikaike Information Criterion (AIC) (Hosmer and Lemeshow, 2000). Outcome probabilities, adjusting for comodeled covariates, were calculated from adjusted odds ratios (aORs). Since cefepime MIC was the primary variable of interest, it was the major variable considered for CART splitting; however, as previous studies demonstrated the importance of morbidity indices for CART analyses (Lodise et al., 2007), a tree with 2 splits was created with APACHE II score at the first node and cefepime MIC at the second node. We assessed competing logistic regression models for mortality with cefepime MIC as a predictor using: the CART-derived cefepime MIC breakpoint (model 1), categorical cefepime MIC (model 2), and log2-transformed cefepime MIC (model 3). Modified APACHE II was forced into all models, as it is a known predictor of mortality (Hamilton et al., 2007; Thom et al., 2008). As additional exploratory analyses, time-to-event analyses were conducted by assessing variables identified in multivariate logistic regression models using Cox regression models for the categorical MIC (model 4) and CART-derived breakpoint MIC (model 5). Kaplan–Meier analyses with log-rank tests were conducted for bivariate significance determination and graph generation for all time-to-event analyses. All tests were 2 tailed, with an a priori level of alpha set at 0.05 for statistical significance. 3. Results 3.1. Baseline characteristics A total of 223 patients were screened for inclusion, and 91 patients met inclusion criteria during the study period. CART defined a clinical MIC inhospital mortality breakpoint when controlling for modified APACHE II score on day 0. Splits were identified at a modified APACHE II score of ≤16.5 and N16.5 and at a cefepime MIC of ≤2 mg/L and ≥4 mg/L (Fig. 1). Patients were then stratified according to MIC breakpoint category (≤2 mg/L or ≥4 mg/L) with baseline demographics displayed in Table 1. Predictors were relatively similar between
N.J. Rhodes et al. / Diagnostic Microbiology and Infectious Disease 82 (2015) 165–171
167
Table 1 Baseline characteristics by CART-derived breakpoint group and inhospital mortality outcome. Cefepime MIC ≤2 mg/L
Cefepime MIC ≥4 mg/L
Parameter
n = 55
n = 36
Age in years, mean (SD) Male, n (%) Race White, n (%) Black, n (%) Asian, n (%) Hispanic, n (%) Other, n (%) Modified APACHE II on day of culture, mean (SD) Admit from Emergency department, n (%) Outside hospital, n (%) Outpatient clinic, n (%) Community, n (%) Serum creatinine on day of culture, median (IQR) ICU at culture, n (%) ANC b500 cells/mm3 during admission, n (%) Mechanical ventilation in prior 12 months, n (%) History of leukemia, n (%) History of lymphoma, n (%) History of myeloma, n (%) Renal dysfunction within 5 days of culture, n (%) Diabetes, n (%) Receipt of immune suppressants within 12 months, n (%) Days to positive culture, median (IQR) Days to cefepime therapy, median (IQR) Received other active therapy in addition to cefepime, n (%) Source known, n (%) Source, n (%) Urine (n = 10) Respiratory (n = 1) Catheter tip (n = 7) GI or skin (n = 11) Central line (n = 21) Gram-negative genera, n (%) Acinetobacter baumannii (n = 3) Acinetobacter spp. (n = 1) Citrobacter spp. (n = 4) Delftia spp. (n = 1) Enterobacter spp. (n = 5) Escherichia coli (n = 32) Klebsiella spp. (n = 15) Pseudomonas spp. (n = 30) Polymicrobial infection, n (%) ESBL positive, n (%)
62.1(14.0) 26 (47.3)
59.9 (13.9) 24(66.7)
40 (72.7) 7 (12.7) 0 (0) 4 (7.2) 4 (7.2) 17.3 (4.32)
18 (50) 11 (30.5) 2 (5.6) 3 (8.3) 2 (5.6) 16.1 (5.12)
24 (43.6) 6 (10.9) 8 (14.6) 17 (30.9) 0.88 (0.7–1.5) 12 (21.8) 41 (74.6) 9 (16.4) 11 (20) 8 (14.6) 8 (14.6) 22 (40) 14 (25.5) 44 (80) 3.7 (1.6–14.6) 0.09 (0.03–0.44) 54 (98.2) 33 (60)
20 (55.6) 9 (25) 3 (8.3) 4 (11.1) 1.02 (0.77–1.33) 13 (36.1) 20 (55.6) 12 (33.3) 8 (22.2) 7 (19.4) 3 (8.3) 17 (47.2) 6 (16.7) 25 (69.4) 7.2 (1.9–18.3) 0.24 (0–0.97) 32 (88.9) 17 (47.2)
7 (21.2) 0 (0) 4 (12.1) 11 (33.3) 11 (33.3)
3 (17.7) 1 (5.88) 3 (17.7) 0 (0) 10 (58.8)
0 (0) 1 (1.82) 3 (5.45) 1 (1.82) 3 (5.45) 26 (47.3) 10 (18.2) 11 (20) 18 (32.7) 3 (5.45)
3 (8.33) 0 (0) 1 (2.78) 0 (0) 2 (5.56) 6 (16.7) 5 (13.9) 19 (52.8) 13 (36.1) 13 (36.1)
P-value
0.45 0.07 0.06 0.03 0.04 0.15 N0.99 N0.99 0.22 0.052 0.27 0.08 0.52 0.04 0.44 0.14 0.06 0.07 0.80 0.54 0.52 0.50 0.32 0.25 0.35 0.64 0.08 0.23 0.019 N0.99 0.34 0.67 0.009 0.084 0.002 0.06 N0.99 N0.99 N0.99 N0.99 0.003 0.59 0.001 0.74 b0.001
Survived inhospital
Died inhospital
n = 72
n = 19
61 (14.7) 39 (54.2)
62.1 (10.6) 11 (57.9)
48 (66.7) 13 (18.1) 2 (2.8) 6 (8.3) 3 (4.2) 16.6 (4.8)
10 (52.6) 5 (26.3) 0 (0) 1 (5.3) 3 (15.8) 17.5 (4.2)
34 (47.2) 10 (13.9) 10 (13.9) 18 (25) 0.90 (0.77–1.30) 12 (16.7) 46 (63.9) 19 (26.8) 12 (16.7) 12 (16.7) 9 (12.5) 30 (41.7) 19 (26.4) 51 (70.8) 2.5 (1.7–12.5) 0.19 (0.02–0.76) 67 (93.1) 46 (63.9)
10 (52.6) 5 (26.3) 1 (5.3) 3 (15.8) 1.15 (0.57–1.70) 13 (68.4) 15 (79.0) 2 (10.5) 7 (36.8) 3 (15.8) 2 (10.5) 9 (47.4) 1 (5.26) 18 (94.7) 15.7 (6.5–28.8) 0.04 (0–0.44) 19 (100) 4 (21.1)
9 (19.6) 1 (2.17) 6 (13.0) 10 (21.7) 20 (43.5)
1 (25) 0 (0) 1 (25) 1 (25) 1 (25)
2 (2.78) 1 (1.39) 3 (4.17) 1 (1.39) 3 (4.17) 27 (37.5) 12 (16.7) 23 (31.9) 23 (31.9) 13 (18.1)
1 (5.26) 0 (0) 1 (5.26) 0 (0) 2 (10.53) 5 (26.3) 3 (15.8) 7 (36.8) 8 (42.1) 3 (15.8)
P-value
0.77 0.77 0.31 0.26 0.42 N0.99 N0.99 0.10 0.45 0.42 0.68 0.19 0.45 0.55 0.82 b0.001 0.28 0.22 0.054 N0.99 N0.99 0.66 0.06 0.035 0.006 0.28 0.58 0.001 0.90 N0.99 N0.99 0.46 N0.99 0.63 0.81 0.51 N0.99 N0.99 N0.99 0.28 0.36 N0.99 0.69 0.41 N0.99
Abbreviation: ANC = absolute neutrophil count.
