β-Lactamase Characterization of Gram-Negative Pathogens Recovered from Patients Enrolled in the Phase 2 Trials for Ceftazidime-Avibactam: Clinical Efficacies Analyzed against Subsets of Molecularly Characterized Isolates

The correlation of the clinical efficacies of ceftazidime-avibactam and comparators (carbapenems) was evaluated against baseline Gram-negative isolates having characterized β-lactam resistance mechanisms from complicated urinary tract infection (cUTI) and complicated intra-abdominal infection (cIAI) phase 2 trials. Enterobacteriaceae displaying ceftriaxone and/or ceftazidime MICs of ≥2 μg/ml (69 isolates) and nonfermentative Gram-negative bacilli (NF-GNB [three isolates]) with ceftazidime MICs of ≥16 μg/ml were characterized for their narrow- and extended-spectrum β-lactamase (ESBL) content. Enterobacteriaceae (one isolate) and NF-GNB (three isolates) with imipenem/meropenem MICs of ≥2 and ≥16 μg/ml, respectively, were tested for carbapenemases. All cUTI E. coli had the lineage background investigated (ST131-like versus non-ST131-like). The primary efficacy endpoint was microbiological response (eradication) at test of cure (TOC) for cUTI and clinical response (inferred microbiological eradication) at TOC for cIAI. A total of 34.1% of baseline cUTI (36.4%) and cIAI (33.1%) pathogens met the MIC-based screening criteria (screen positive). All screen-positive cUTI pathogens were CTX-M-producing E. coli, except for one E. cloacae isolate with AmpC overexpression. The majority (66.7%) of screen-positive cIAI isolates produced CTX-M-type coupled with a diverse array of other β-lactamases. Similar favorable responses were observed with ceftazidime-avibactam (93.3%) and carbapenems (90.9%), when a non-ESBL Enterobacteriaceae isolate was recovered at the baseline visit. When an ESBL Enterobacteriaceae isolate was present, the favorable responses were 85.7% and 80.0% with ceftazidime-avibactam and carbapenems, respectively. Higher favorable responses were observed with ceftazidime-avibactam (75.0%) than with carbapenems (66.7%) when an ST131-like E. coli isolate was recovered at baseline, as when a non-ST131-like isolate was present (93.8% versus 86.7%, respectively). The efficacy of ceftazidime-avibactam was similar to that of carbapenems for treatment of cUTI and cIAI caused by ESBL organisms.

and tolerability of ceftazidime-avibactam versus comparator agents for treatment of complicated urinary tract infections (cUTI) and complicated intra-abdominal infections (cIAI) (11,12). The present study characterized the ␤-lactamase genes in baseline pathogens recovered during those phase 2 trials. In addition, this study correlates the efficacy of ceftazidime-avibactam and comparators against subsets of isolates harboring ␤-lactam resistance mechanisms.

MATERIALS AND METHODS
Patients, clinical isolates, study treatment, and endpoints. Male and female patients between the ages of 18 and 90 years were enrolled in the phase 2 clinical trials for ceftazidime-avibactam (clinicaltrials.gov identifiers NCT00752219 and NCT00690378) (11,12). Hospitalized patients were enrolled from medical centers located in Guatemala, India, Jordan, Lebanon, and the United States for the cUTI trial and Bulgaria, France, India, Lebanon, Poland, Romania, Russia, and the United States for the cIAI trial.
Patients eligible for the cUTI trial were stratified by diagnosis (acute pyelonephritis or other cUTI) and randomized 1:1 to 0.5 g ceftazidime-0.125 g avibactam (here, "0.5/0.125 g") (intravenous [i.v.] every 8 h) or imipenem-cilastatin (0.5 g i.v. every 6 h). Oral ciprofloxacin (0.5 g twice daily) or alternative oral therapy was permitted after at least 4 days of initial i.v. therapy. Pathogens were recovered via urine culture at the baseline (the first one being termed "baseline") and follow-up visits. Blood cultures were performed when clinically indicated or in patients with indwelling catheters and stents. The primary efficacy endpoint was microbiological response at test of cure (TOC), 5 to 9 days after end of treatment in the microbiologically evaluable (ME) population. Clinical response was a secondary endpoint.
