Outcomes of children with enterobacteriaceae bacteremia with reduced susceptibility to ceftriaxone: do the revised breakpoints translate to improved patient outcomes?

Evaluation of parameters affecting performance and reliability of machine learning-based antibiotic susceptibility testing from whole genome sequencing data

Prediction of antibiotic resistance phenotypes from whole genome sequencing data by machine learning methods has been proposed as a promising platform for the development of sequence-based diagnostics. However, there has been no systematic evaluation of factors that may influence performance of such models, how they might apply to and vary across clinical populations, and what the implications might be in the clinical setting. Here, we performed a meta-analysis of seven large Neisseria gonorrhoeae datasets, as well as Klebsiella pneumoniae and Acinetobacter baumannii datasets, with whole genome sequence data and antibiotic susceptibility phenotypes using set covering machine classification, random forest classification, and random forest regression models to predict resistance phenotypes from genotype. We demonstrate how model performance varies by drug, dataset, resistance metric, and species, reflecting the complexities of generating clinically relevant conclusions from machine learning-derived models. Our findings underscore the importance of incorporating relevant biological and epidemiological knowledge into model design and assessment and suggest that doing so can inform tailored modeling for individual drugs, pathogens, and clinical populations. We further suggest that continued comprehensive sampling and incorporation of up-to-date whole genome sequence data, resistance phenotypes, and treatment outcome data into model training will be crucial to the clinical utility and sustainability of machine learning-based molecular diagnostics.

Querying automated antibiotic susceptibility testing instruments: a novel population-based active surveillance method for multidrug-resistant gram-negative bacilli

OBJECTIVE

To describe the implementation of a population-based surveillance system for multidrug-resistant gram-negative bacilli (MDR-GNB).

DESIGN

Population-based active surveillance by the Georgia Emerging Infections Program.

SETTING

Metropolitan Atlanta, starting November 2010.

PATIENTS

Residents with MDR-GNB isolated from urine or a normally sterile site culture.

METHODS

Surveillance was implemented in 3 phases: (1) surveying laboratory antibiotic susceptibility testing practices, (2) piloting surveillance to estimate the proportion of GNB that were MDR, and (3) maintaining ongoing active surveillance for carbapenem-nonsusceptible Enterobacteriaceae and Acinetobacter baumannii using the 2010 Clinical and Laboratory Standards Institute (CLSI) breakpoints. Pilot surveillance required developing and installing queries for GNB on the 3 types of automated testing instruments (ATIs), such as MicroScan, in Atlanta's clinical laboratories. Ongoing surveillance included establishing a process to extract data from ATIs consistently, review charts, manage data, and provide feedback to laboratories.

RESULTS

Output from laboratory information systems typically used for surveillance would not reliably capture the CLSI breakpoints, but queries developed for the 3 ATIs did. In November 2010, 0.9% of Enterobacteriaceae isolates and 35.7% of A. baumannii isolates from 21 laboratories were carbapenem nonsusceptible. Over a 5-month period, 82 Enterobacteriaceae and 59 A. baumannii were identified as carbapenem nonsusceptible.

CONCLUSIONS

Directly querying ATIs, a novel method of active surveillance for MDR-GNB, proved to be a reliable, sustainable, and accurate method that required moderate initial investment and modest maintenance. Ongoing surveillance is critical to assess the burden of and changes in MDR-GNB to inform prevention efforts.

Treatment of KPC-producing Enterobacteriaceae: suboptimal efficacy of polymyxins

Treatment of Klebsiella pneumoniae carbapenemase-producing Enterobacteriaceae infections (KPC-EI) remains a challenge. Combined therapy has been proposed as the best choice, but there are no clear data showing which combination therapy is superior. Our aim was to evaluate the effectiveness of antimicrobial regimens for treating KPC-EI. This was a retrospective cohort study of KPC-EI nosocomial infections (based on CDC criteria) between October 2009 and June 2013 at three tertiary Brazilian hospitals. The primary outcomes were the 30-day mortality for all infections and the 30-day mortality for patients with bacteraemia. Risk factors for mortality were evaluated by comparing clinical variables of survivors and nonsurvivors. In this study, 118 patients were included, of whom 78 had bacteraemia. Catheter-related bloodstream infections were the most frequent (43%), followed by urinary tract infections (n = 27, 23%). Monotherapy was used in 57 patients and combined treatment in 61 patients. The most common therapeutic combination was polymyxin plus carbapenem 20 (33%). Multivariate analysis for all infections (n = 118) and for bacteremic infections (n = 78) revealed that renal failure at the end of treatment, use of polymyxin and older age were prognostic factors for mortality. In conclusion, polymyxins showed suboptimal efficacy and combination therapy was not superior to monotherapy.