breakpoint groups with the following exceptions: infection source (P = 0.019), type of GN pathogen (P = 0.002), and the presence of confirmed ESBL-producing pathogen (P b 0.001), as shown in Table 1. Likewise,
predictors of inhospital death were relatively similar with the following exceptions: being in the ICU at time of culture (P b 0.001), the number of days to positive culture (P = 0.006), prior immunosuppression history
Study Population ( n = 91 )
Modified APACHE II ≤16.5 [ n = 6/44 (13.6%) ]
Modified APACHE II >16.5 [ n = 13/47 (27.7%) ]
P=0.125
MIC ≤ 2 mg/L [ n = 7/31 (22.6%) ]
MIC ≤ 2 mg/L [ n = 1/24 (4.2%) ] P=0.077
P=0.28 MIC ≥ 4 mg/L [ n = 5/20 (25%) ]
MIC ≥ 4 mg/L [ n = 6/16 (37.5%) ]
Fig. 1. Incidence of inhospital death: classification and regression tree determined clinical MIC breakpoint for cefepime.
168
N.J. Rhodes et al. / Diagnostic Microbiology and Infectious Disease 82 (2015) 165–171
(P = 0.035), and those with a known source of infection (P = 0.001), as shown in Table 1. Clinical outcomes by the CART-derived breakpoint are shown in Table 2. Inhospital death was numerically greater in the cefepime MIC ≥4 mg/L group (30.6% versus 14.6%; P = 0.066). 3.2. Cefepime dosing intensity The dosing intensity of patients is shown in Table 3. Overall, 82.4% of patients received the renal equivalent of 2 g every 8 hours (n = 75), 16.5% of patients received the renal equivalent of 2 g every 12 hours (n = 15), and 1.1% of patients received the renal equivalent of 1 g every 12 hours (n = 1). The incidence of maximal dosing intensity (i.e., the equivalent of 2 g every 8 hours) was similar between CARTderived breakpoint groups (P = 0.71) and between survivors and those who died (P N 0.99). Cefepime dose did not improve multivariate predictions of mortality. Similarly, dose-to-MIC ratios were collinear with the observed cefepime MICs and did not significantly improve the model estimates of mortality (data not shown), thus allowing evaluation the contribution of the more clinically relevant parameter of cefepime MIC without regard to dose in all subsequent models.
In the survival analysis, controlling for log10-normalized days to positive culture and modified APACHE II score on day 0 (model 4), cumulative risk of mortality increased in MIC categories of 4 mg/L (hazard ratio [HR] 3.75, 95% CI 1.18–12.0; P = 0.025) and 64 mg/L (HR 4.61, 95% CI 1.31–16.2; P = 0.017). Similar findings were noted in 30-day mortality analyses (data not shown). In survival analyses using the CART identified MIC categories (i.e. ≤2 versus ≥4 mg/L) (model 5) and controlling for log10 days to positive culture and modified APACHE II score N16.5 on day 0, hospital death increased at MICs ≥4 mg/L (HR 2.54, 95% CI 1.01–6.37; P = 0.048; Figure S1). Relatively similar results were noted for 30-day mortality. Overall, median hospital LOS postculture for survivors was numerically but not statistically shorter for patients with cefepime MICs ≤2 versus ≥4 mg/L (16 days, interquartile range [IQR] 9–38 days versus 24 days, IQR 12–40 days; P = 0.19). Median ICU LOS postculture for survivors was significantly shorter for patients with a cefepime MIC ≤2 versus ≥4 mg/L (2 days, IQR 2–3 days versus 16 days, IQR 4–27 days; P = 0.026). All-cause 30-day readmission rates and antibiotic days postculture did not differ significantly between patients (Table 2). 3.4. Contribution of time to cefepime therapy and other active therapies administered
3.3. Multivariate models of survival according to cefepime MIC The multivariate models for mortality according to cefepime MIC are shown in Tables 4 and 5. Model 1 evaluated the influence of the CARTderived categorical MIC breakpoint and controlled for the CART-derived modified APACHE II breakpoint of N16.5 and days to positive culture (log10 normalized) on the outcome of inhospital mortality. Cefepime MIC ≥4 mg/L independently predicted increased mortality (aOR 3.21, 95% confidence interval [CI] 1.02–10.2; P = 0.047). Model 2 evaluated the influence of each cefepime MIC category (compared to a referent category of an MIC of 1 mg/L) adjusting for modified APACHE II score and days to positive culture (log10 normalized) on the outcome of inhospital mortality. The cefepime MIC categories of 4 mg/L and 64 mg/L independently predicted increased mortality (MIC = 4 mg/L: aOR 6.47, 95% CI 1.25–33.4; P = 0.026, and MIC = 64 mg/L: aOR 6.54 95% CI 1.03–41.4; P = 0.046). The influence of cefepime MIC on the actual and predicted odds of mortality is depicted in Fig. 2. Log-linear transformations of MIC (model 3) did not result in improved mortality predictions (Table S2). There was no significant effect of pathogen type on multivariate mortality estimates (Table S2).