For cIAI, eligible patients with a presumed (preoperative) or definitive (intraoperative or postoperative) diagnosis of cIAI were randomized 1:1 to ceftazidime-avibactam (2/0.5 g i.v. every 8 h) and metronidazole (0.5 g i.v. every 8 h) or meropenem (1 g i.v. every 8 h) plus placebo. Isolates were obtained during the study qualifying operative procedure (i.e., laparotomy, laparoscopy, or percutaneous drainage) and during any subsequent operative procedures. Blood cultures were also collected from all patients at the baseline visit. The primary efficacy endpoint was clinical response at TOC, 2 weeks after the end of treatment in the ME population. The ME population was defined as clinically evaluable patients with a confirmed cUTI or cIAI, at least one baseline pathogen susceptible to the study therapy agents, who received treatment, and who had a clinical assessment at the TOC visit (see references 11 and 12 for additional information).
For the cUTI trial, a favorable response was defined as eradication of all pathogens from urine (Ͻ10 4 CFU/ml) and blood, while in the cIAI trial, a favorable response was defined as complete resolution or significant improvement of signs/symptoms of infection with no requirements for additional antimicrobial therapy (11,12).
The transcription levels of the chromosomally encoded AmpC were determined in Pseudomonas aeruginosa, E. coli, and Enterobacter cloacae. The transcription levels of the chromosomal ampC gene were determined by the quantification of the target gene mRNA using a normalized expression analysis method and relative comparison to susceptible control strains (18,19). A given strain was considered to overexpress the ampC gene when at least a 10-fold greater difference of ampC transcripts was detected compared with a species-specific wild-type reference control strain.
Determination of E. coli strains associated with ST131 and PFGE. All E. coli strains recovered during the cUTI trial were investigated by pulsed-field gel electrophoresis (PFGE) for the presence of two single nucleotide polymorphisms (SNPs), namely, thymine 144 and adenine 450 in the pabB gene, as previously described by Dhanji et al. (20). The presence of these two SNPs is known to be unique to sequence type 131 (ST131) strains, and they were described here as "ST131-like."

Study population.
A total of 192 ME patients were included in the study, with 26 and 20 ME patients included in the ceftazidimeavibactam and carbapenem arms, respectively, for the cUTI trial, while 60 and 70 patients, respectively, were included in the arms for the cIAI trial (11,12). Most ME patients had a single baseline aerobic Gram-negative pathogen recovered during the screening visit, except for four (4/62 [6.5%]) patients in the cUTI trial and 14 (14/130 [10.8%]) patients in the cIAI trial who had multiple aerobic Gram-negative organisms recovered at the baseline visit (see Tables 2 and 3).
Antimicrobial susceptibility profiles. Table 1 lists the MIC values obtained for baseline isolates recovered from ME patients enrolled in both the cIAI and cUTI trials. Baseline Enterobacteriaceae isolates recovered from cIAI demonstrated a bimodal MIC distribution when tested against ceftazidime, with one modal MIC of 0.12 g/ml and a second mode at Ͼ32 g/ml. Avibactam restored the ceftazidime activity, including against ESBL-positive isolates, with a modal MIC value of 0.06 g/ml (the highest MIC of 2 g/ml). One exception was observed against a carbapenemase (NDM-1)-producing K. pneumoniae strain exhibiting ceftazidime-avibactam MIC results of Ͼ32 g/ml. All baseline NF-GNB from the cIAI trial showed ceftazidime-avibactam MICs at Յ8 g/ml, except for one P. aeruginosa isolate (32 g/ml) expressing VIM-2 and two A. baumannii strains producing PER-1 (32 g/ml) or OXA-23 (Ͼ32 g/ml). Against baseline Enterobacteriaceae from cUTI patients, ceftazidime alone showed a bimodal MIC distribution and inhibited 68.3% of tested isolates at the breakpoint for susceptibility (i.e., Յ4 g/ml). Ceftazidime-avibactam and imipenem had the same modal MIC value (0.12 g/ml) against Enterobacteriaceae isolates causing cUTI.