Effects of breakpoint changes on carbapenem susceptibility rates of Enterobacteriaceae: Results from the SENTRY Antimicrobial Surveillance Program, United States, 2008 to 2012

In the absence of clinical resistance, breakpoints for many antimicrobial agents are often set high. Clinical failures following use of the agents over time requires re-evaluation of breakpoints. This is based on patient response, pharmacokinetic/pharmacodynamic information and in vitro minimal inhibitory concentration data. Data from the SENTRY Antimicrobial Surveillance Program has shown that Clinical and Laboratory Standards Institute breakpoint changes for carbapenems that occurred between 2008 and 2012 in North America have resulted in decreased levels of susceptibility for some species. In particular, reduced susceptibility to imipenem was observed for Proteus mirabilis (35%) and Morganella morganii (80%). Minor decreases in susceptibility were also noted for Enterobacter species with ertapenem (5%) and imipenem (4.3%), and Serratia species with imipenem (6.4%). No significant decreases in susceptibility were observed for meropenem following the breakpoint changes. There were no earlier breakpoints established for doripenem. Very few of these Enterobacteriaceae produce carbapenamase enzymes; therefore, the clinical significance of these changes has not yet been clearly determined. In conclusion, ongoing surveillance studies with in vitro minimum inhibitory concentration data are essential in predicting the need for breakpoint changes and in identifying the impact of such changes on the percent susceptibility of different species.

Determining the optimal ceftriaxone MIC for triggering extended-spectrum β-lactamase confirmatory testing

As routine testing of clinical isolates for extended-spectrum β-lactamase (ESBL) production (screen plus phenotypic confirmatory testing) is no longer required by the Clinical and Laboratory Standards Institute (CLSI), a number of clinical microbiology laboratories use ceftriaxone MICs as a proxy means of identifying bacteria as potential ESBL producers. Data from 1,386 clinical isolates suggest that a ceftriaxone MIC cutoff of 8 μg/ml is an excellent predictor of ESBL production, with a positive predictive value and negative predictive value approaching 100% and 99.5%, respectively.

The use of cefepime for treating AmpC β-lactamase-producing Enterobacteriaceae

BACKGROUND

 AmpC β-lactamase-producing organisms are associated with significant morbidity and mortality. Induction of resistance to third-generation cephalosporins after exposure to these agents complicates treatment options and carbapenems are considered optimal therapy. The role of cefepime, however, remains unclear. Our objective was to compare clinical outcomes for patients receiving cefepime compared with meropenem for invasive infections caused by organisms expressing AmpC β-lactamases.

METHODS

 Hospitalized patients with blood, bronchoalveolar lavage, or intra-abdominal fluid cultures growing Enterobacter spp, Serratia spp, or Citrobacter spp were evaluated using the cefotetan-boronic acid disk test and the cefotetan-cloxacillin Etest to identify organisms with AmpC β-lactamase production from February 2010 to January 2011. In patients with organisms hyperproducing AmpC β-lactamases (positive by both methods), clinical outcomes for patients receiving cefepime or meropenem therapy were compared. To minimize the possibility of treatment selection bias, 1:1 nearest neighbor propensity score matching was performed prior to regression analysis.

RESULTS

 Of 399 patients meeting eligibility criteria, 96 (24%) had confirmed infections with AmpC β-lactamase-producing organisms. Propensity score matching of patients infected with AmpC β-lactamase-positive organisms treated with cefepime or meropenem yielded 32 well-balanced patient pairs with no difference in 30-day mortality (odds ratio, 0.63; 95% confidence interval [CI], .23-2.11; P = .36) or length of hospital stay after infection (relative risk, 0.96; 95% CI, .79-1.26; P = .56) between the 2 groups.

CONCLUSIONS

 Cefepime may be a reasonable option for the treatment of invasive infections due to AmpC β-lactamase-producing organisms, particularly when adequate source control is achieved.