Time to receipt of cefepime and receipt of other active agents were also assessed as potential predictors of mortality. All patients received cefepime therapy within 4 days of the index blood culture, and days to initiation of cefepime therapy did not predict increased mortality (P = 0.49 by log-rank test, Fig. 3). Receipt of other active agents (administered concomitantly or serial fashion) as directed therapies was common as 91% (83/91) of patients within the study population received at least 1 other active agent in addition to or after cefepime including the following classes: fluoroquinolones (27.7%), carbapenems (24.1%), piperacillin-tazobactam (16.9%), alternate cephalosporins (15.7%), aminoglycosides (8.4%), polymyxins (2.4%), tigecycline (2.4%), and trimethoprim-sulfamethoxazole (2.4%). Receipt of other active agents in addition to or after cefepime did not predict increased mortality at the bivariate level (93.1% [n = 67] among survivors versus 100% [n = 19] among those who died; P = 0.60) nor did it improve multivariate models (data not shown). Antecedent carbapenem use was rare with only 3 cases. As such, mortality rates were not significantly different between subjects receiving a carbapenem prior to cefepime therapy and those who did not (0% versus 21.6%; P N 0.99). Antecedent
Table 2 Clinical outcomes by CART-derived breakpoint group and inhospital mortality outcome. MIC ≤2 mg/L
MIC ≥4 mg/L
Parameter
n = 55
n = 36
Died, n (%) ICU transfer postculture, n (%) Days to negative culture, median (IQR) Days to death postculture (n = 19), median (IQR) Hospital LOS postculture for survivors (n = 72), median (IQR) ICU LOS postculture for survivors (n = 22), median (IQR) Readmission within 30 days (n = 72), n (%) Discharge from ICU, n (%) Discharge from general medicine, n (%) Disposition, n (%) Death (n = 19) Discharged home (n = 53) Long-term acute care (n = 1) Skilled nursing facility (n = 4) Inpatient rehabilitation (n = 13) Temporary housing (n = 1) Inpatient days of antibiotics postculture, median (IQR)
8 (14.6) 11 (30.6) 19 (34.6) 13 (36.1) 2.06 (1.47–2.40) n = 52 2.00 (1.63–3.10) n = 33 8.38 (1.82–11.5) 8.21 (1.83–17.8)
0.066 0.88 0.56 0.97
19 (100) 19 (26.4) 13 (68.4) 1.99 (1.46–2.51) n = 71 2.22 (1.83–3.08) n = 14 8.2 (1.8–14.1)
10 (6.22–18.4)
10.3 (7.08–29.0)
0.30
10.2 (6.46–20.9)
2 (2–3)
16 (4–27)
0.026
2.91 (2.00–12.5)
14 (29.8), n = 47 5 (9.09) 50 (90.9)
10 (40), n = 25 10 (27.8) 26 (72.2)
0.38 0.019
24 (33.3), n = 72 1 (1.39) 71 (98.6)
8 (14.6) 38 (69.1) 0 (0) 0 (0) 8 (14.6) 1 (1.82) 9.06 (6.13–15.3)
11 (30.6) 15 (41.7) 1 (2.78) 4 (11.1) 5 (13.9) 0 (0) 10.2 (6.32–20.1)
P-value Survived inhospital n = 72
0.007 0.066 0.009 0.40 0.022 N0.99 N0.99 0.27
Died inhospital
P-value
n = 19
14 (73.7) 5 (26.3)
0.001 0.19
b0.001 b0.001
19 (100) 53 (73.6) 1 (1.39) 4 (5.56) 13 (18.1) 1 (1.39) 9.13 (6.33–17.0)
9.10 (1.29–12.6)
0.11
N.J. Rhodes et al. / Diagnostic Microbiology and Infectious Disease 82 (2015) 165–171
169