␤-Lactamase profiles. Totals of 38.7% (24/62) and 35.4% (46/ 130) of the ME patients in the cUTI and cIAI trials, respectively, had baseline pathogens that met the MIC screening criteria (i.e., were screen positive) (Tables 2 and 3). E. coli alone comprised the vast majority (88.7%) of the causative pathogens responsible for cUTI, and 38.2% met the MIC screening criteria. These isolates produced CTX-M-14 (9.5%) or CTX-M-15 (90.5%), and almost half (42.1%) of the CTX-M-15 E. coli producers also carried bla OXA-1/30 . In addition, hyperproduction of the intrinsic chromosomally encoded AmpC enzyme was detected in one CTX-M-15 E. coli isolate. Except for one E. cloacae strain hyperproducing AmpC, other baseline cUTI pathogens did not meet the screening criteria (i.e., were screen negative). Among aerobic Gram-negative bacteria, E. coli alone represented the vast majority of baseline cIAI pathogens (70.0%), and approximately one-third (36.3%) of those isolates met the MIC screening criteria. Among these, 75.8% (25/33) carried bla CTX-M-15 alone or in combination with bla SHV and/or bla OXA-1/30 and/or plasmid AmpC genes (i.e., bla ACC and bla CMY variants). Other E. coli isolates carried bla OXA-1/30 and/or bla SHV and/or bla CMY enzymes. Among K. pneumoniae isolates, 50.0% met the MIC screening criteria, and bla CTX-M-15 , either alone or with bla OXA-1/30 and/or bla SHV-5 , was present in 50.0% (3/6) of investigated bacteria. One K. pneumoniae isolate carried bla NDM-1 and bla CTX-M-15 . NF-GNB (three P. aeruginosa isolates) recovered from cUTI patients were screen negative (Table 2), while three out of 15 cIAI NF-GNB met the MIC screening criteria. These three isolates were two A. baumannii isolates-one producing OXA-23 and the Screen negative for E. coli (1) and C. koseri (1) E. coli-P. mirabilis (1) CTX-M-15 for E. coli (1) and screen negative for P. mirabilis (1) P. mirabilis (1) Screen negative (1) P. aeruginosa (1) Screen negative (1) P. aeruginosa-E. coli (2) Screen negative for P. aeruginosa (1)  other PER-1-and one VIM-2-producing P. aeruginosa isolate (Table 3). Efficacy analysis of ceftazidime-avibactam and carbapenems in the ME population. Table 4 summarizes the favorable response at the TOC assessment among the ME patients enrolled in the cUTI and cIAI phase 2 trials for ceftazidime-avibactam. Organisms associated with clinical or microbiological failure are presented in Table 5. Overall, favorable responses at the TOC visit were observed in 93.3 and 90.9% of ceftazidime-avibactam and comparator patients, respectively, for both trials combined when baseline MIC-based screen-negative Enterobacteriaceae isolates were recovered and in 85.7 and 80.0% of patients, respectively, when an ESBL-producing Enterobacteriaceae isolate was recovered at baseline. Favorable outcome was noted in 84.6% (22/26) of patients in the ceftazidime-avibactam arm when CTX-M-15 (alone or with additional enzymes)-producing Enterobacteriaceae was recovered at the baseline visit and in 79.2% (19/24) of patients in the carbapenem arms. In neither the ceftazidime-avibactam nor carbapenem arms of either of the trials could the microbiological or clinical failures be ascribed to high MIC of the antibacterial agent or possession of particular ␤-lactamase genes ( Table 5).
Similar proportions of patients with favorable outcomes were observed for the ceftazidime-avibactam and carbapenem arms in the phase 2 cIAI trial, regardless of the presence of MIC-based screen-positive or -negative Enterobacteriaceae recovered at baseline (91.7 and 90.3% for ceftazidime-avibactam and 94.4 and 90.9% for carbapenem, respectively). Favorable outcome in the ceftazidime-avibactam arm of the phase 2 cUTI trial was less frequent (72.7%) when an isolate that met the MIC screening criteria was recovered at baseline than for patients with MIC-based screen-negative isolates (93.8%). The favorable outcome rate (61.5%) in the imipenem cUTI arm when MIC-based screen-positive Enterobacteriaceae were identified at baseline was lower than that (90.9%) for patients with a screen-negative isolate.

