Antibiotic Strategies in the Era of Multidrug Resistance

George Karam; Jean Chastre; Mark H. Wilcox; Jean-Louis Vincent


Crit Care. 2016;20(136) 

In This Article

Management: Patient Stratification and Antibiotic Choices

Perhaps the greatest challenge imposed by MDR Gram-negative bacteria is attainment of the appropriate degree of balance between efficacy and ecology. It is well recognized that there are significant increases in mortality rates when antibiotic therapy is delayed,[29,30] challenging the clinician to prescribe therapy broad enough to cover all of the most likely pathogens. What is simultaneously recognized is the collateral damage of selecting resistant bacteria that can occur with prolonged receipt of broad-spectrum antibiotics.[31]

Three important categories can influence antimicrobial choices: patient characteristics; risk factors for infection with specific pathogens; and severity of illness. Of the demographic data of patients with infections, advanced age and comorbid illnesses have been associated with decreased reserve and, hence, an increase in mortality. Contact with the health care system is an important patient characteristic that influences the broadness of initial antibiotic therapy. Previous hospitalization (within the past 30 days as a defining feature of hospital-acquired infection and within the past 90 days as a defining feature of healthcare-associated infection) increases the risk of infection by a resistant pathogen that was acquired in the healthcare environment. Invasive procedures, even those occurring in the outpatient setting, increase the risk of colonization by MDR organisms, with such colonization serving as an antecedent predisposition to infection with MDR microorganisms. Local microbiologic data are extremely important in predicting the type of resistance that may be present in the etiologic agent causing a clinical infection. It is important to recognize that such local data may vary from unit to unit within a hospital, and when possible unit-specific data are optimal. In recognition of the fact that resistance can rapidly develop within a practice environment, it is important that microbiologic data should be as current as possible. Antibiograms reporting data that are from a previous year may be many months behind what is presently occurring. Very important is the role of prior antibiotic exposure, recognizing the risk of eliminating normal flora and allowing the selection of resistant bacteria. Although various timelines have been suggested in the medical literature about what defines "recent" antibiotic therapy, within the previous 90 days is a consistently cited number. A second element of importance in stratification regarding antibiotic therapy is the likely type of microorganism. As noted previously, risk stratification related to P. aeruginosa has been included in multiple clinical guidelines.[25,26] Of all the variables that influence stratification of patients into a category requiring coverage of MDR bacteria, severity of illness may be the most important. As severity of illness increases, the margin for error in initial antimicrobial therapy diminishes and this requires consideration of coverage against multiple patterns of bacterial resistance.

Since the reviews by Jacoby and Munoz-Price[32] and by Paterson and Bonomo[5] listing carbapenems as drugs of choice for infections caused by ESBL-producing bacteria, this class of antibiotics has been considered first-line therapy for such infections. This recommendation is influenced by the fact that ESBL-producing pathogens are often resistant to fluoroquinolones and aminoglycosides since resistance mechanisms for these classes of antibiotics are often carried on the same large plasmids that contain the genetic elements for ESBL production. Paradoxically, even though carbapenems are β-lactam agents, they are stable in the presence of ESBLs. More recently and in response to a large volume of infections in Spain caused by ESBL-producing bacteria, a group of investigators from that country conducted a post hoc analysis of patients with bloodstream infections due to ESBL-producing E. coli from six published prospective cohorts.[33] In this trial, mortality and length of hospital stay were compared in patients treated with either amoxicillin–clavulanic acid (AMC) or piperacillin–tazobactam (PTZ) versus those receiving a carbapenem. In this analysis, the infection source was the urinary or biliary tract in two-thirds of the patients. The results suggest that AMC and PTZ are suitable alternatives to carbapenems for treating patients with bloodstream infections due to ESBL-producing E. coli if there is in vitro activity. However, the variable in vitro susceptibility of ESBL-producing bacteria to β-lactam/β-lactamase inhibitor combinations warrants consideration before these agents are used as empiric therapy for serious infections.

Further addressing the question of whether PTZ is clinically as efficacious as carbapenems in the treatment of bacteremia caused by ESBL-producing organisms, investigators compared 14-day mortality when PTZ versus a carbapenem was used as empiric therapy (defined in this trial as antibiotic therapy administered to a patient before their ESBL status was known) in a cohort of 331 patients with ESBL bacteremia who received definitive therapy with a carbapenem.[34] The adjusted risk of death was 1.9 times higher for patients receiving empiric PTZ compared with empiric carbapenem (95 % confidence interval 1.07–3.45).