Table 3 Cefepime renal dose equivalents received according to product labeling. MIC ≤2 mg/L
MIC ≥4 mg/L
Parameter
n = 55
n = 36
Cefepime dosing 2 g every 8 hours 2 g every 12 hours 1 g every 12 hours 500 mg every 12 hours
46 (83.6) 9 (16.4) 0 (0) 0 (0)
29 (80.6) 6 (16.7) 1 (2.78) 0 (0)
fluoroquinolone use during the admission of the GNBSI was also uncommon with only 13 cases. Mortality rates were also not significantly different between subjects who received a fluoroquinolone prior to cefepime therapy and those who did not (15.4% versus 21.8%; P = 0.73). 3.5. Microbiology of infections The distribution of pathogens according to CART-derived clinical MIC breakpoint and inhospital mortality outcome is shown in Table 1. BSIs due to Acinetobacter spp. and Pseudomonas spp. more frequently had MICs ≥4 mg/L (8.3% versus 0%; P = 0.059 and 52.8% versus 20%; P = 0.001, respectively), while BSIs due to Escherichia coli more frequently had MICs ≤2 mg/L (47.3% versus 16.7%; P = 0.003). Organism category was not significantly associated with death in univariate analyses (Table 1) nor did it significantly improve mortality predictions in multivariate analyses (Table S2). The addition of GN genera in place of categorical cefepime MIC was less explanatory than the categorical treatment of cefepime MIC (AIC = 84.7 versus 98.5) (Table S2). 4. Discussion Elevated cefepime MICs predicted inhospital mortality among patients with GNBSI. The association of higher MICs with worse outcomes is supported by the difference in outcomes observed in the clinical breakpoint model, the categorical MIC model, and the time-to-death model. The near homogenous cefepime dosing administered to our patients allowed exploration of MIC as a mortality predictor, which is important since this is the most accessible parameter available to clinicians and is used by the CLSI to classify susceptible, susceptible dose dependent, and resistant. This study is unique in that 1) most patients received aggressive doses (the renal equivalent of 2 g every 8 hours) of cefepime as per our institutional practice for patients with severe infections; 2) our patients represented a highly comorbid population, with a majority of patients having had neutropenia (61/91, 67%); and 3) a variety of organisms with cefepime MICs between 2 and 8 mg/L were reasonably represented. Our data support a clinical MIC breakpoint of 2 mg/L for various GNBSI, which is similar to the new CLSI recommendation (CLSI, 2014) and the EUCAST breakpoint for Enterobacteriaceae (EUCAST, 2014) but lower than the CLSI and EUCAST breakpoint for nonfermentative organisms like Pseudomonas spp. Our data also suggest caution in treating MICs 4–8 mg/L with cefepime, though additional study is necessary before formal recommendations are possible. Table 4 Multivariate model of the contribution of cefepime MIC ≥4 mg/L to the odds of inhospital mortality. Parameter
Log10 days to positive culture Cefepime MIC ≥4 mg/L Modified APACHE II (day 0) N16.5 Abbreviation: OR = odds ratio.