DISCUSSION
Overall, 35.2% (38.1 and 33.8% of cUTI and cIAI isolates, respectively) of all baseline Enterobacteriaceae met the MIC-basedscreening criteria at the baseline visit. Previous studies have reported rates of ESBL isolates of 3.7 and 2.4% for cUTI and cIAI, respectively, in large phase 3 clinical trials (21) or 16.2% in community-acquired cIAI (22). Recent phase 3 clinical trials for cef-
pneumoniae (1) Screen negative for P. fluorescens (1), E. coli (1), and K. pneumoniae (1) P. stutzeri-E. coli (1) Screen negative for P. stutzeri (1) and E. coli (1) a The cIAI results shown represent 130 ME patients. b Screen negative, isolates that did not meet the MIC-based screening criteria; screen positive, isolates that met the screening criteria (i.e., ceftriaxone and/or ceftazidime MICs of Ն2 g/ml, NF-GNB with ceftazidime MICs of Ն16 g/ml, and Enterobacteriaceae and NF-GNB isolates exhibiting imipenem and meropenem MICs of Ն2 and Ն16 g/ml, respectively). CMY-42 is a single-amino-acid-change (V231S) variant of CMY-2. c Some isolates met the phenotypic MIC screening criteria, but targeted ␤-lactamase resistance mechanisms were not detected.
tolozane-tazobactam reported ESBL rates in Enterobacteriaceae of 14.8 and 7.2% for cUTI and cIAI, respectively (23,24). These rates can be influenced by geography, population, and enrollment criteria, among other factors. However, the ESBL rates reported here seem to be higher than most rates documented previously. In addition, E. coli comprised the most common pathogen in both trials, consistent with previous results (23)(24)(25)(26), and ESBL rates among this pathogen were 36.3% among baseline E. coli isolates causing cIAI and 38.2% in those causing cUTI. However, other smaller studies have reported ESBL rates of up to 60% in E. coli isolates causing cUTI (18,27). Among the MIC-based screen-positive Enterobacteriaceae presented here, production of CTX-M-type enzymes, with or without the presence of OXA-1/30, was the most prevalent ␤-lactam resistance mechanism. Recent publications have reported the increased prevalence of CTX-M-type enzymes in Enterobacteriaceae, especially E. coli, regardless of infection source (i.e., hospital versus community acquired) (14,15,28). Interestingly, compared with the baseline Enterobacteriaceae isolates causing cUTI, those responsible for cIAI demonstrated a more heterogeneous ␤-lactamase genetic profiling. The reasons for these differences are un-clear but might perhaps be attributed to geographical differences or other patient differences between trial populations.
Overall, the rate of favorable responses was 90 to 93% for both arms (ceftazidime-avibactam and carbapenems) in the aggregated data analysis (both trials combined) when an Enterobacteriaceae isolate that did not meet the screening criteria was cultured at the baseline visit. The rate of favorable responses decreased when infections were caused by isolates that met the MIC screening criteria (screen-positive isolates), with a rate for ceftazidime-avibactam (85.7%) slightly higher than that obtained for the carbapenems (80.0%). The latter overall rates were driven by those responses obtained from the cUTI trial, where the presence of a screen-positive isolate at the baseline visit predicted a less favorable outcome (72.7 and 61.5% for ceftazidime-avibactam and imipenem, respectively) compared to when a screen-negative isolate was recovered (93.8 and 90.9% for ceftazidime-avibactam and imipenem, respectively). The reasons for these findings are uncertain, although they may be related to underlying patient factors associated with acquisition of drug-resistant bacteria. The phase 2 trials were not powered to demonstrate noninferiority, a Screen negative, isolates that did not meet the MIC-based screening criteria; screen positive, isolates that met the screening criteria (i.e., ceftriaxone and/or ceftazidime MICs of Ն2 g/ml and NF-GNB with ceftazidime MICs of Ն16 g/ml, and Enterobacteriaceae and NF-GNB isolates exhibiting imipenem and meropenem MICs of Ն2 and Ն16 g/ml, respectively). b Results are expressed as the number of patients with favorable responses/total number of patients in each category (percentage) and represent the ceftazidime-avibactam or imipenem-cilastatin arms for cUTI and the ceftazidime-avibactam plus metronidazole or meropenem arms for cIAI. c Patients from whom Enterobacteriaceae pathogens were recovered at the baseline visit. and