It is important to acknowledge that the data regarding the role of PTZ in the treatment of ESBL-producing bacteria may not be an accurate predictor of how new β-lactamase inhibitors (e.g., avibactam or relebactam) or new combinations (e.g., ceftolozane/tazobactam) may perform in the treatment of these infections. Ceftolozane/tazobactam in combination with metronidazole was shown to be noninferior to meropenem in a multinational, double-blind randomized phase 3 trial of adult patients hospitalized with complicated intraabdominal infections.[35] Of the isolated pathogens, 7.2 % were ESBL-producing Enterobacteriaceae and for this group of patients clinical cure rates were 95.8 % in the ceftolozane/tazobactam group and 88.5 % in the meropenem group. Similar findings were reported from a randomized trial of adult patients hospitalized with complicated lower urinary tract infections or pyelonephritis, in which 7.6 % of the isolated pathogens were ESBL-producing Enterobacteriaceae.[36]

In a recent review, Harris et al.[37] suggested three clinical scenarios in which β-lactam/β-lactamase inhibitors might be appropriate instead of using a carbapenem in the treatment of infections caused by ESBL-producing organisms: urinary tract infections (including with bacteremia); nonurinary tract infections in which the isolate is susceptible at a low minimum inhibitory concentration (MIC); and when adequate source control has been achieved.

In a retrospective study of monomicrobial bacteremia caused by ESBL-producing organisms, patients were definitively treated with cefepime if there was in vitro activity to that antibiotic or with a carbapenem.[38] Patients who received cefepime as definitive therapy were more likely to have clinical failure, and the survival analysis consistently found that individuals who received cefepime therapy had a lower survival rate. Based on these observations, the authors concluded that cefepime empirical therapy was inferior to carbapenem in the treatment of patients with bacteremia due to cefepime-susceptible ESBL-producing bacteria. It is noteworthy that the cefepime breakpoint for susceptibility in this study was ≤8 μg/ml, which may have overstated the susceptibility rate, compared with the more recent breakpoints of ≤2 μg/ml from the Clinical and Laboratory Standards Institute (CLSI) and ≤1 μg/ml from the European Committee on Antimicrobial Susceptibility Testing (EUCAST), which would result in lower rates of susceptibility.[39]

A major challenge in the ICU is achieving adequate therapy for infections caused by carbapenemase-producing bacteria. An important early observation regarding therapy of such infections was that monotherapy with an agent like polymyxin might not provide optimal efficacy. In an analysis of 15 studies and case reports of the therapy of infections caused by KPCs, the success of polymyxin monotherapy was 14 % in contrast to 73 % with polymyxin combinations.[40] Considerations regarding such failure include the lack of optimal dosing of polymyxin (potentially due to variables such as augmented renal clearance in patients with severe illness) versus the lack of an additive or even synergistic effect that might occur with combination therapy. Early data suggested a potential role for carbapenems as a therapeutic option if the MIC of the infecting organism was ≤4 mg/l and if the carbapenem was given in combination with another active agent.[41] In an observational study from Greece regarding the treatment of infections due to carbapenemase-producing K. pneumoniae, patients who received a carbapenem in combination with other active agents had a mortality rate of 19.3 % if the carbapenem MIC of the infecting organism was ≤8 μg/ml versus a mortality rate of 35.5 % if the MIC was >8 μg/ml.[42] An overall observation in this trial was that mortality was higher in patients who received monotherapy than in those treated with combination regimens. In a retrospective analysis of the data from this trial, the lowest mortality was observed in patients treated with a carbapenem-containing regimen. However, a recent analysis of 12 retrospective cohort studies or case series, two prospective observational studies, and two randomized controlled trials showed no difference in mortality between colistin alone and colistin/carbapenem combination therapy.[43] Several clinical trials are ongoing to further assess this issue ( NCT01732250 and NCT01597973). Until these data are available, a major challenge for the clinician will be to interpret existing data—most of which are observational or retrospective—and make clinical decisions regarding the therapy for such infections. Important considerations in those analyses will be the microbiologic data of the organism (including the specific carbapenem MICs), the timing of the administered antibiotics, the effectiveness of therapy based on pharmacokinetic/pharmacodynamic considerations, and the potential for additive or even synergistic effects with combination therapy.