Cefepime MIC ≥4 mg/L (model 1) Univariate analysis OR (95% CI)
Multivariate analysis OR (95% CI)
4.99 (1.67–14.9) 2.59 (0.92–7.25) 2.42 (0.83–7.08)
5.38 (1.67–17.3) 3.21 (1.02–10.2) 3.06 (0.93–10.1)
P-value
0.63 0.71 0.97 0.40
Survived inhospital
Died inhospital
n = 72
n = 19
59 (81.9) 12 (16.7) 1 (1.39) 0 (0)
16 (84.2) 3 (15.8) 0 (0) 0 (0)
P-value
N0.99 N0.99 N0.99 N0.99
Our results are generally consistent with others studies. Bhat et al. (2007) studied 204 episodes of GNBSIs in patients receiving cefepime 1–2 g intravenously every 12 hours. They found higher 28-day mortality at a cefepime MIC of ≥8 mg/L than at a cefepime MIC of b8 mg/L (54.8% versus 24.1%; P = 0.001). Although immune compromise and neutropenia in our population was higher than Bhat et al. (2007), neutropenia status was evaluated and did not influence our findings. Chopra et al. (2012) evaluated the extent to which empiric or directed use of cefepime or a carbapenem for ESBL-producing GNBSI impacted clinical outcomes. When empiric cefepime use was evaluated in the multivariate model, there was no association with increased mortality. Lee et al. (2013) studied patients treated with cefepime for ESBL-producing GNBSI in Taiwan. The authors observed increased odds of 30-day mortality and a lower survival rate when cefepime monotherapy was utilized. In our limited dataset, we did not observe a significant association of ESBL status, drug dose, or receipt of combination therapy on inpatient mortality in univariate or multivariate analyses. Limitations must be considered. First, this was a single-center, retrospective cohort using a sample size of convenience and is subject to inherent biases. Second, though pharmacokinetic and clinical studies have shown higher likelihood of target attainment (Kuti et al., 2003; Nicasio et al., 2009) and improved mortality (Lodise et al., 2007) by utilizing more frequent or prolonged infusions, the majority of patients in this study received maximal standard infusions, and only 2 (2.2%) received continuous infusion, both of whom survived. Though one may criticize an MIC-only analysis, PK data (serum drug monitoring) were not available within our cohort, and thus, a PK/PD analysis could not be undertaken. However, as MIC values represent the most commonly available clinical indicators of resistance upon which to base drug selection, the impact of this variable on mortality was evaluated. Third, many
Table 5 Multivariate model of the contribution of increasing categorical cefepime MIC relative to a referent category of 1 mg/L to the odds of inhospital mortality. Parameter
Cefepime MICc 1 mg/L 2 mg/L 4 mg/L 8 mg/L 16 mg/L 32 mg/L 64 mg/L Modified APACHE II (day 0) Log10 days to positive culture a
Cefepime MIC as a categorical variable (model 2) Univariate analysis OR (95% CI)a
Multivariate analysis OR (95% CI)b
0.58 (0.20–1.65) 0.31 (0.04–2.55) 4.79 (1.22–18.8) 0.82 (0.16–4.17)
0.79 (0.08–7.88) 6.47 (1.23–33.4) 1.58 (0.25–10.0)
d
d
b
d
d
6.13 (1.24–30.3) 1.04 (0.94–1.16) 4.99 (1.67–14.9)
6.54 (1.03–41.4) 1.07 (0.94–1.23) 4.85 (1.41–16.7)
Univariate referent category was all other MIC categories for each MIC. Multivariate referent category was MIC = 1 mg/L. To test the importance of the inclusion of cefepime MIC in the multivariate model of mortality, we conducted a likelihood ratio test of the final model 2 to a model without categorical cefepime MIC (referent category of 1 mg/L). We found that model 2 with categorical cefepime MIC was significantly more informative than a model without categorical cefepime MIC (AIC [9 df] = 85 versus AIC [3 df] = 86.5; P = 0.036). d Parameter estimates were poor at MICs of 16 and 32 mg/L, and subjects with MICs in these categories were excluded from the model as the outcome did not vary in these categories. b c
N.J. Rhodes et al. / Diagnostic Microbiology and Infectious Disease 82 (2015) 165–171
0.2
0.4
Death
0.6
0.8
Yes
No
0
Probability of Death
1.0
170
1
2
4
8
16
32
64
Cefepime MIC (mg/L) Mean Predicted Death (%)
Individual Predicted Death (%)
95% CI Probability of Death (%)
Observed Death (Yes/No)
Fig. 2. Adjusted probabilities of death⁎ and observed inhospital mortality according to cefepime MIC category.⁎⁎. ⁎Probabilities of death adjusted for regression covariates as shown in Table 3, model 2. Each MIC category was regressed against the outcome of death while holding modified APACHE II score and the log10-normalized LOS prior to culture at their mean values (16.9 and 5.6 days, respectively). ⁎⁎Individual predicted death and observed death are jittered for visual appreciation of the count at each MIC.