additional studies with larger numbers of patients are needed to further evaluate and confirm these results. In this study, the number of ST131 isolates causing cUTI was relatively small (19.3% versus 80.7% for non-ST131 strains). The prevalence of ST131 among E. coli isolates varies among studies, mostly due to different organism inclusion criteria (29), and rates of ST131 isolates causing UTI of 12% have been reported in one study (30), while another study documented a rate of 30% of E. coli isolates causing pyelonephritis in Australia (31). The rates of a favorable cUTI outcome due to ST131 strains reported here were 75.0 and 66.7% for ceftazidime-avibactam and imipenem-cilastatin, respectively, while higher rates were documented when non-ST131 strains were recovered at the baseline visit (93.8 and 86.7%, respectively). These results indicate a higher treatment failure for patients infected with ST131, which corroborates a recent report published by Can et al. (31). E. coli ST131 is a multidrug-resistant (MDR) clonal group that has spread throughout the world (32). Overall, clonal expansion of ST131 is the predominant mechanism for the rising prevalence of both fluoroquinolone resistance and CTX-M-15 producers among E. coli, and consequently the ST131 E. coli population causing infections in humans (33). The dissemination of MDR strains of E. coli has challenged the empirical treatment, especially UTI, and increased morbidity and mortality (34). However, a few studies have reported similar mortality rates between infections caused by ST131 and non-ST131 strains (35,36).
The results presented here document high rates of MIC-based screen-positive pathogens causing cUTI and cIAI, which support the current recommendations for empirical treatment of such infections (carbapenems for both cUTI and cIAI). These clinical trial results suggest that the efficacy of ceftazidime-avibactam (with metronidazole for cIAI) may be similar to that of imipenem for the treatment of cUTI or meropenem for the treatment of cIAI caused by ESBL producers, including prevalent CTX-M-15 Enterobacteriaceae.
Limitations of this study include the relatively small number of ME patients available for analysis in each arm of each trial and the even smaller numbers of patients infected by ESBL organisms. Therefore, trial results were also analyzed in aggregate to increase the number of patients for a more robust data analysis. Therefore, it is also important to mention that the patients originated from two different trial designs with distinct primary efficacy endpoints. Similarly, the association of specific E. coli lineages (ST131) with outcomes may be limited as a consequence of the small number sampled. In addition, this study evaluated aerobic Gram-negative pathogens only and did not take into consideration the presence of polymicrobial infections caused by Grampositive bacteria and/or anaerobes, especially in the cIAI arms. Moreover, the isolates investigated here may possess ␤-lactam resistance mechanisms other than those screened in this study, such as other ␤-lactamases and/or decreased permeability.
In summary, this study shows high proportions (36.4 and 33.1% in the cUTI and cIAI trials, respectively) of aerobic Gramnegative isolates that met the screening criteria for ESBL production. All but one baseline pathogen from the cUTI trial that met the MIC screening criteria produced CTX-M enzymes, showing the dominance of CTX-M isolates causing cUTI. In contrast, a greater diversity of ESBL-encoding genes was observed among baseline isolates causing cIAI. Moreover, the efficacy of ceftazidime-avibactam was similar to that of carbapenems for treatment of cUTI and cIAI caused by ESBL organisms. However, these phase 2 clinical trials were not powered to demonstrate noninferiority, and 95% confidence intervals for the differences in responses between ceftazidime-avibactam and carbapenem treatments were expectedly wide (11,12). Likewise, although favorable responses were similar between ceftazidime-avibactam and carbapenem for the subgroups presented here, formal statistical comparisons were not performed due to the small sample sizes and consequent large 95% confidence intervals for the treatment differences. Further and more robust analyses with a larger sample size are necessary to confirm these results, which can be expected when the phase 3 trials' results for ceftazidime-avibactam become available.