0.75 0.25
0.50
P=0.49
0.00
Proportion on Cefepime
1.00
patients in our study were neutropenic, which qualifies for maximal dosing of cefepime at our institution. Results may be less applicable to healthier patients. Fourth, we did not observe significant increases in mortality in the cefepime MIC categories of 8, 16, or 32 mg/L; however, sample size was constrained within these MIC categories, and multivariate predictions were poor at MICs of 16 and 32 mg/L. Despite the less robust fit, the log-linear assessment was consistent with our categorical MIC assessment, supporting higher mortality with higher cefepime MICs. Additionally, sample size constraints limited our ability to identify whether specific GN genera were more closely associated with in hospital mortality. Fifth, the receipt of other active agents in addition to or after cefepime administration represented a confounding variable. However, the evaluation of this effect in the multivariate analysis mitigated the impact of this limitation for assessing the impact of cefepime MIC on mortality outcomes. Sixth, MICs were obtained using automated susceptibility testing by the Vitek 2 system as opposed to agar dilution or broth microdilution; however, Vitek 2 is a Food and Drug Administration–approved method and frequently used for clinical decision making and was the only source of MICs available during the study period. We have shown that there is clinical evidence to support lower cefepime susceptibility breakpoints (≤2 mg/L) for various GNBSI when using Vitek 2 for susceptibility testing. Although the impact of increasing MIC requires confirmation in larger and more robust studies, we recognize that quantifying outcomes for patients with more resistant organisms and for
0
.5
1
1.5
2
2.5
3
3.5
4
Time in Days Died = No
Died = Yes
Fig. 3. Time to cefepime initiation stratified by outcome of inhospital mortality.
those treated with cefepime monotherapy is challenging. Until a larger study is conducted with a greater representation of MICs between 8 and 32 mg/L and a more diverse representation or organisms at each MIC, we believe that it is prudent to consider GNBSIs in highly comorbid patients with an MIC ≥4 mg/L as potentially clinically resistant. Acknowledgments We would like to acknowledge the following individuals who supported the development of this manuscript and provided reviews of the data presented herein: Sonia Rao, PharmD, and Sheila Wang, PharmD. We thank Dr. Steven J.H. Samuels at the University at Albany, State University of New York, and Dr. Nicholas J. Cox at Durham University, United Kingdom, for their kind input regarding the Stata code for CIs of marginal effects. Funding No external funding was received. Transparency declaration All authors: No conflicts of interest. Appendix A. Supplementary data Supplementary data to this article can be found online at http://dx. doi.org/10.1016/j.diagmicrobio.2015.03.005. References Bellomo R, Ronco C, Kellum JA, Mehta RL, Palevsky P, Acute Dialysis Quality Initiative, w. Acute renal failure—definition, outcome measures, animal models, fluid therapy and information technology needs: the Second International Consensus Conference of the Acute Dialysis Quality Initiative (ADQI) Group. Crit Care 2004;8:R204–12. Bhat SV, Peleg AY, Lodise TP, Shutt KA, Capitano B, Potoski BA, et al. Failure of current cefepime breakpoints to predict clinical outcomes of bacteremia caused by gramnegative organisms. Antimicrob Agents Chemother 2007;51:4390–5. Chopra T, Marchaim D, Veltman J, Johnson P, Zhao JJ, Tansek R, et al. Impact of cefepime therapy on mortality among patients with bloodstream infections caused by extended-spectrum-beta-lactamase-producing Klebsiella pneumoniae and Escherichia coli. Antimicrob Agents Chemother 2012;56:3936–42. Clinical and Laboratory Standards Institute. Performance standards for antimicrobial susceptibility testing. M100-S20; 2010 [Wayne, PA]. Clinical and Laboratory Standards Institute (CLSI). Performance standards for antimicrobial susceptibility testing. M100-S23; 2013 [Wayne, PA]. Clinical and Laboratory Standards Institute (CLSI). Performance standards for antimicrobial susceptibility testing. M100-S24; 2014 [Wayne, PA]. Esterly JS, Wagner J, McLaughlin MM, Postelnick MJ, Qi C, Scheetz MH. Evaluation of clinical outcomes in patients with bloodstream infections due to Gram-negative bacteria according to carbapenem MIC stratification. Antimicrob Agents Chemother 2012;56: 4885–90. Hamilton KW, Bilker WB, Lautenbach E. Controlling for severity of illness in assessment of the association between antimicrobial-resistant infection and mortality: impact of calculation of Acute Physiology and Chronic Health Evaluation (APACHE) II scores at different time points. Infect Control Hosp Epidemiol 2007;28:832–6. Hosmer DW, Lemeshow S. Applied logistic regression. 2nd ed. New York: Wiley; 2000. Hospira. MAXIPIME (cefepime hydrochloride) IV package insert. Lake Forest, IL: Hospira, Inc.; 2013. Kuti JL, Dandekar PK, Nightingale CH, Nicolau DP. Use of Monte Carlo simulation to design an optimized pharmacodynamic dosing strategy for meropenem. J Clin Pharmacol 2003;43:1116–23. Kuti JL, Nightingale CH, Nicolau DP. Optimizing pharmacodynamic target attainment using the MYSTIC antibiogram: data collected in North America in 2002. Antimicrob Agents Chemother 2004;48:2464–70. Lee SY, Kuti JL, Nicolau DP. Cefepime pharmacodynamics in patients with extended spectrum beta-lactamase (ESBL) and non-ESBL infections. J Infect 2007;54:463–8. Lee NY, Lee CC, Huang WH, Tsui KC, Hsueh PR, Ko WC. Cefepime therapy for monomicrobial bacteremia caused by cefepime-susceptible extended-spectrum beta-lactamase-producing Enterobacteriaceae: MIC matters. Clin Infect Dis 2013;56: 488–95. Lodise Jr TP, Lomaestro B, Drusano GL. Piperacillin-tazobactam for Pseudomonas aeruginosa infection: clinical implications of an extended-infusion dosing strategy. Clin Infect Dis 2007;44:357–63. Nicasio AM, Ariano RE, Zelenitsky SA, Kim A, Crandon JL, Kuti JL, et al. Population pharmacokinetics of high-dose, prolonged-infusion cefepime in adult critically ill patients with ventilator-associated pneumonia. Antimicrob Agents Chemother 2009;53: 1476–81. Song W, Moland ES, Hanson ND, Lewis JS, Jorgensen JH, Thomson KS. Failure of cefepime therapy in treatment of Klebsiella pneumoniae bacteremia. J Clin Microbiol 2005;43: 4891–4.
N.J. Rhodes et al. / Diagnostic Microbiology and Infectious Disease 82 (2015) 165–171 The European Committee on Antimicrobial Susceptibility Testing (EUCAST) 2014. Breakpoint tables for interpretation of MICs and zone diameters. Version 4.0, vol. 2014. ; 2014. Thom KA, Shardell MD, Osih RB, Schweizer ML, Furuno JP, Perencevich EN, et al. Controlling for severity of illness in outcome studies involving infectious diseases: impact of measurement at different time points. Infect Control Hosp Epidemiol 2008;29:1048–53. Tumbarello M, Sanguinetti M, Montuori E, Trecarichi EM, Posteraro B, Fiori B, et al. Predictors of mortality in patients with bloodstream infections caused by extended-
171
spectrum-beta-lactamase-producing Enterobacteriaceae: importance of inadequate initial antimicrobial treatment. Antimicrob Agents Chemother 2007;51:1987–94. Ury HK. Efficiency of case-control studies with multiple controls per case: continuous or dichotomous data. Biometrics 1975;31:643–9. Wi YM, Kang CI, Cheong HS, Chung DR, Jung CW, Ko KS, et al. Failure of cefepime therapy in neutropenic patients with extended-spectrum beta-lactamase-producing Gramnegative bacteraemia. Int J Antimicrob Agents 2009;33:384–6.